Research ArticleA Comparative Study of Two-Compartment Exchange Models forDynamic Contrast-Enhanced MRI in Characterizing UterineCervical Carcinoma

Xue Wang1 Wenxiao Lin1 Yiting Mao2 Wenwen Peng1 Jiao Song1 Yi Lu1 Yu Zhao3

Tong San Koh 45 Zujun Hou6 and Zhihan Yan 1

1Department of Radiology e Second Aliated Hospital and Yuying Childrenrsquos Hospital of Wenzhou Medical University109 Xueyuanxi Road 325027 Wenzhou China2Wenzhou Medical University Gaojiaoyuan District 325027 Wenzhou China3Department of Gynaecology and Obstetrics e Second Aliated Hospital and Yuying Childrenrsquos Hospital of Wenzhou MedicalUniversity 109 Xueyuanxi Road 325027 Wenzhou China4Department of Oncologic Imaging National Cancer Center 169610 Singapore5Duke-NUS Graduate Medical School 169547 Singapore6Suzhou Institute of Biomedical Engineering and Technology Chinese Academy of Sciences Suzhou 25163 China

Correspondence should be addressed to Zhihan Yan yanzhihanwz163com

Received 25 June 2019 Accepted 14 October 2019 Published 7 November 2019

Academic Editor Alexander R Haug

Copyright copy 2019 Xue Wang et al is is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

A variety of tracer kinetic methods have been employed to assess tumor angiogenesis e Standard two-Compartmentmodel (SC) used in cervix carcinoma was less frequent and Adiabatic Approximation to the Tissue Homogeneity (AATH)and Distributed Parameter (DP) model are lacking is study compares two-compartment exchange models (2CXM)(AATH SC and DP) for determining dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) parameters incervical cancer with the aim of investigating the potential of various parameters derived from 2CXM for tumor diagnosisand exploring the possible relationship between these parameters in patients with cervix cancer Parameters (tissue bloodow Fp tissue blood volume Vp interstitial volume Ve and vascular permeability PS) for regions of interest (ROI) ofcervix lesions and normal cervix tissue were estimated by AATH SC and DP models in 36 patients with cervix cancer and17 healthy subjects All parameters showed signicant dierences between lesions and normal tissue with a P value lessthan 005 except for PS from the AATH model Fp from the SC model and Vp from the DP model Parameter Ve from theAATHmodel had the largest AUC (r 085) Parameters Fp and Vp from SC and DP models and Ve and PS from AATH andDP models were highly correlated respectively (r gt 08) in cervix lesions Cervix cancer was found to have a very unusualmicrocirculation pattern with over-growth of cancer cells but without evident development of angiogenesis Ve has thebest performance in identifying cervix cancer Most physiological parameters derived from AATH SC and DP models arelinearly correlated in cervix cancer

1 Introduction

Neovascularization plays a fundamental role in the growthof solid tumors [1] Various experimental models as well asclinicopathological observations have shown that solid tu-mors (eg breast [2] lung [3] and cervical carcinoma [4 5])

cannot attain diameters gt2-3mmwithout their own vascularsupply Dynamic contrast-enhanced magnetic resonanceimaging (DCE-MRI) is commonly used for the assessmentof tumor angiogenesis Analysis of DCE-MRI data can beperformed using tracer kinetic models to derive quantitativeparameters of tissue microcirculation A variety of tracer

HindawiContrast Media amp Molecular ImagingVolume 2019 Article ID 3168416 13 pageshttpsdoiorg10115520193168416

kinetic methods have been employed to characterize varioustumors and to assess the effects of antiangiogenic andantivascular drugs in clinical trials [6ndash15]

To date the Generalized Kinetic (GK or Tofts) modeland the Extended Generalized Kinetic (EGK or extendedTofts) model are frequently used for analysis of DCE-MRIdata in oncology and drug trials It is commonly assumedthat Ktrans yielded from the two models encompasses theeffects of both blood flow and vessel permeability such thatKtrans reflects permeability when blood flow is much largerthan permeability and Ktrans reflects blood flow when per-meability dominates [16] However novel vascular targetingdrugs could reduce both permeability and blood flow andmight exert different pharmacokinetic effects on blood flowand permeability [17 18]us a method that can separatelyestimate tissue blood (plasma) flow (Fp) and vessel per-meability (PS) is more valuable for the assessment of drugeffects Improvements in the temporal resolution of DCE-MRI sequences [19] have facilitated independent measure-ment of Fp and PS using the two-compartment exchangemodel (2CXM) A few such methods have been proposedwhich include the Standard two-Compartment model (SC)Adiabatic Approximation to the Tissue Homogeneity(AATH) and the two-compartment Distributed Parametermodel (DP) [20ndash23]

Several studies have applied these models in cervix carci-nomae relationship between parameters from various tracerkinetic methods [24 25] has been reported However thesestudies mainly focused on GK and EGK models and appli-cation of the SC model to cervix cancer was less frequentKallehauge et al found that Ktrans mainly reflects tissue bloodflow even though PSltFp which is contradictory with theunderstanding on Ktrans which will reflect flow when PS isinfinite [25] Donaldson et al observed that the correlationcoefficient of Ktrans with PS is low (r 031) while the corre-lation coefficient of Ktrans with Fp is high (r 094) [24] Veyielded from the EGK model was correlated with 2CXM butcorrelations between Vp from EGK and 2CXM were differentfor different studies [24 25] e application of AATH and DPmodels in cervix carcinoma is still lacking So parametervariations assessed by the two-compartment exchange models(AATH SC and DP) with respect to tumor microcirculationwould be clinically desirable Moreover to facilitate the con-sistent interpretation of treatment effects formulticenter clinicaltrials employing different tracer kinetic models it would be ofinterest to understand the relationships between parameters ofthese tracer kinetic models so that if one kinetic model is usedin a certain trial one could get a sense of what would be thelikely changes in another kineticmodel In this study three two-compartment models (AATH SC and DP) were applied incervix carcinoma to investigate the potential of various pa-rameters in tumor diagnosis and to examine the possible re-lationship between parameters of these tracer kinetic models

2 Materials and Methods

21 Patients is retrospective study was approved by theinstitutional research ethics review board and informedconsent was obtained from all patients Sixty-eight

consecutive female patients (mean age 504 years age range41ndash75 years) clinically suspected of cervix cancer presentedto our department in the period of April 2016 to July 2018Patients were excluded for the following reasons (1) poorimage quality of DCE-MRI such as significant motion ar-tifacts (n 1) or incomplete images (n 2) (2) patients witha history of targeted chemotherapy or radiation therapybefore examination (n 4) (3) patients diagnosed of othercervical lesions such as submucous myoma of uterus (n 2)and the endometrial carcinoma (n 1) and (4) no mass wasidentified for patients with stage Ia on DCE and other MRIsequences (n 5) In the end 36 patients with cervix cancerand 17 healthy subjects were included in this retrospectivestudy Cervix cancer was clinically staged according to theInternational Federation of Gynecology and Obstetricsclassifications [26] ese 36 patients were classified intostage Ib (n 17) IIa (n 14) and IIb (n 5) All patientswere confirmed histopathologically Hysterectomy wasperformed for cervix cancer patients and biopsy for healthysubjects Patient characteristics are summarized in Table 1

22 MR Imaging Protocol All scans were performed on a 30Tscanner (DiscoverytradeMR750w General Electric USA) usingan 8-channel torso phased-array coil Routine clinical MRIscan sequences included a transverse fast spin-echo T1-weighted sequence (repetition timeecho time 550ndash7007ndash10ms field of view 320times 320mm matrix size 512times 512slice thickness 50mm intersection gap 6mm) a shorttime inversion recovery (STIR) T2-weighted sequence (rep-etition timeecho time 3000ndash400070ndash80ms field ofview 256ndash320times 256ndash320mm matrix size 512times 512 slicethickness 50mm intersection gap 6ndash7mm) and diffu-sion-weighted imaging (DWI) (repetition timeecho time- 240060ms field of view 320times 320mm matrixsize 256times 256mm slice thickness 50mm intersectiongap 7mm b value 0 1000 smm2)

DCE-MRI was performed using a three-dimensional T1-weighted spoiled gradient echo sequence (LAVA repetitiontimeecho time 31ms flip angle 4deg 8deg and 11deg field ofview 360times 360mm matrix size 256times 256 slice thick-ness 5mm 6 slices per slab) Ten precontrast scans of eachflip angle (4deg 8deg and 11deg) were acquired in the axial planeunder quiet respiration Dynamic postcontrast scans wereacquired using the same sequence and a flip angle of 11degwith the intravenous injection of gadopentetated imeglu-mine (Magnevist Bayer Healthcare Pharmaceuticals IncNJ) at an injection rate of 2mLsecond standard dose of01mmolkg A total of 180 consecutive scans were acquiredfor the dynamic series with a temporal resolution 2 sSubsequently a routine late contrast-enhanced T1-wieghtedscan (repetition timeecho time 42ms flip angle 13degfield of view 280times 280mm matrix size 512times 512 slicethickness 3mm) was acquired in the sagittal plane

23 Tracer Kinetic Models Details of the four tracer kineticmodels used in this study can be found in several reviewpapers [21 22] Here we would only list the essential op-erational equations for these models which specify the

2 Contrast Media amp Molecular Imaging

dependence of tissue tracer concentration Ctiss(t) (as afunction of time t) on the arterial input function (AIF) andrelevant physiological parameters

Standard Compartment (SC) model

Ctiss(t) AIFotimesFp[A exp(α t) +(1 minus A)exp(β t)] (1a)

where

α

β⎛⎝ ⎞⎠

12

⎡⎣ minusPSVp

+PSVe

+Fp

Vp1113888 1113889

plusmn

PSVp

+PSVe

+Fp

Vp1113888 1113889

2

minus 4PSVe

Fp

Vp

11139741113972

⎤⎦

(1b)

A α + PSVp1113872 1113873 + PSVe( 1113857

α minus β (1c)

and otimes denotes the convolution operatorDistributed Parameter (DP) model

Ctiss(t) AIFotimes

Fp u(t) minus u t minusVp

Fp1113888 1113889 + u t minus

Vp

Fp1113888 1113889 1 minus exp minus

PSFp

1113888 1113889 1 + 1113946tminus VpFp( )

0exp minus

PSVe

τ1113888 1113889

PSVe

PSFp

1τ

1113971

I1 2

PSVe

PSFp

τ

1113971

⎛⎝ ⎞⎠dτ⎡⎢⎢⎣ ⎤⎥⎥⎦⎧⎨

⎩

⎫⎬

⎭

⎧⎨

⎩

⎫⎬

⎭

(2)

where u(t) denotes the Heaviside unit-step functionand I1 is the modified Bessel functionAdiabatic Approximation to the Tissue Homogeneity(AATH) model

Ctiss(t) AIFotimes

Fp 1 minus exp minusPSFp

1113888 11138891113890 1113891exp minusFp

Ve1 minus exp minus

PSFp

1113888 11138891113890 1113891 t minusFp

Vp1113888 11138891113896 11138971113896 1113897

(3)

Interested readers can refer to recent review papers[21 22] for more details of the three tracer kinetic models

24 Image Postprocessing To avoid possible effects of inflowand inhomogeneity near boundaries only the central fourslices from the imaging volume (of 6 slices) were selected forprocessing For each patient regions of interest (ROIs) fortumor lesions and normal cervix tissue were manually de-lineated on the central four slices (as illustrated inFigure 1(a)) by two experienced radiologists with more than10 years of experience in gynecological radiology RoutineT1-weighted T2-weighted and DW images were used forcross-referencing to confirm the location and size of the

lesions when contrast-enhanced scans were evaluated esize of ROI is no less than 10 voxels to ensure the robustnessof measurement e normal ROIs were selected in thenormal cervical tissue away from the lesions e areas ofnecrotic cystic and hemorrhages were avoided whendrawing the lesion ROIs Finally 107 ROIs for cervix cancerwere obtained from 36 patients 103 ROIs for normal tissueswere obtained from 17 healthy subjects and 24 patients withsmaller masses which can be delineated accurately AIF wassampled from a voxel that clearly resided within the iliacartery on one of the four central slices as shown inFigure 1(b) Desirable features for AIF selection included anearly bolus arrival time and high peak value and signal-to-noise ratio Fitting of voxel-level tissue concentration-timecurves Ctiss(t) and generation of parametric maps wereperformed using a commercially available software (MIta-lytics Fitpu Healthcare Singapore) e software allows forthe selection of an individual AIF for each patient case andemploys a constrained nonlinear optimization algorithm infitting the various models

25 Statistical Analysis For each patient the median pa-rameter value of all voxels within the tumor ROIs onmultiple slices is taken as a representative statistic of theparameter in the tumor e median values of the fitted

Table 1 Patient characteristics

Parameters Noof patientsAge mean (range) years 42 (42ndash75)Histological subtypeAdenocarcinoma (AC) 2Squamous cell carcinoma (SCC) 34

Tumor gradeWell 6Moderate 30

FIGOa stageIb 17IIa 14IIb 5

Abbreviation FIGO International Federation of Gynecology and Obstet-rics aAccording to FIGO 2009 staging criteria

Contrast Media amp Molecular Imaging 3

(a)

200

150

100

50

00 50 100

Mea

n co

ncen

trat

ion

(10ndash1

middot 1 s)

150 200Time (s)

250 300 350 400

(b)

Figure 1 Continued

4 Contrast Media amp Molecular Imaging

AATH-Fp (mLmin100mL)40

20

0

35

175

0

23

115

0

SC-Fp (mLmin100mL)

DP-Fp (mLmin100mL)

5

25

0

25

125

0

20

10

0

AATH-Vp (mL100mL)

SC-Vp (mL100mL)

DP-Vp (mL100mL)

55

275

0

55

275

0

55

275

0

AATH-Ve (mL100mL)

SC-Ve (mL100mL)

DP-Ve (mL100mL)

30

15

0

35

175

0

20

10

0

AATH-PS (mLmin100mL)

SC-PS (mLmin100mL)

DP-PS (mLmin100mL)

(c)

0

5

45

4

35

3

25

2

15

1

05

ndash05

0

50 100 150 200Time (s)

250 300 350

Con

cent

ratio

n (times

10ndash1

middotsndash1)

DataAATH [233 30 64 183]

SC [342 34 66 182]DP [117 30 74 178]

(d)

Figure 1 Example of a patient with stage IIb cervix cancer (a) ROIs for cervix carcinoma (blue) and normal cervix tissue (red) are shown forthe central four slices of the DCE-MRI dataset and the location within the iliac artery where the AIF was sampled was marked with a red dot(b) Sampled AIF used in model fitting (c) Parameter maps generated using the three models (AATH SC and DP) for tumor and the normaltissue ROIs (d) Examples of curve fittings for a tumor voxel In the legend the four numbers within square brackets beside each model aretheir respective parameter values (Fp (mLmin100mL) Vp (mL100mL) PS (mLmin100mL) Ve (mL100mL))

Contrast Media amp Molecular Imaging 5

parameters were used because the median is more robust tooutliers (that could occur during data fitting) than the meanA two-way model average measure intraclass correlationcoefficient (ICC) was used to test the interobserver con-sistency en all the measurements from the two observerswere averaged for further comparison Agreement wasinterpreted according to the ICC as follows gt080 excellent061ndash080 good 041ndash060 moderate 021ndash040 fair andlt02 poor agreement [27]

e normality of the distribution of all parameters wasanalysed by the KolmogorovndashSmirnov test e receiveroperating characteristic (ROC) curves of all parameters wereobtained and the areas under the curves (AUC) wereevaluated to determine the discriminating power of DCEparameters between the lesion and normal tissues In-terpretation of AUC values is application dependent and ingeneral it is appropriate that values ge09 would be ldquoexcel-lentrdquo ge08 ldquogoodrdquo ge07 ldquofairrdquo and lt07 ldquopoorrdquo [28]

e Pearson correlation coefficient r was used to explorepossible relationship between median parameter values ofthe three models A strong correlation was assumed for08lt rle 1 a moderate correlation for 05lt rle 08 a weakcorrelation for 03lt rle 05 and no correlation for rle 03[29] e BlandndashAltman plot was used to show agreementsbetween parameter values of the three models All statisticalanalysis were performed using SPSS software 180 (ChicagoIL USA) and Plt 005 was considered statisticallysignificant

3 Results

For the case shown in Figure 1(a) the parameter maps of oneslice generated using the three methods (AATH SC andDP) are shown in Figure 1(c) It is evident that threemethods attain similar results in Ve and PS and both es-timates are smaller in tumor ROI than in normal tissue ROIFor Fp the estimate by AATH is smaller in tumor ROI andthe estimates by SC and DP are apparently close betweentumor and normal tissue ROI For Vp the estimates varyfrom close between tumor and normal tissue ROI (AATH)to smaller in tumor ROI (SC and DP)

Table 2 shows the ICC values for the measured pa-rameters where most ICC values are greater than 09 in-dicating very good agreement between measurements fromtwo observers us the parameter values as measured bytwo observers are averaged and utilized in the analysis asfollows

Table 3 shows the nonparametric statistical analysis ofparameters between cervix cancer and normal tissue ROIse parameters Fp Ve and PS of the three DCE models(AATH SC and DP models) were smaller in cervix cancerlesions than in normal cervix tissue and the differencesbetween lesions and normal tissue were mostly significantexcept for PS from the AATH model and Fp from the SCmodelVp by AATH and SC was significantly larger in cervixcancer lesions than in normal cervix tissue and DP yieldedcontradictory results though the difference between lesionsand normal tissue was not statistically significant ParameterVe attained fair to good AUC values (gt083 for AATH and

DP models and gt075 for the SC model) Other parameterseg Vp from the AATH model and PS from the SC modelalso showed fair AUC values

Results of the Pearson correlation between the sameparameters estimated by different models are shown inTable 4 Comparing parameter estimates of the three models(AATH SC and DPmodels) for cervix cancer lesions it wasfound that Fp and Vp from SC and DPmodels and Ve and PSfrom AATH and DP models were highly correlated re-spectively (rgt 08) Moderate correlations were observedbetween Vp from AATH and Vp from SC and DPmodels Veand PS from AATH and SC models and Ve and PS from SCand DP models (05ltrlt08) No correlation was observedbetween Fp from AATH and Fp from SC and DP modelsrespectively (rlt 03) Comparing parameter estimates of thethree models (AATH SC and DPmodels) for normal cervixtissues it was found that good agreement existed betweenthe estimates of Fp Ve from the three models and PS fromAATH and DP models (rgt 08) Moderate correlations wereobserved between PS from AATH and SC models and Vpfrom SC and DP models (05lt rlt 08) A weak correlationwas observed between PS from SC and DP models(03lt rlt 05) No correlation was observed between Vp fromAATH and Vp from SC and DP models respectively(rlt 03) Results of the Bland and Altman test for com-parison of the parameter with the same biophysical meaningobtained using the three models are shown in Figure 2BlandndashAltman plots demonstrated good agreements be-tween the estimates of Ve from the three models for bothnormal cervix tissue and cervix lesion because their y-co-ordinate values were centered at zero (Figures 2(a) and 2(b))Good agreements were observed between Vp from SC andDP models and PS from AATH and SC models in normalcervix tissue (Figure 2(a)) e y-coordinate value of otherparameters estimated by AATH SC and DP modes stayedaway from zero indicating less agreement between them innormal cervix tissue and cervix lesion

4 Discussion

is preliminary study evaluated the performance ofphysiological parameters derived from AATH SC and DPmodels with respect to tumor microcirculation in cervixcancer e angiogenic activity of cervix cancer assessed byvarious DCE-MRI models improved in vivo understandingof the fundamental processes involved in tumor angio-genesis Parameter Ve has the best performance in identi-fying cervix cancer among all parameters Mostphysiological parameters derived from the three DCE-MRImodels are linearly correlated in cervix cancer

e parameter Ve of all three DCE models showed thatthe Ve value of cervix cancer tissue was significantly smallerthan that of normal cervix tissue e corresponding AUCvalues were greater than 075 and Ve of the AATH modelhad the largest AUC (085) among all parameters indicatinga fairly good performance in terms of diagnostic value Intracer kinetic modeling Ve stands for the fractional volumeof extravascular extracellular space which is closely per-taining to the degree of cell growth e more the cells grow

6 Contrast Media amp Molecular Imaging

Tabl

e2

Interobserverconsistency

forcervix

lesio

nandtheno

rmal

cervix

tissue

AATH

SCDP

F pVp

Ve

PSF p

Vp

Ve

PSF p

Vp

Ve

PSLesio

nIC

C0922

0975

0959

0939

0935

0980

0968

0857

0965

0989

0958

0958

95

CI

0886ndash

0947

0963ndash

0983

0940ndash

0972

0911ndash

0958

0905ndash

0956

0971ndash

0987

0953ndash

0978

0790ndash

0903

0949ndash

0976

0984ndash

0993

0938ndash

0971

0938ndash

0971

Normal

ICC

0934

0661

0908

0948

0978

0945

0910

0887

0906

0812

0943

0892

95

CI

0902ndash

0955

0499ndash

0770

0864ndash

0938

0924ndash

0965

0967ndash

0985

0919ndash

0963

0867ndash

0939

0833ndash

0923

0861ndash

0936

0702ndash

0882

0916ndash

0961

0841ndash

0927

ICCintraclasscorrelationcoeffi

cientCIconfi

denceinterval

ofdifference

Contrast Media amp Molecular Imaging 7

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

dependence of tissue tracer concentration Ctiss(t) (as afunction of time t) on the arterial input function (AIF) andrelevant physiological parameters

Standard Compartment (SC) model

Ctiss(t) AIFotimesFp[A exp(α t) +(1 minus A)exp(β t)] (1a)

where

α

β⎛⎝ ⎞⎠

12

⎡⎣ minusPSVp

+PSVe

+Fp

Vp1113888 1113889

plusmn

PSVp

+PSVe

+Fp

Vp1113888 1113889

2

minus 4PSVe

Fp

Vp

11139741113972

⎤⎦

(1b)

A α + PSVp1113872 1113873 + PSVe( 1113857

α minus β (1c)

and otimes denotes the convolution operatorDistributed Parameter (DP) model

Ctiss(t) AIFotimes

Fp u(t) minus u t minusVp

Fp1113888 1113889 + u t minus

Vp

Fp1113888 1113889 1 minus exp minus

PSFp

1113888 1113889 1 + 1113946tminus VpFp( )

0exp minus

PSVe

τ1113888 1113889

PSVe

PSFp

1τ

1113971

I1 2

PSVe

PSFp

τ

1113971

⎛⎝ ⎞⎠dτ⎡⎢⎢⎣ ⎤⎥⎥⎦⎧⎨

⎩

⎫⎬

⎭

⎧⎨

⎩

⎫⎬

⎭

(2)

where u(t) denotes the Heaviside unit-step functionand I1 is the modified Bessel functionAdiabatic Approximation to the Tissue Homogeneity(AATH) model

Ctiss(t) AIFotimes

Fp 1 minus exp minusPSFp

1113888 11138891113890 1113891exp minusFp

Ve1 minus exp minus

PSFp

1113888 11138891113890 1113891 t minusFp

Vp1113888 11138891113896 11138971113896 1113897

(3)

Interested readers can refer to recent review papers[21 22] for more details of the three tracer kinetic models

24 Image Postprocessing To avoid possible effects of inflowand inhomogeneity near boundaries only the central fourslices from the imaging volume (of 6 slices) were selected forprocessing For each patient regions of interest (ROIs) fortumor lesions and normal cervix tissue were manually de-lineated on the central four slices (as illustrated inFigure 1(a)) by two experienced radiologists with more than10 years of experience in gynecological radiology RoutineT1-weighted T2-weighted and DW images were used forcross-referencing to confirm the location and size of the

lesions when contrast-enhanced scans were evaluated esize of ROI is no less than 10 voxels to ensure the robustnessof measurement e normal ROIs were selected in thenormal cervical tissue away from the lesions e areas ofnecrotic cystic and hemorrhages were avoided whendrawing the lesion ROIs Finally 107 ROIs for cervix cancerwere obtained from 36 patients 103 ROIs for normal tissueswere obtained from 17 healthy subjects and 24 patients withsmaller masses which can be delineated accurately AIF wassampled from a voxel that clearly resided within the iliacartery on one of the four central slices as shown inFigure 1(b) Desirable features for AIF selection included anearly bolus arrival time and high peak value and signal-to-noise ratio Fitting of voxel-level tissue concentration-timecurves Ctiss(t) and generation of parametric maps wereperformed using a commercially available software (MIta-lytics Fitpu Healthcare Singapore) e software allows forthe selection of an individual AIF for each patient case andemploys a constrained nonlinear optimization algorithm infitting the various models

25 Statistical Analysis For each patient the median pa-rameter value of all voxels within the tumor ROIs onmultiple slices is taken as a representative statistic of theparameter in the tumor e median values of the fitted

Table 1 Patient characteristics

Parameters Noof patientsAge mean (range) years 42 (42ndash75)Histological subtypeAdenocarcinoma (AC) 2Squamous cell carcinoma (SCC) 34

Tumor gradeWell 6Moderate 30

FIGOa stageIb 17IIa 14IIb 5

Abbreviation FIGO International Federation of Gynecology and Obstet-rics aAccording to FIGO 2009 staging criteria

Contrast Media amp Molecular Imaging 3

(a)

200

150

100

50

00 50 100

Mea

n co

ncen

trat

ion

(10ndash1

middot 1 s)

150 200Time (s)

250 300 350 400

(b)

Figure 1 Continued

4 Contrast Media amp Molecular Imaging

AATH-Fp (mLmin100mL)40

20

0

35

175

0

23

115

0

SC-Fp (mLmin100mL)

DP-Fp (mLmin100mL)

5

25

0

25

125

0

20

10

0

AATH-Vp (mL100mL)

SC-Vp (mL100mL)

DP-Vp (mL100mL)

55

275

0

55

275

0

55

275

0

AATH-Ve (mL100mL)

SC-Ve (mL100mL)

DP-Ve (mL100mL)

30

15

0

35

175

0

20

10

0

AATH-PS (mLmin100mL)

SC-PS (mLmin100mL)

DP-PS (mLmin100mL)

(c)

0

5

45

4

35

3

25

2

15

1

05

ndash05

0

50 100 150 200Time (s)

250 300 350

Con

cent

ratio

n (times

10ndash1

middotsndash1)

DataAATH [233 30 64 183]

SC [342 34 66 182]DP [117 30 74 178]

(d)

Figure 1 Example of a patient with stage IIb cervix cancer (a) ROIs for cervix carcinoma (blue) and normal cervix tissue (red) are shown forthe central four slices of the DCE-MRI dataset and the location within the iliac artery where the AIF was sampled was marked with a red dot(b) Sampled AIF used in model fitting (c) Parameter maps generated using the three models (AATH SC and DP) for tumor and the normaltissue ROIs (d) Examples of curve fittings for a tumor voxel In the legend the four numbers within square brackets beside each model aretheir respective parameter values (Fp (mLmin100mL) Vp (mL100mL) PS (mLmin100mL) Ve (mL100mL))

Contrast Media amp Molecular Imaging 5

parameters were used because the median is more robust tooutliers (that could occur during data fitting) than the meanA two-way model average measure intraclass correlationcoefficient (ICC) was used to test the interobserver con-sistency en all the measurements from the two observerswere averaged for further comparison Agreement wasinterpreted according to the ICC as follows gt080 excellent061ndash080 good 041ndash060 moderate 021ndash040 fair andlt02 poor agreement [27]

e normality of the distribution of all parameters wasanalysed by the KolmogorovndashSmirnov test e receiveroperating characteristic (ROC) curves of all parameters wereobtained and the areas under the curves (AUC) wereevaluated to determine the discriminating power of DCEparameters between the lesion and normal tissues In-terpretation of AUC values is application dependent and ingeneral it is appropriate that values ge09 would be ldquoexcel-lentrdquo ge08 ldquogoodrdquo ge07 ldquofairrdquo and lt07 ldquopoorrdquo [28]

e Pearson correlation coefficient r was used to explorepossible relationship between median parameter values ofthe three models A strong correlation was assumed for08lt rle 1 a moderate correlation for 05lt rle 08 a weakcorrelation for 03lt rle 05 and no correlation for rle 03[29] e BlandndashAltman plot was used to show agreementsbetween parameter values of the three models All statisticalanalysis were performed using SPSS software 180 (ChicagoIL USA) and Plt 005 was considered statisticallysignificant

3 Results

For the case shown in Figure 1(a) the parameter maps of oneslice generated using the three methods (AATH SC andDP) are shown in Figure 1(c) It is evident that threemethods attain similar results in Ve and PS and both es-timates are smaller in tumor ROI than in normal tissue ROIFor Fp the estimate by AATH is smaller in tumor ROI andthe estimates by SC and DP are apparently close betweentumor and normal tissue ROI For Vp the estimates varyfrom close between tumor and normal tissue ROI (AATH)to smaller in tumor ROI (SC and DP)

Table 2 shows the ICC values for the measured pa-rameters where most ICC values are greater than 09 in-dicating very good agreement between measurements fromtwo observers us the parameter values as measured bytwo observers are averaged and utilized in the analysis asfollows

Table 3 shows the nonparametric statistical analysis ofparameters between cervix cancer and normal tissue ROIse parameters Fp Ve and PS of the three DCE models(AATH SC and DP models) were smaller in cervix cancerlesions than in normal cervix tissue and the differencesbetween lesions and normal tissue were mostly significantexcept for PS from the AATH model and Fp from the SCmodelVp by AATH and SC was significantly larger in cervixcancer lesions than in normal cervix tissue and DP yieldedcontradictory results though the difference between lesionsand normal tissue was not statistically significant ParameterVe attained fair to good AUC values (gt083 for AATH and

DP models and gt075 for the SC model) Other parameterseg Vp from the AATH model and PS from the SC modelalso showed fair AUC values

Results of the Pearson correlation between the sameparameters estimated by different models are shown inTable 4 Comparing parameter estimates of the three models(AATH SC and DPmodels) for cervix cancer lesions it wasfound that Fp and Vp from SC and DPmodels and Ve and PSfrom AATH and DP models were highly correlated re-spectively (rgt 08) Moderate correlations were observedbetween Vp from AATH and Vp from SC and DPmodels Veand PS from AATH and SC models and Ve and PS from SCand DP models (05ltrlt08) No correlation was observedbetween Fp from AATH and Fp from SC and DP modelsrespectively (rlt 03) Comparing parameter estimates of thethree models (AATH SC and DPmodels) for normal cervixtissues it was found that good agreement existed betweenthe estimates of Fp Ve from the three models and PS fromAATH and DP models (rgt 08) Moderate correlations wereobserved between PS from AATH and SC models and Vpfrom SC and DP models (05lt rlt 08) A weak correlationwas observed between PS from SC and DP models(03lt rlt 05) No correlation was observed between Vp fromAATH and Vp from SC and DP models respectively(rlt 03) Results of the Bland and Altman test for com-parison of the parameter with the same biophysical meaningobtained using the three models are shown in Figure 2BlandndashAltman plots demonstrated good agreements be-tween the estimates of Ve from the three models for bothnormal cervix tissue and cervix lesion because their y-co-ordinate values were centered at zero (Figures 2(a) and 2(b))Good agreements were observed between Vp from SC andDP models and PS from AATH and SC models in normalcervix tissue (Figure 2(a)) e y-coordinate value of otherparameters estimated by AATH SC and DP modes stayedaway from zero indicating less agreement between them innormal cervix tissue and cervix lesion

4 Discussion

is preliminary study evaluated the performance ofphysiological parameters derived from AATH SC and DPmodels with respect to tumor microcirculation in cervixcancer e angiogenic activity of cervix cancer assessed byvarious DCE-MRI models improved in vivo understandingof the fundamental processes involved in tumor angio-genesis Parameter Ve has the best performance in identi-fying cervix cancer among all parameters Mostphysiological parameters derived from the three DCE-MRImodels are linearly correlated in cervix cancer

e parameter Ve of all three DCE models showed thatthe Ve value of cervix cancer tissue was significantly smallerthan that of normal cervix tissue e corresponding AUCvalues were greater than 075 and Ve of the AATH modelhad the largest AUC (085) among all parameters indicatinga fairly good performance in terms of diagnostic value Intracer kinetic modeling Ve stands for the fractional volumeof extravascular extracellular space which is closely per-taining to the degree of cell growth e more the cells grow

6 Contrast Media amp Molecular Imaging

Tabl

e2

Interobserverconsistency

forcervix

lesio

nandtheno

rmal

cervix

tissue

AATH

SCDP

F pVp

Ve

PSF p

Vp

Ve

PSF p

Vp

Ve

PSLesio

nIC

C0922

0975

0959

0939

0935

0980

0968

0857

0965

0989

0958

0958

95

CI

0886ndash

0947

0963ndash

0983

0940ndash

0972

0911ndash

0958

0905ndash

0956

0971ndash

0987

0953ndash

0978

0790ndash

0903

0949ndash

0976

0984ndash

0993

0938ndash

0971

0938ndash

0971

Normal

ICC

0934

0661

0908

0948

0978

0945

0910

0887

0906

0812

0943

0892

95

CI

0902ndash

0955

0499ndash

0770

0864ndash

0938

0924ndash

0965

0967ndash

0985

0919ndash

0963

0867ndash

0939

0833ndash

0923

0861ndash

0936

0702ndash

0882

0916ndash

0961

0841ndash

0927

ICCintraclasscorrelationcoeffi

cientCIconfi

denceinterval

ofdifference

Contrast Media amp Molecular Imaging 7

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

(a)

200

150

100

50

00 50 100

Mea

n co

ncen

trat

ion

(10ndash1

middot 1 s)

150 200Time (s)

250 300 350 400

(b)

Figure 1 Continued

4 Contrast Media amp Molecular Imaging

AATH-Fp (mLmin100mL)40

20

0

35

175

0

23

115

0

SC-Fp (mLmin100mL)

DP-Fp (mLmin100mL)

5

25

0

25

125

0

20

10

0

AATH-Vp (mL100mL)

SC-Vp (mL100mL)

DP-Vp (mL100mL)

55

275

0

55

275

0

55

275

0

AATH-Ve (mL100mL)

SC-Ve (mL100mL)

DP-Ve (mL100mL)

30

15

0

35

175

0

20

10

0

AATH-PS (mLmin100mL)

SC-PS (mLmin100mL)

DP-PS (mLmin100mL)

(c)

0

5

45

4

35

3

25

2

15

1

05

ndash05

0

50 100 150 200Time (s)

250 300 350

Con

cent

ratio

n (times

10ndash1

middotsndash1)

DataAATH [233 30 64 183]

SC [342 34 66 182]DP [117 30 74 178]

(d)

Figure 1 Example of a patient with stage IIb cervix cancer (a) ROIs for cervix carcinoma (blue) and normal cervix tissue (red) are shown forthe central four slices of the DCE-MRI dataset and the location within the iliac artery where the AIF was sampled was marked with a red dot(b) Sampled AIF used in model fitting (c) Parameter maps generated using the three models (AATH SC and DP) for tumor and the normaltissue ROIs (d) Examples of curve fittings for a tumor voxel In the legend the four numbers within square brackets beside each model aretheir respective parameter values (Fp (mLmin100mL) Vp (mL100mL) PS (mLmin100mL) Ve (mL100mL))

Contrast Media amp Molecular Imaging 5

parameters were used because the median is more robust tooutliers (that could occur during data fitting) than the meanA two-way model average measure intraclass correlationcoefficient (ICC) was used to test the interobserver con-sistency en all the measurements from the two observerswere averaged for further comparison Agreement wasinterpreted according to the ICC as follows gt080 excellent061ndash080 good 041ndash060 moderate 021ndash040 fair andlt02 poor agreement [27]

e normality of the distribution of all parameters wasanalysed by the KolmogorovndashSmirnov test e receiveroperating characteristic (ROC) curves of all parameters wereobtained and the areas under the curves (AUC) wereevaluated to determine the discriminating power of DCEparameters between the lesion and normal tissues In-terpretation of AUC values is application dependent and ingeneral it is appropriate that values ge09 would be ldquoexcel-lentrdquo ge08 ldquogoodrdquo ge07 ldquofairrdquo and lt07 ldquopoorrdquo [28]

e Pearson correlation coefficient r was used to explorepossible relationship between median parameter values ofthe three models A strong correlation was assumed for08lt rle 1 a moderate correlation for 05lt rle 08 a weakcorrelation for 03lt rle 05 and no correlation for rle 03[29] e BlandndashAltman plot was used to show agreementsbetween parameter values of the three models All statisticalanalysis were performed using SPSS software 180 (ChicagoIL USA) and Plt 005 was considered statisticallysignificant

3 Results

For the case shown in Figure 1(a) the parameter maps of oneslice generated using the three methods (AATH SC andDP) are shown in Figure 1(c) It is evident that threemethods attain similar results in Ve and PS and both es-timates are smaller in tumor ROI than in normal tissue ROIFor Fp the estimate by AATH is smaller in tumor ROI andthe estimates by SC and DP are apparently close betweentumor and normal tissue ROI For Vp the estimates varyfrom close between tumor and normal tissue ROI (AATH)to smaller in tumor ROI (SC and DP)

Table 2 shows the ICC values for the measured pa-rameters where most ICC values are greater than 09 in-dicating very good agreement between measurements fromtwo observers us the parameter values as measured bytwo observers are averaged and utilized in the analysis asfollows

Table 3 shows the nonparametric statistical analysis ofparameters between cervix cancer and normal tissue ROIse parameters Fp Ve and PS of the three DCE models(AATH SC and DP models) were smaller in cervix cancerlesions than in normal cervix tissue and the differencesbetween lesions and normal tissue were mostly significantexcept for PS from the AATH model and Fp from the SCmodelVp by AATH and SC was significantly larger in cervixcancer lesions than in normal cervix tissue and DP yieldedcontradictory results though the difference between lesionsand normal tissue was not statistically significant ParameterVe attained fair to good AUC values (gt083 for AATH and

DP models and gt075 for the SC model) Other parameterseg Vp from the AATH model and PS from the SC modelalso showed fair AUC values

Results of the Pearson correlation between the sameparameters estimated by different models are shown inTable 4 Comparing parameter estimates of the three models(AATH SC and DPmodels) for cervix cancer lesions it wasfound that Fp and Vp from SC and DPmodels and Ve and PSfrom AATH and DP models were highly correlated re-spectively (rgt 08) Moderate correlations were observedbetween Vp from AATH and Vp from SC and DPmodels Veand PS from AATH and SC models and Ve and PS from SCand DP models (05ltrlt08) No correlation was observedbetween Fp from AATH and Fp from SC and DP modelsrespectively (rlt 03) Comparing parameter estimates of thethree models (AATH SC and DPmodels) for normal cervixtissues it was found that good agreement existed betweenthe estimates of Fp Ve from the three models and PS fromAATH and DP models (rgt 08) Moderate correlations wereobserved between PS from AATH and SC models and Vpfrom SC and DP models (05lt rlt 08) A weak correlationwas observed between PS from SC and DP models(03lt rlt 05) No correlation was observed between Vp fromAATH and Vp from SC and DP models respectively(rlt 03) Results of the Bland and Altman test for com-parison of the parameter with the same biophysical meaningobtained using the three models are shown in Figure 2BlandndashAltman plots demonstrated good agreements be-tween the estimates of Ve from the three models for bothnormal cervix tissue and cervix lesion because their y-co-ordinate values were centered at zero (Figures 2(a) and 2(b))Good agreements were observed between Vp from SC andDP models and PS from AATH and SC models in normalcervix tissue (Figure 2(a)) e y-coordinate value of otherparameters estimated by AATH SC and DP modes stayedaway from zero indicating less agreement between them innormal cervix tissue and cervix lesion

4 Discussion

is preliminary study evaluated the performance ofphysiological parameters derived from AATH SC and DPmodels with respect to tumor microcirculation in cervixcancer e angiogenic activity of cervix cancer assessed byvarious DCE-MRI models improved in vivo understandingof the fundamental processes involved in tumor angio-genesis Parameter Ve has the best performance in identi-fying cervix cancer among all parameters Mostphysiological parameters derived from the three DCE-MRImodels are linearly correlated in cervix cancer

e parameter Ve of all three DCE models showed thatthe Ve value of cervix cancer tissue was significantly smallerthan that of normal cervix tissue e corresponding AUCvalues were greater than 075 and Ve of the AATH modelhad the largest AUC (085) among all parameters indicatinga fairly good performance in terms of diagnostic value Intracer kinetic modeling Ve stands for the fractional volumeof extravascular extracellular space which is closely per-taining to the degree of cell growth e more the cells grow

6 Contrast Media amp Molecular Imaging

Tabl

e2

Interobserverconsistency

forcervix

lesio

nandtheno

rmal

cervix

tissue

AATH

SCDP

F pVp

Ve

PSF p

Vp

Ve

PSF p

Vp

Ve

PSLesio

nIC

C0922

0975

0959

0939

0935

0980

0968

0857

0965

0989

0958

0958

95

CI

0886ndash

0947

0963ndash

0983

0940ndash

0972

0911ndash

0958

0905ndash

0956

0971ndash

0987

0953ndash

0978

0790ndash

0903

0949ndash

0976

0984ndash

0993

0938ndash

0971

0938ndash

0971

Normal

ICC

0934

0661

0908

0948

0978

0945

0910

0887

0906

0812

0943

0892

95

CI

0902ndash

0955

0499ndash

0770

0864ndash

0938

0924ndash

0965

0967ndash

0985

0919ndash

0963

0867ndash

0939

0833ndash

0923

0861ndash

0936

0702ndash

0882

0916ndash

0961

0841ndash

0927

ICCintraclasscorrelationcoeffi

cientCIconfi

denceinterval

ofdifference

Contrast Media amp Molecular Imaging 7

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

AATH-Fp (mLmin100mL)40

20

0

35

175

0

23

115

0

SC-Fp (mLmin100mL)

DP-Fp (mLmin100mL)

5

25

0

25

125

0

20

10

0

AATH-Vp (mL100mL)

SC-Vp (mL100mL)

DP-Vp (mL100mL)

55

275

0

55

275

0

55

275

0

AATH-Ve (mL100mL)

SC-Ve (mL100mL)

DP-Ve (mL100mL)

30

15

0

35

175

0

20

10

0

AATH-PS (mLmin100mL)

SC-PS (mLmin100mL)

DP-PS (mLmin100mL)

(c)

0

5

45

4

35

3

25

2

15

1

05

ndash05

0

50 100 150 200Time (s)

250 300 350

Con

cent

ratio

n (times

10ndash1

middotsndash1)

DataAATH [233 30 64 183]

SC [342 34 66 182]DP [117 30 74 178]

(d)

Figure 1 Example of a patient with stage IIb cervix cancer (a) ROIs for cervix carcinoma (blue) and normal cervix tissue (red) are shown forthe central four slices of the DCE-MRI dataset and the location within the iliac artery where the AIF was sampled was marked with a red dot(b) Sampled AIF used in model fitting (c) Parameter maps generated using the three models (AATH SC and DP) for tumor and the normaltissue ROIs (d) Examples of curve fittings for a tumor voxel In the legend the four numbers within square brackets beside each model aretheir respective parameter values (Fp (mLmin100mL) Vp (mL100mL) PS (mLmin100mL) Ve (mL100mL))

Contrast Media amp Molecular Imaging 5

parameters were used because the median is more robust tooutliers (that could occur during data fitting) than the meanA two-way model average measure intraclass correlationcoefficient (ICC) was used to test the interobserver con-sistency en all the measurements from the two observerswere averaged for further comparison Agreement wasinterpreted according to the ICC as follows gt080 excellent061ndash080 good 041ndash060 moderate 021ndash040 fair andlt02 poor agreement [27]

e normality of the distribution of all parameters wasanalysed by the KolmogorovndashSmirnov test e receiveroperating characteristic (ROC) curves of all parameters wereobtained and the areas under the curves (AUC) wereevaluated to determine the discriminating power of DCEparameters between the lesion and normal tissues In-terpretation of AUC values is application dependent and ingeneral it is appropriate that values ge09 would be ldquoexcel-lentrdquo ge08 ldquogoodrdquo ge07 ldquofairrdquo and lt07 ldquopoorrdquo [28]

e Pearson correlation coefficient r was used to explorepossible relationship between median parameter values ofthe three models A strong correlation was assumed for08lt rle 1 a moderate correlation for 05lt rle 08 a weakcorrelation for 03lt rle 05 and no correlation for rle 03[29] e BlandndashAltman plot was used to show agreementsbetween parameter values of the three models All statisticalanalysis were performed using SPSS software 180 (ChicagoIL USA) and Plt 005 was considered statisticallysignificant

3 Results

For the case shown in Figure 1(a) the parameter maps of oneslice generated using the three methods (AATH SC andDP) are shown in Figure 1(c) It is evident that threemethods attain similar results in Ve and PS and both es-timates are smaller in tumor ROI than in normal tissue ROIFor Fp the estimate by AATH is smaller in tumor ROI andthe estimates by SC and DP are apparently close betweentumor and normal tissue ROI For Vp the estimates varyfrom close between tumor and normal tissue ROI (AATH)to smaller in tumor ROI (SC and DP)

Table 2 shows the ICC values for the measured pa-rameters where most ICC values are greater than 09 in-dicating very good agreement between measurements fromtwo observers us the parameter values as measured bytwo observers are averaged and utilized in the analysis asfollows

Table 3 shows the nonparametric statistical analysis ofparameters between cervix cancer and normal tissue ROIse parameters Fp Ve and PS of the three DCE models(AATH SC and DP models) were smaller in cervix cancerlesions than in normal cervix tissue and the differencesbetween lesions and normal tissue were mostly significantexcept for PS from the AATH model and Fp from the SCmodelVp by AATH and SC was significantly larger in cervixcancer lesions than in normal cervix tissue and DP yieldedcontradictory results though the difference between lesionsand normal tissue was not statistically significant ParameterVe attained fair to good AUC values (gt083 for AATH and

DP models and gt075 for the SC model) Other parameterseg Vp from the AATH model and PS from the SC modelalso showed fair AUC values

Results of the Pearson correlation between the sameparameters estimated by different models are shown inTable 4 Comparing parameter estimates of the three models(AATH SC and DPmodels) for cervix cancer lesions it wasfound that Fp and Vp from SC and DPmodels and Ve and PSfrom AATH and DP models were highly correlated re-spectively (rgt 08) Moderate correlations were observedbetween Vp from AATH and Vp from SC and DPmodels Veand PS from AATH and SC models and Ve and PS from SCand DP models (05ltrlt08) No correlation was observedbetween Fp from AATH and Fp from SC and DP modelsrespectively (rlt 03) Comparing parameter estimates of thethree models (AATH SC and DPmodels) for normal cervixtissues it was found that good agreement existed betweenthe estimates of Fp Ve from the three models and PS fromAATH and DP models (rgt 08) Moderate correlations wereobserved between PS from AATH and SC models and Vpfrom SC and DP models (05lt rlt 08) A weak correlationwas observed between PS from SC and DP models(03lt rlt 05) No correlation was observed between Vp fromAATH and Vp from SC and DP models respectively(rlt 03) Results of the Bland and Altman test for com-parison of the parameter with the same biophysical meaningobtained using the three models are shown in Figure 2BlandndashAltman plots demonstrated good agreements be-tween the estimates of Ve from the three models for bothnormal cervix tissue and cervix lesion because their y-co-ordinate values were centered at zero (Figures 2(a) and 2(b))Good agreements were observed between Vp from SC andDP models and PS from AATH and SC models in normalcervix tissue (Figure 2(a)) e y-coordinate value of otherparameters estimated by AATH SC and DP modes stayedaway from zero indicating less agreement between them innormal cervix tissue and cervix lesion

4 Discussion

is preliminary study evaluated the performance ofphysiological parameters derived from AATH SC and DPmodels with respect to tumor microcirculation in cervixcancer e angiogenic activity of cervix cancer assessed byvarious DCE-MRI models improved in vivo understandingof the fundamental processes involved in tumor angio-genesis Parameter Ve has the best performance in identi-fying cervix cancer among all parameters Mostphysiological parameters derived from the three DCE-MRImodels are linearly correlated in cervix cancer

e parameter Ve of all three DCE models showed thatthe Ve value of cervix cancer tissue was significantly smallerthan that of normal cervix tissue e corresponding AUCvalues were greater than 075 and Ve of the AATH modelhad the largest AUC (085) among all parameters indicatinga fairly good performance in terms of diagnostic value Intracer kinetic modeling Ve stands for the fractional volumeof extravascular extracellular space which is closely per-taining to the degree of cell growth e more the cells grow

6 Contrast Media amp Molecular Imaging

Tabl

e2

Interobserverconsistency

forcervix

lesio

nandtheno

rmal

cervix

tissue

AATH

SCDP

F pVp

Ve

PSF p

Vp

Ve

PSF p

Vp

Ve

PSLesio

nIC

C0922

0975

0959

0939

0935

0980

0968

0857

0965

0989

0958

0958

95

CI

0886ndash

0947

0963ndash

0983

0940ndash

0972

0911ndash

0958

0905ndash

0956

0971ndash

0987

0953ndash

0978

0790ndash

0903

0949ndash

0976

0984ndash

0993

0938ndash

0971

0938ndash

0971

Normal

ICC

0934

0661

0908

0948

0978

0945

0910

0887

0906

0812

0943

0892

95

CI

0902ndash

0955

0499ndash

0770

0864ndash

0938

0924ndash

0965

0967ndash

0985

0919ndash

0963

0867ndash

0939

0833ndash

0923

0861ndash

0936

0702ndash

0882

0916ndash

0961

0841ndash

0927

ICCintraclasscorrelationcoeffi

cientCIconfi

denceinterval

ofdifference

Contrast Media amp Molecular Imaging 7

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

parameters were used because the median is more robust tooutliers (that could occur during data fitting) than the meanA two-way model average measure intraclass correlationcoefficient (ICC) was used to test the interobserver con-sistency en all the measurements from the two observerswere averaged for further comparison Agreement wasinterpreted according to the ICC as follows gt080 excellent061ndash080 good 041ndash060 moderate 021ndash040 fair andlt02 poor agreement [27]

e normality of the distribution of all parameters wasanalysed by the KolmogorovndashSmirnov test e receiveroperating characteristic (ROC) curves of all parameters wereobtained and the areas under the curves (AUC) wereevaluated to determine the discriminating power of DCEparameters between the lesion and normal tissues In-terpretation of AUC values is application dependent and ingeneral it is appropriate that values ge09 would be ldquoexcel-lentrdquo ge08 ldquogoodrdquo ge07 ldquofairrdquo and lt07 ldquopoorrdquo [28]

e Pearson correlation coefficient r was used to explorepossible relationship between median parameter values ofthe three models A strong correlation was assumed for08lt rle 1 a moderate correlation for 05lt rle 08 a weakcorrelation for 03lt rle 05 and no correlation for rle 03[29] e BlandndashAltman plot was used to show agreementsbetween parameter values of the three models All statisticalanalysis were performed using SPSS software 180 (ChicagoIL USA) and Plt 005 was considered statisticallysignificant

3 Results

For the case shown in Figure 1(a) the parameter maps of oneslice generated using the three methods (AATH SC andDP) are shown in Figure 1(c) It is evident that threemethods attain similar results in Ve and PS and both es-timates are smaller in tumor ROI than in normal tissue ROIFor Fp the estimate by AATH is smaller in tumor ROI andthe estimates by SC and DP are apparently close betweentumor and normal tissue ROI For Vp the estimates varyfrom close between tumor and normal tissue ROI (AATH)to smaller in tumor ROI (SC and DP)

Table 2 shows the ICC values for the measured pa-rameters where most ICC values are greater than 09 in-dicating very good agreement between measurements fromtwo observers us the parameter values as measured bytwo observers are averaged and utilized in the analysis asfollows

Table 3 shows the nonparametric statistical analysis ofparameters between cervix cancer and normal tissue ROIse parameters Fp Ve and PS of the three DCE models(AATH SC and DP models) were smaller in cervix cancerlesions than in normal cervix tissue and the differencesbetween lesions and normal tissue were mostly significantexcept for PS from the AATH model and Fp from the SCmodelVp by AATH and SC was significantly larger in cervixcancer lesions than in normal cervix tissue and DP yieldedcontradictory results though the difference between lesionsand normal tissue was not statistically significant ParameterVe attained fair to good AUC values (gt083 for AATH and

DP models and gt075 for the SC model) Other parameterseg Vp from the AATH model and PS from the SC modelalso showed fair AUC values

Results of the Pearson correlation between the sameparameters estimated by different models are shown inTable 4 Comparing parameter estimates of the three models(AATH SC and DPmodels) for cervix cancer lesions it wasfound that Fp and Vp from SC and DPmodels and Ve and PSfrom AATH and DP models were highly correlated re-spectively (rgt 08) Moderate correlations were observedbetween Vp from AATH and Vp from SC and DPmodels Veand PS from AATH and SC models and Ve and PS from SCand DP models (05ltrlt08) No correlation was observedbetween Fp from AATH and Fp from SC and DP modelsrespectively (rlt 03) Comparing parameter estimates of thethree models (AATH SC and DPmodels) for normal cervixtissues it was found that good agreement existed betweenthe estimates of Fp Ve from the three models and PS fromAATH and DP models (rgt 08) Moderate correlations wereobserved between PS from AATH and SC models and Vpfrom SC and DP models (05lt rlt 08) A weak correlationwas observed between PS from SC and DP models(03lt rlt 05) No correlation was observed between Vp fromAATH and Vp from SC and DP models respectively(rlt 03) Results of the Bland and Altman test for com-parison of the parameter with the same biophysical meaningobtained using the three models are shown in Figure 2BlandndashAltman plots demonstrated good agreements be-tween the estimates of Ve from the three models for bothnormal cervix tissue and cervix lesion because their y-co-ordinate values were centered at zero (Figures 2(a) and 2(b))Good agreements were observed between Vp from SC andDP models and PS from AATH and SC models in normalcervix tissue (Figure 2(a)) e y-coordinate value of otherparameters estimated by AATH SC and DP modes stayedaway from zero indicating less agreement between them innormal cervix tissue and cervix lesion

4 Discussion

is preliminary study evaluated the performance ofphysiological parameters derived from AATH SC and DPmodels with respect to tumor microcirculation in cervixcancer e angiogenic activity of cervix cancer assessed byvarious DCE-MRI models improved in vivo understandingof the fundamental processes involved in tumor angio-genesis Parameter Ve has the best performance in identi-fying cervix cancer among all parameters Mostphysiological parameters derived from the three DCE-MRImodels are linearly correlated in cervix cancer

e parameter Ve of all three DCE models showed thatthe Ve value of cervix cancer tissue was significantly smallerthan that of normal cervix tissue e corresponding AUCvalues were greater than 075 and Ve of the AATH modelhad the largest AUC (085) among all parameters indicatinga fairly good performance in terms of diagnostic value Intracer kinetic modeling Ve stands for the fractional volumeof extravascular extracellular space which is closely per-taining to the degree of cell growth e more the cells grow

6 Contrast Media amp Molecular Imaging

Tabl

e2

Interobserverconsistency

forcervix

lesio

nandtheno

rmal

cervix

tissue

AATH

SCDP

F pVp

Ve

PSF p

Vp

Ve

PSF p

Vp

Ve

PSLesio

nIC

C0922

0975

0959

0939

0935

0980

0968

0857

0965

0989

0958

0958

95

CI

0886ndash

0947

0963ndash

0983

0940ndash

0972

0911ndash

0958

0905ndash

0956

0971ndash

0987

0953ndash

0978

0790ndash

0903

0949ndash

0976

0984ndash

0993

0938ndash

0971

0938ndash

0971

Normal

ICC

0934

0661

0908

0948

0978

0945

0910

0887

0906

0812

0943

0892

95

CI

0902ndash

0955

0499ndash

0770

0864ndash

0938

0924ndash

0965

0967ndash

0985

0919ndash

0963

0867ndash

0939

0833ndash

0923

0861ndash

0936

0702ndash

0882

0916ndash

0961

0841ndash

0927

ICCintraclasscorrelationcoeffi

cientCIconfi

denceinterval

ofdifference

Contrast Media amp Molecular Imaging 7

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

Tabl

e2

Interobserverconsistency

forcervix

lesio

nandtheno

rmal

cervix

tissue

AATH

SCDP

F pVp

Ve

PSF p

Vp

Ve

PSF p

Vp

Ve

PSLesio

nIC

C0922

0975

0959

0939

0935

0980

0968

0857

0965

0989

0958

0958

95

CI

0886ndash

0947

0963ndash

0983

0940ndash

0972

0911ndash

0958

0905ndash

0956

0971ndash

0987

0953ndash

0978

0790ndash

0903

0949ndash

0976

0984ndash

0993

0938ndash

0971

0938ndash

0971

Normal

ICC

0934

0661

0908

0948

0978

0945

0910

0887

0906

0812

0943

0892

95

CI

0902ndash

0955

0499ndash

0770

0864ndash

0938

0924ndash

0965

0967ndash

0985

0919ndash

0963

0867ndash

0939

0833ndash

0923

0861ndash

0936

0702ndash

0882

0916ndash

0961

0841ndash

0927

ICCintraclasscorrelationcoeffi

cientCIconfi

denceinterval

ofdifference

Contrast Media amp Molecular Imaging 7

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

the less the Ve value is Tumor is typically characterized bythe over-growth of cells us the observed smaller value ofVe in cervix cancer tissue suggests the over-growth of cells incervix cancer tissue in comparison with normal cervix tissuewhich is consistent with the feature of other cancer tissue

In the tissue region of interest there are three types oftransportation process for the tracer molecule the flowingprocess within the vascular space the exchange processbetween the vascular and extravascular extracellular spaceand the diffusion process within the extravascular extra-cellular space e first two processes are of interest in DCEtracer kinetic modeling and the associated parameters are Fpfor the flowing process and PS for the exchange process bothof which are linked to nutrition supply to cell growth Inmost solid tumors both parameters usually increase due tothe over-growth of tumor cells From Table 2 it is observedthat both parameters as estimated by the three models aresurprisingly smaller in cervix cancer tissue

Another parameter in DCE modeling which is alsoclosely relevant to cell growth is Vp the fractional volume ofvascular space characterizing the degree of microvascularityin tissue In general tumor tissue would develop moremicrovessels as a response to nutrition necessity in tumorcell over-growth Parameter Vp can be validated throughpathohistological experiment which can visualize andquantify the degree of microvascularity through the measureof microvessel density (MVD) Early experiments showedthat the MVD of cervix cancer tissue was larger than that ofnormal cervix tissue [30ndash32] In this study it is observed that

AATH and CC models yielded Vp values consistent withprevious pathohistological results and the DP measurementis on the contrary but with insignificant difference in value

e unusual observation in the microcirculation patternof cervix cancer tissue is also evidently reflected in Figure 3which exemplifies the change of tissue intensity after venousinjection of contrast media e baseline corresponds to theperiod the contrast media has not flowed into the tissue ofinterest Here the intensity of cervix cancer tissue is slightlybrighter than that of normal cervix tissue indicating thatcervix cancer tissue has shorter spin-lattice relaxation timeWith the arrival of contrast media tissue intensity rapidlyincreases till a ldquopeakrdquo value followed by a part with intensityslowly decreasing which stands for the wash-in phase andwash-out phase of contrast media respectively e speed ofuptake in the wash-in phase depends on the value of bloodflow and the larger the value the sharper the slope of theuptake curve e sharper slope of normal cervix tissuesuggests a lower blood flow for the cervix cancer tissuewhich is well in line with the observation in DCE modelinge ldquopeakrdquo value is primarily related to the degree ofmicrovascularity If the tissue develops more microvesselsthere would accumulate more tracer molecules hence toobserve brighter intensity value e lower ldquopeakrdquo value ofcervix cancer tissue suggests that cervix cancer tissue mightdevelop fewer vessels is is consistent with the finding bythe DP model but not consistent with either that by AATHand CC models or previous pathohistological experimentis apparent contradiction might be due to the dysfunction

Table 3 Comparison of model parameters between cervix carcinoma and normal cervix tissue

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)AATH-normal cervix 4457plusmn 2323 127plusmn 125 3215plusmn 1432 1040plusmn 643AATH- cervix cancer 3413plusmn 1515 255plusmn 230 1703plusmn 1080 888plusmn 428P lt0001a lt0001a lt005a gt005aAUC 067 075 085 056SC-normal cervix 1951plusmn 1596 600plusmn 765 2995plusmn 1733 1114plusmn 718SC-cervix cancer 1848plusmn 731 680plusmn 424 1921plusmn 2005 621plusmn 440P gt005b lt0001a lt0001a lt0001aAUC 050 064 075 075DP-normal cervix 1606plusmn 2122 490plusmn 324 3188plusmn 1841 967plusmn 562DP-cervix cancer 1483plusmn 523 438plusmn 306 1605plusmn 1334 813plusmn 396P lt001a gt005b lt0001a lt005aAUC 060 054 083 058AUC area under the ROC curve aComparison was performed by the MannndashWhitney U test bComparison was performed by the independent t test

Table 4 Results of the Pearson correlation between parameters of the three models (AATH SC and DP) in cervix cancer carcinoma andnormal cervix tissue rgt 05 and Plt 005 are indicated bylowast

Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) PS (mLmin100mL)LesionAATH-CC 0121 (P 0214) 0710 (Plt 0001)lowast 0572 (Plt 0001)lowast 0667 (Plt 0001)lowastAATH-DP 0095 (P 0329) 0732 (Plt 0001)lowast 0899 (Plt 0001)lowast 0921 (Plt 0001)lowastCC-DP 0813 (Plt 0001)lowast 0850 (Plt 0001)lowast 0565 (Plt 0001)lowast 0749 (Plt 0001)lowast

NormalAATH-CC 0829 (Plt 0001)lowast 0261 (P 0008) 0824 (Plt 0001)lowast 0554 (Plt 0001)lowastAATH-DP 0815 (Plt 0001)lowast 0247 (P 0012) 0912 (Plt 0001)lowast 0890 (Plt 0001)lowastCC-DP 0981 (Plt 0001)lowast 0696 (Plt 0001)lowast 0886 (Plt 0001)lowast 0385 (Plt 0001)

8 Contrast Media amp Molecular Imaging

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

70

+196 SD514

Mean251

ndash196 SDndash12

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

NSC

)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NSC)200 250

+196 SD98

Meanndash47

ndash196 SDndash193

+196 SD214

Mean22

ndash196 SDndash170

+196 SD159

Mean03

ndash196 SDndash153

+196 SD149

Meanndash19

ndash196 SDndash188

+196 SD119

Meanndash07

ndash196 SDndash134

+196 SD65

Mean07

ndash196 SDndash50

+196 SD156

Mean15

ndash196 SDndash127

+196 SD26

Meanndash36

ndash196 SDndash98

+196 SD126

Mean11

ndash196 SDndash104D

iffer

ence

in V

p (N

AAT

H an

d N

SC)

0 20 40 60 80 100 120 0 20 40

0 2010 30 400 2010 30 40 50

60 80 100 120 0 20 40 60 80 100 120

50 10 15 20 25 0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 2 4 6 8 10

70

+196 SD553

Mean285

ndash196 SD18

50

60

40

30

Diff

eren

ce in

Fp (

NA

ATH

and

ND

P)

20

10

0

ndash100 50 100 150

Mean value Fp (NAATH and NDP)200 250

+196 SD160

Mean35

ndash196 SDndash91

Diff

eren

ce in

Fp (

NSC

and

ND

P)

Diff

eren

ce in

Vp (

NA

ATH

and

ND

P)

Diff

eren

ce in

Vp (

NSC

and

ND

P)

Diff

eren

ce in

Ve (

NA

ATH

and

ND

P)

Diff

eren

ce in

Ve (

NSC

and

ND

P)

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50 ndash16 ndash15ndash10

ndash505

101520253035

ndash14ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash50

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash10

0

ndash5

5

10

15

20

Diff

eren

ce in

Ve (

NA

ATH

and

NSC

)

Diff

eren

ce in

PS

(NA

ATH

and

ND

P)

Diff

eren

ce in

PS

(NSC

and

ND

P)

Diff

eren

ce in

PS

(NA

ATH

and

NSC

)

ndash40ndash30ndash20ndash10

0

20304050

10

ndash20

ndash10

0

20

30

10

ndash40

ndash30

ndash20

ndash10

0

10

20

30

40

ndash30

ndash20

ndash10

20

30

40

10

0

0 50 100 150Mean value Fp (NSC and NDP)

Mean value Vp (NAATH and NSC) Mean value Vp (NAATH and NDP) Mean value Vp (NSC and NDP)

Mean value Ve (NAATH and NSC) Mean value Ve (NAATH and NDP) Mean value Ve (NSC and NDP)

Mean value PS (NAATH and NSC) Mean value PS (NAATH and NDP) Mean value PS (NSC and NDP)

200 250

(a)

Figure 2 Continued

Contrast Media amp Molecular Imaging 9

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

of vessels newly developed in cancer tissue [33] which mayaccumulate less tracer molecules resulting in the lowerldquopeakrdquo value of cervix cancer

e wash-out phase involves two processes one processwhere the contrast media flows out of the tissue of interestand another process where the contrast media leaks out ofvascular space through vascular wall and enters the extra-vascular extracellular space and returns from the extravas-cular extracellular space to the vascular space and flows outof the tissue of interest e flow-out process would lead tolower tissue intensity and the exchange process woulddetermine the speed of intensity decreasing e higher theblood flow is the faster the intensity decreases e more

leakier the vessel wall is the slower the intensity decreasese trend of wash-out phase would be determined by bothfactors If the blood flow is large and the effect of flow-outprocess is dominant a downward trend would be observedIf the blood flow is small and the vessel wall is leaky a slowlyupward or flat trend could be observed As there is an ev-ident downward trend in the wash-out phase of normalcervix tissue it demonstrates the higher blood flow rate innormal cervix tissue Consequently the intensity differencebetween cervix cancer tissue and normal cervix tissue isdecreasing in the early stage of wash-out phase In spite ofthis it is clear that the intensity difference becomes stabletowards the tail stage of the observed wash-out phase which

+196 SD470

Mean156

ndash196 SDndash157

+196 SD498

Mean193

ndash196 SDndash112

+196 SD121

Mean37

ndash196 SDndash48

+196 SD18

Meanndash43

ndash196 SDndash103

+196 SD23

Meanndash18

ndash196 SDndash59

+196 SD69

Mean24

ndash196 SDndash21

+196 SD301

Meanndash22

ndash196 SDndash344

+196 SD127

Mean10

ndash196 SDndash107

+196 SD358

Mean32

ndash196 SDndash295

+196 SD96

Mean27

ndash196 SDndash43

+196 SD40

Mean07

ndash196 SDndash25

+196 SD39

Meanndash19

ndash196 SDndash78

0 20 40 60 80

0 2010 30 40 50 60 0 2010 30 40 50 0 2010 30 40 506070

0 20 40 60 80 100 0 20 40 60 80 100

50 10 15 20 25 50 10 15 20 25 50 10 15 20 25

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20

Diff

eren

ce in

Fp (

LAAT

H an

d LS

C)D

iffer

ence

in P

S (L

AAT

H an

d LS

C)

Diff

eren

ce in

Fp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Fp (

LSC

and

LDP)

Diff

eren

ce in

Vp (

LAAT

H an

d LS

C)

Diff

eren

ce in

Vp (

LAAT

H an

d LD

P)

Diff

eren

ce in

Vp (

LSC

and

LDP)

ndash10

ndash5

0

5

10

ndash10

ndash5

0

5

10

15

20

ndash50

ndash60

ndash40

ndash30

ndash20

ndash10

0

10

20

ndash40

ndash20

0

20

40

60

80

ndash20

ndash14ndash16

ndash80

ndash60

ndash40

ndash20

0

20

40

ndash12ndash10

ndash8ndash6ndash4ndash2

024

ndash8ndash6ndash4ndash2

02468

ndash6 ndash10

ndash5

0

5

10

ndash4

ndash2

0

2

4

6

81012

ndash12

ndash10

ndash8

ndash6

ndash4

ndash2

0

2

4

ndash10

203040506070

100

ndash20

20

40

60

80

0

Mean value Fp (LAATH and LSC) Mean value Fp (LAATH and LDP) Mean value Fp (LSC and LDP)

Diff

eren

ce in

Ve (

LAAT

H an

d LS

C)

Diff

eren

ce in

Ve (

LAAT

H an

d LD

P)

Diff

eren

ce in

Ve (

LSC

and

LDP)

Diff

eren

ce in

PS

(LA

ATH

and

LDP)

Diff

eren

ce in

PS

(LSC

and

LDP)

Mean value Ve (LAATH and LSC) Mean value Ve (LAATH and LDP) Mean value Ve (LSC and LDP)

Mean value PS (LAATH and LSC) Mean value PS (LAATH and LDP) Mean value PS (LSC and LDP)

Mean value Vp (LAATH and LSC) Mean value Vp (LAATH and LDP) Mean value Vp (LSC and LDP)

(b)

Figure 2 (a) BlandndashAltman plots for Fp Vp Ve and PS from AATH SC and DP models in normal cervix tissue (N) (b) BlandndashAltmanplots for FpVp Ve and PS from AATH SC and DPmodels in the cervix lesion (L)e unit of these parameters is listed as follows Fp (mLmin100mL) Vp (mL100mL) Ve (mL100mL) and PS (mLmin100mL)

10 Contrast Media amp Molecular Imaging

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

is more influenced by the exchange process As aforemen-tioned the measured permeability PS in DCE modeling islarger in normal cervix tissue thus the downward trend innormal cervix tissue encounters larger resistance whichhelps precisely to explain the observed phenomena in thewash-out phase

From the aforementioned analysis cervix cancer exhibitsvery unusual pattern in microcirculation with over-growthof cells decreased blood flow decreased vessel wall per-meability and without well development of increased andfunctional microvascularity One immediately notices whatis the growing mechanism for tumor cells without additionalnutrition supply through the blood transportation system Itis known that microvascular proliferation is not the onlyfactor of cancer cell growth and some other factors likeinterstitial fluid pressure oxygen and human papillomavirus (HPV) [9 34 35] but may also promote tumor growthBut which of these pathophysiological mechanisms is thedominating factor for the growth of cervix cancer cells is notclear so far It would require a histopathologic and etiologicstudy to uncover the difference of growth condition betweencervix carcinoma and other tumors For clinicians to provideindividual cancer patients with optimal treatment andprecise prognosis further knowledge on the tumor micro-environment would be of great value

In the situation where two different kinetic models wereemployed in a multicenter clinical trial for the purposes ofassessment and monitoring of treatment effect using DCE-MRI it would be desirable that the kinetic parameters es-timated by the two models can be related ie that they wereat least linear correlated ParameterVe derived from all threemodels Vp estimated by SC and DP models and PS from

AATH and SCmodels in cervix lesion were highly correlatedas well as in good agreement indicating a fairly goodperformance in terms of evaluation of treatment effect inmulticenters In most situations parameters derived fromdifferent models were not quantitatively equivalent buthighly correlated For example if the SC and DP modelswere employed in different centers and one center observeda decrease in perfusion (tissue blood flow) of cervix cancerestimated by SC-Fp due to a particular treatment thenanother center employing the DP model should also observea similar percentage decrease in DP-Fp Because SC-Fp andDP-Fp were highly correlated but not quantitativelyequivalent we could only compare their relative difference(percentage change) in Fp before and after treatment and nottheir absolute values In a DCE CT study of brain tumors[36] parameters Fp Vp Ve and PS from the SC model werefound to be highly correlated with those from the DP modelese correlation results for these parameters in meningi-omas derived from DCE CT are very similar to the presentresults derived from DCE-MRI for cervix cancer (rgt 05)However comparing parameter estimates of the threemodels (AATH SC and DPmodels) for cervix in this studyit was found that the Pearson correlation between param-eters of the three models in cervix cancer lesion and normalcervix tissue yield inconsistent conclusion ie Fp fromAATH model and Fp from SC and DP models were highlycorrelated respectively in normal cervix tissue but veryweak correlation in cervix cancer lesionsVp from the AATHmodel and Vp from SC and DP models were highly cor-related respectively in cervix cancer lesions but very weakcorrelation in normal cervix tissue (Table 4) us theparameter correlations between kinetic models of tumors

0 50 100 150 200Time (s)

250 300 350 400

200

300

400

500

600

700

800

100

0

Mea

n sig

nal (

au)

NormalLesion

Figure 3 e signal intensity-time curve of cervix cancer ROI and the normal tissue for the same patient in Figure 1

Contrast Media amp Molecular Imaging 11

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

may be different from normal tissue and it should not begeneralized without validation for different tumors char-acterized with different perfusion Parameter relationshipsunder various tracer kinetic models in other tumors needfurther investigation

is study has a few limitations ere was a lack ofpathohistological experiment and the previous pathohisto-logical results by other researchers were employed for com-parison in this studye present DCE-MRI data were acquiredusing a protocol with high temporal resolution (2 s) in order tocapture rapid changes in arterial and tumor concentration withtime which led to a tradeoff in organ coverage and image spatialresolution As a result effects of lesion heterogeneity might notbe fully accounted for Possible error sources that could affectcontrast concentration estimation in this study include the lackof correction for effects of (i) B1 field inhomogeneity and (ii)T2lowast relaxation (i) Concentration estimation based on variableflip angle T1 mapping would be sensitive to spatial non-uniformity in the radio-frequency transmit field (B1) whichcauses variations in the prescribed flip angles (ii) At high-contrast concentrations effects of T2lowast relaxation become in-creasingly substantial in higher field strengths and could not beignored during concentration estimation Previous studies haveshown that failure to account for T2lowast effects could result inunderestimation of the AIF especially for concentrationsaround the AIF peak [37 38] Both effects (i) and (ii) can resultin errors in the arterial andor tissue concentration-time curveswhich further translate to errors in the estimation of kineticparameters [37 38] Amajor limitation of this study is thereforethe lack of quantification of such errors and their propagation tothe estimated kinetic parameters which would be the subject offurther study in our future work

In conclusion two-compartment exchange modelsturned out to be very promising tools in analyzing DCE dataand assessing the microcirculation pattern in cervix cancertissue hence with great value to assist cervix cancer di-agnosis and prognosis Parameter Ve has the best perfor-mance in identifying cervix cancer tissue among allparameters

Data Availability

e data of this study are already presented in this paper

Ethical Approval

All procedures performed involving human participantswere in accordance with the ethical standards of the SecondAffiliated Hospital and Yuying Childrenprimes Hospital ofWenzhou Medical University

Consent

All patients signed informed consent forms allowing theirdata to be used in this study

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is work was supported by Zhejiang Province HealthDepartment (grant no 2018KY534) and Wenzhou Scienceand Technology Bureau in China (Y20170825)

References

[1] J Folkman ldquoIntratumoral microvascular density as a prog-nostic factor in cancerrdquo e American Journal of Pathologyvol 147 pp 9ndash19 1995

[2] N Weidner J P Semple W R Welch and J FolkmanldquoTumor angiogenesis and metastases-correlation in invasivebreast carcinomardquoNew England Journal of Medicine vol 324no 1 pp 1ndash8 1991

[3] P Macchiarini G Fontanini F Squartini C A Angeletti andM J Hardin ldquoRelation of neovascularization to metastasis ofnon-small-cell lung cancerrdquo e Lancet vol 340 no 8812pp 145-146 1992

[4] D LWiggins C O Granai M M Steinhoff and P CalabresildquoTumor angiogenesis as a prognostic factor in cervical car-cinomardquo Gynecologic Oncology vol 56 no 3 pp 353ndash3561995

[5] K Schlenger M Hockel M Mitze et al ldquoTumor vascularity-anovel prognostic factor in advanced cervical carcinomardquoGynecologic Oncology vol 59 no 1 pp 57ndash66 1995

[6] S Mussurakis D L Buckley P J Drew et al ldquoDynamic MRimaging of the breast combined with analysis of contrast agentkinetics in the differentiation of primary breast tumoursrdquoClinical Radiology vol 52 no 7 pp 516ndash526 1997

[7] D L Buckley A E Shurrab C M Cheung A P JonesH Mamtora and P A Kalra ldquoMeasurement of single kidneyfunction using dynamic contrast-enhanced MRI comparisonof two models in human subjectsrdquo Journal of MagneticResonance Imaging vol 24 no 5 pp 1117ndash1123 2006

[8] S M Galbraith M A Lodge N J Taylor et al ldquoRe-producibility of dynamic contrast-enhanced MRI in humanmuscle and tumours comparison of quantitative and semi-quantitative analysisrdquo NMR in Biomedicine vol 15 no 2pp 132ndash142 2002

[9] H Hawighorst P G Knapstein W Weikel et al ldquoAngio-genesis of uterine cervical carcinoma characterization bypharmacokinetic magnetic resonance parameters and histo-logical microvessel density with correlation to lymphaticinvolvementrdquo Cancer Research vol 57 no 21 pp 4777ndash47861997

[10] J A Loncaster B M Carrington J R Sykes et al ldquoPredictionof radiotherapy outcome using dynamic contrast enhancedMRI of carcinoma of the cervixrdquo International Journal ofRadiation OncologylowastBiologylowastPhysics vol 54 no 3pp 759ndash767 2002

[11] F Kiessling M Lichy R Grobholz et al ldquoSimple modelsimprove the discrimination of prostate cancers from theperipheral gland by T1-weighted dynamic MRIrdquo EuropeanRadiology vol 14 no 10 pp 1793ndash1801 2004

[12] M A Rosen and M D Schnall ldquoDynamic contrast-enhancedmagnetic resonance imaging for assessing tumor vascularityand vascular effects of targeted therapies in renal cell carci-nomardquo Clinical Cancer Research vol 13 no 2 pp 770sndash776s2007

[13] A Dowlati K Robertson M Cooney et al ldquoA phase Ipharmacokinetic and translational study of the novel vasculartargeting agent combretastatin a-4 phosphate on a single-dose

12 Contrast Media amp Molecular Imaging

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

intravenous schedule in patients with advanced cancerrdquoCancer Research vol 62 no 12 pp 3408ndash3416 2002

[14] A Lomas B Morgan M A Horsfield et al ldquoPhase I studyof the safety tolerability pharmacokinetics and pharmaco-dynamics of PTK787ZK 222584 administered twice daily inpatients with advanced cancerrdquo Journal of Clinical Oncologyvol 23 no 18 pp 4162ndash4171 2005

[15] K Mross J Drevs M Muller et al ldquoPhase I clinical andpharmacokinetic study of PTKZK a multiple VEGF receptorinhibitor in patients with liver metastases from solid tu-moursrdquo European Journal of Cancer vol 41 no 9pp 1291ndash1299 2005

[16] P S Tofts G Brix D L Buckley et al ldquoEstimating kineticparameters from dynamic contrast-enhanced T(1)-weightedMRI of a diffusable tracer standardized quantities andsymbolsrdquo Journal of Magnetic Resonance Imaging vol 10no 3 pp 223ndash232 1999

[17] P R omas M C Peters A B Ennett and D J MooneyldquoPolymeric system for dual growth factor deliveryrdquo NatureBiotechnology vol 19 no 11 pp 1029ndash1034 2001

[18] G Bergers S Song M-M Nicole E Bergsland andD Hanahan ldquoBenefits of targeting both pericytes and en-dothelial cells in the tumor vasculature with kinase in-hibitorsrdquo Journal of Clinical Investigation vol 111 no 9pp 1287ndash1295 2003

[19] R Stollberger and F Fazekas ldquoImproved perfusion and tracerkinetic imaging using parallel imagingrdquo Topics in MagneticResonance Imaging vol 15 no 4 pp 245ndash255 2004

[20] T-N Isabelle D Balvay C A Cuenod E Daraı C Marsaultand M Bazot ldquoDynamic contrast-enhanced MR imaging toassess physiologic variations of myometrial perfusionrdquo Eu-ropean Radiology vol 20 no 4 pp 984ndash994 2010

[21] F Khalifa A Soliman E-B Ayman et al ldquoModels andmethods for analyzing DCE-MRI a reviewrdquo Medical Physicsvol 41 no 12 Article ID 124301 2014

[22] T S Koh S Bisdas D M Koh and C H ng ldquoFunda-mentals of tracer kinetics for dynamic contrast-enhancedMRIrdquo Journal of Magnetic Resonance Imaging vol 34 no 6pp 1262ndash1276 2011

[23] K S S Lawrence and T-Y Lee ldquoAn adiabatic approximationto the tissue homogeneity model for water exchange in thebrain I eoretical derivationrdquo Journal of Cerebral BloodFlow amp Metabolism vol 18 no 12 pp 1365ndash1377 1998

[24] S B Donaldson C M L West S E Davidson et al ldquoAcomparison of tracer kinetic models for T1-weighted dynamiccontrast-enhanced MRI application in carcinoma of thecervixrdquo Magnetic Resonance in Medicine vol 63 no 3pp 691ndash700 2010

[25] J F Kallehauge K Tanderup C Duan et al ldquoTracer kineticmodel selection for dynamic contrast-enhanced magneticresonance imaging of locally advanced cervical cancerrdquo ActaOncologica vol 53 no 8 pp 1064ndash1072 2014

[26] S Pecorelli ldquoRevised FIGO staging for carcinoma of the vulvacervix and endometriumrdquo International Journal of Gyne-cology amp Obstetrics vol 105 no 2 pp 103-104 2009

[27] X Zou Y Luo Z Li et al ldquoVolumetric apparent diffusioncoefficient histogram analysis in differentiating intrahepaticmass-forming cholangiocarcinoma from hepatocellular car-cinomardquo Journal of Magnetic Resonance Imaging vol 49no 4 pp 975ndash983 2018

[28] E A Youngstrom ldquoA primer on receiver operating charac-teristic analysis and diagnostic efficiency statistics for pedi-atric psychology we are ready to ROCrdquo Journal of PediatricPsychology vol 39 no 2 pp 204ndash221 2014

[29] S Zwick G Brix P S Tofts et al ldquoSimulation-based com-parison of two approaches frequently used for dynamiccontrast-enhanced MRIrdquo European Radiology vol 20 no 2pp 432ndash442 2010

[30] S Mondal S Dasgupta P Mandal S Chatterjee andD Chakraborty ldquoIs there any role of mast cell density andmicrovessel density in cervical squamous cell carcinoma Ahistologic study with special reference to CD-34 immuno-marker stainingrdquo Indian Journal of Medical and PaediatricOncology vol 35 no 2 pp 165ndash169 2014

[31] L Benıtez-Bribiesca A Wong D Utrera and E Castellanosldquoe role of mast cell tryptase in neoangiogenesis of pre-malignant and malignant lesions of the uterine cervixrdquoJournal of Histochemistry amp Cytochemistry vol 49 no 8pp 1061-1062 2001

[32] M Sotiropoulou E Diakomanolis A Elsheikh D LoutradisS Markaki and S Michalas ldquoAngiogenic properties of car-cinoma in situ and microinvasive carcinoma of the uterinecervixrdquo European Journal of Gynaecological Oncology vol 25no 2 pp 219ndash221 2004

[33] P Carmeliet and R K Jain ldquoMolecular mechanisms andclinical applications of angiogenesisrdquo Nature vol 473no 7347 pp 298ndash307 2011

[34] R A Cooper BM Carrington J A Loncaster et al ldquoTumouroxygenation levels correlate with dynamic contrast-enhancedmagnetic resonance imaging parameters in carcinoma of thecervixrdquo Radiotherapy and Oncology vol 57 no 1 pp 53ndash592000

[35] A H Masoom I Sitartchouk T P L Roberts A FylesA T Hashmi and M Milosevic ldquoCorrelations between dy-namic contrast-enhanced magnetic resonance imaging -derived measures of tumor microvasculature and interstitialfluid pressure in patients with cervical cancerrdquo Journal ofMagnetic Resonance Imaging vol 25 no 1 pp 153ndash159 2007

[36] D L H Cheong C C Tchoyoson Lim and T S KohldquoDynamic contrast-enhanced CTof intracranial meningiomacomparison of distributed and compartmental tracer kineticmodelsmdashinitial resultsrdquo Radiology vol 232 no 3pp 921ndash930 2004

[37] M C Schabel and D L Parker ldquoUncertainty and bias incontrast concentration measurements using spoiled gradientecho pulse sequencesrdquo Physics in Medicine and Biologyvol 53 no 9 pp 2345ndash2373 2008

[38] D De Naeyer I Debergh Y De Deene W P CeelenP Segers and P Verdonck ldquoFirst order correction for T2lowast-relaxation in determining contrast agent concentration fromspoiled gradient echo pulse sequence signal intensityrdquo Journalof Magnetic Resonance Imaging vol 34 no 3 pp 710ndash7152011

Contrast Media amp Molecular Imaging 13

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom