single slice versus whole organ/tumorclinical-mri.com/wp-content/uploads/2014/11/...thus, 43 kidneys...

Dynamic Contrast-Enhanced Magnetic Resonance ImagingAssessment of Kidney Function and Renal Masses

Single Slice Versus Whole Organ/Tumor

Katharina S. Winter,* Andreas D. Helck, MD,* Michael Ingrisch, PhD,Þ Michael Staehler, MD,þChristian Stief, MD,þ Wieland H. Sommer, MD, MPH,* Margarita Braunagel, MD,*

Philipp M. Kazmierczak, MD,* Maximilian F. Reiser, MD, FACR, FRCR,* Konstantin Nikolaou, MD,*§and Mike Notohamiprodjo, MD*§

Objectives: The aim of this study was to compare single-slice and 3-dimensional(3D) analysis for magnetic resonance renography (plasma flow [FP], plasma volume[VP], and glomerular filtration rate [GFR]) and for dynamic contrast-enhanced mag-netic resonance imaging (MRI) of renal tumors (FP, VP, permeability-surface areaproduct), respectively.Material and Methods: We prospectively included 22 patients (43 kidneys with22 suspicious renal lesions) and performed preoperative and postoperative imagingbefore and after partial nephrectomy, respectively. Of the 22 renal lesions, 15 turnedout to be renal cell carcinoma and were included in the tumor analysis, altogetherleading to 86 renal and 15 tumor MRI scans, respectively. Dynamic contrast-enhanced MRI was performed with a time-resolved angiography with stochastictrajectories sequence (spatial resolution, 2.6� 2.6� 2.6 mm3; temporal resolution,2.5 seconds) at 3 T (MagnetomVerio; Siemens Healthcare Sector) after injection of0.05 mmol/kg body weight Gadobutrol (Bayer Healthcare Pharmaceuticals).Analysis was performed using regions of interest encompassing a single centralslice and the whole kidney/tumor, respectively. A 2-compartment model yieldingFP, VP, GFR, or tumor permeability-surface area product was used for kineticmodelling. Modelling was performed based on relative contrast enhancement toaccount for coil-related inhomogeneity. Significance in difference, agreement, andgoodness of fit of the data to the curve was assessed with paired t tests, Bland-Altman plots, and W2 test, respectively.Results: Bland-Altman analysis revealed a good agreement between both types ofmeasurement for kidneys and tumors, respectively. Results between single-slice andwhole-kidney regions of interest showed significant differences for Fp (single slice,256.1 T 104.1 mL/100 mL/min; whole kidney, 217.2 T 92.5 mL/100 mL/min;P G 0.01). Regarding VP and GFR, no significant differences were observed.The W2 test showed a significantly better goodness of fit of the data to the curve forwhole kidneys (0.30% T 0.18%) than for single slices (0.43% T 0.26%) (P G 0.01).In contrast to renal assessment, tumor analysis showed no significant differencesregarding functional parameters and W2 test, respectively.Conclusion: In dynamic contrast-enhanced MRI of the kidney, both 3D whole-organ/tumor and single-slice analyses provide roughly comparable values in func-tional analysis. However, 3D assessment is considerablymore precise and should bepreferred if available.

Key Words: DCE-MRI, 3D analysis, kidney function, RCC

(Invest Radiol 2014;49: 720Y727)

Dynamic contrast-enhanced (DCE) magnetic resonance imaging(MRI) provides information on perfusion and filtration of kidneys1Y4

as well as perfusion and permeability of renal tumors.5,6 Data analysis iscomplex and can be affected by several factors, for example, the choice ofdifferent contrast agents7 or the definition of the region of interest (ROI).8

Precise kinetic modelling requires a high temporal and adequatespatial resolution. Consequently, DCE-MRI is usually performed on ROIsdefined on only a single central slice.9Y12 However, recently developed fastview-sharing sequences combined with parallel imaging and higher fieldstrengths allow for coverage of a whole volume of interest with a hightemporal resolution.13Y15 Principally, assessment of thewhole organ/lesionshould provide more representative information on perfusion and filtra-tion. Given the potential inhomogeneity of renal tumors, this seems to beparticularly important for the assessment of large tumors.

To our knowledge, the results and robustness of single-slice versuswhole-kidney/tumor DCE-MRI measurements have not been comparedyet. There are only few remotely similar studies investigating the extent ofbody fat. These studies showed benefits for multiple-slice imaging be-cause of the fact that a smaller sample size was needed to achieve sig-nificant results compared with single-slice imaging.16,17

The aim of this study, therefore, was to compare single centralslice with whole-organ/tumor analysis for DCE-MRI of the kidney andrenal tumors.

MATERIAL AND METHODS

Study SubjectsThe study was approved by the local ethics committee and written

informed consent was obtained from all patients. From October 2011 toMarch 2013, 22 consecutive patients (6 women, 16men;mean age, 61.7 T15.3 years) with ultrasound-suspected renal masses were prospectivelyincluded. Each patient underwent 2 examinations, the first was 3 monthsbefore and the second was 3 months after partial nephrectomy (Fig. 1).Overall, 7 nontumorous lesions like cysts or fatty tissue as well as tumorssmaller than 2 cm were excluded. One of the patients had only 1 kidney.Thus, 43 kidneys and 15 solid renal masses were assessed. A poweranalysis was performed and showed a power of greater than 80% for allexamined parameters of the kidneys and the renal tumors, respectively.Common criteria for exclusion from MRI (eg, pacemaker, contrast aller-gies, glomerular filtration rate [GFR] G30 mL/min) were applied.

Imaging ProtocolDynamic contrast-enhancedMRI data volumeswere acquired on a

clinical 3.0-T scanner (Magnetom Verio; Siemens Healthcare Sector,Erlangen, Germany) with a dedicated 6-element body array matrix coiland 6 elements of the integrated spine coil.Dynamic contrast-enhancedMRIwas performed during free-breathing with a coronal dynamic 3-dimensional(3D) magnetic resonance (MR) angiography time-resolved angiographywith stochastic trajectories13,15,18Y20 sequence. This is an ultrafast 3Dgradient echo sequence that combines view sharing with parallel imagingto achieve a high temporal and spatial resolution within a large field of

ORIGINAL ARTICLE

720 www.investigativeradiology.com Investigative Radiology & Volume 49, Number 11, November 2014

Received for publication February 9, 2014; and accepted for publication, after re-vision, April 14, 2014.

From the *Institute for Clinical Radiology, †Josef-Lissner Institute for BiomedicalImaging, Institute for Clinical Radiology, and ‡Department of Urology,University Hospitals Munich, Munich; and §Department of Diagnostic andInterventional Radiology, University Hospital Tuebingen, Tuebingen, Germany.

Katharina S. Winter and Andreas D. Helck contributed equally to this work.Conflicts of interest and sources of funding: none declared.Reprints: Katharina S. Winter, Institute for Clinical Radiology, University-Hospitals

Munich, Campus Grosshadern, Marchioninistrasse 15, 81377 Munich, Germany.E-mail: [email protected].

Copyright * 2014 by Lippincott Williams & WilkinsISSN: 0020-9996/14/4911Y0720

Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

mailto:[email protected]

view. View sharing comprises sampling of an inner region A and anouter region B of k-space in an alternating fashion. Data from ad-jacent acquisitions of B are then shared for reconstruction of in-dividual timeframes. A total of 40 slices with an isotropic voxelsize (2.6 � 2.6 � 2.6 mm3) were acquired with a temporal resolu-tion of 2.5 seconds for a total measurement time of 5 minutes. Thecentral region A of k-space was 20%, and the sampling density inthe outer region B was 25%. All sequence parameters are providedin Table 1. Acquisition started after injection of 0.05 mmol/kgbody weight gadobutrol (Bayer Healthcare Pharmaceuticals)7 witha flow of 2 mL/s.

The morphological imaging protocol including T1- and T2-weighted sequences is provided in Table 2.

PostprocessingThe in-house built IDL software PMI 0.421was used for quantitative

analysis (KSWand MN). An individual arterial input function was drawnin the aorta on a coronal slice. All arterial input function ROIs had astandardized size (4 voxels) and were defined at the height of the exitof the renal arteries in the middle of the aorta. The contrast agent

concentrations were approximated from the signal intensities using therelative signal enhancement22 to account for inhomogeneity in the coilsensitivity profiles. First, model-free deconvolution analysis wasperformed on a pixel basis, yielding whole-organ/tumor plasma flow(FP) and plasma volume (VP) maps. Regions of interest encompassing thekidney cortex or tumor were segmented semiautomatically on FP mapsusing an individual threshold. Voxels outside the organ were removedmanually. The single central slices were chosen from the middle of theorgan and were located at the level of the renal hilus. We excluded largevessels to prevent inflow effects. To ensure an equivalent threshold,single-slice ROIs were selected from the center of each ROI coveringthe whole organ/tumor (Figs. 2 and 3). There were 2 readers with8 years (MN) and 2 years (KSW) of experience in MRI of the kidney,respectively, who performed the image analysis in consensus. Tracer-kinetic modelling of kidney tissue is fundamentally different from tu-mor tissue, because the fraction of tracer that is extracted from theblood via the GFR does not return to the blood stream but insteadpasses through the tubular system and is then evacuated to thebladder. Hence, different tracer-kinetic models are necessary for thequantification of kidney and tumor perfusion.

FIGURE 1. Flowchart of patient collective. A total of 22 patients with 43 kidneys and 15 tumors were included in this study. The kidneyswere examined twice, before and 3 months after partial nephrectomy. The tumors were examined once, before partial nephrectomy.Overall, there were 86 kidney investigations and 15 tumor investigations.

TABLE 1. Acquisition Parameters of the Coronal TWIST Sequence

Sequence TR TE FA FOV-Read FOV-Phase Parallel Imaging View Sharing Slice Thickness Slices TA

TWIST 2.56 ms 0.85 ms 15- 500 mm 100.0% GRAPPA, accelerationfactor 4, 24 reference lines

Central region A, 20%;sampling density inregion B, 25%

2.6 mm 40 300 s

TWIST indicates time-resolved imaging with stochastic trajectories; TR, repetition time; TE, echo time; FA, flip angle; FOV, field of view; TA, acquisition time;GRAPPA, generalized autocalibrating partially parallel acquisition.

Investigative Radiology & Volume 49, Number 11, November 2014 Single-Slice vs. Whole-Kidney/Tumor DCE-MRI

* 2014 Lippincott Williams & Wilkins www.investigativeradiology.com 721


In these regions, we used a 2-compartment filtration model to assesskidney function and a 2-compartment exchange model to evaluate tumorperfusion and permeability-surface area product (PS) (Fig. 4).22Y26 Specifi-cally, we calculated FP, VP, GFR, and tubular volume in the kidney cortexregions with a 2-compartment filtration model, and FP, VP, PS, and inter-stitial volume in the tumor regions with a 2-compartment exchange model.

Statistical AnalysisWe used a 2-tailed paired t test to determine significant dif-

ferences between single-slice and volumetric results. Statistical sig-nificance was considered for P G 0.05. Bland-Altman plots weregenerated to define mean difference and 95% limits of agreement.

We used the W2 test to compare the goodness of fit of the deriveddata with the curves that were provided by the 2-compartment models.

W2 ¼~n

i

�mðt iÞjcðt iÞ

�2

Differences between each observed [m(ti)] and expected [c(ti)]data points were calculated, squared, and summed up. Lower resultsindicate a better fit of the data. Statistical analysis was performedwith Excel 2013 (Microsoft Inc, Redmond, WA), GraphPad program(GraphPad Software, Inc, La Jolla, CA), and MedCalc for Windows,version 12.5 (MedCalc Software, Ostend, Belgium).

RESULTS

ParticipantsAll examinations were diagnostic and no data set was

discarded. No adverse reactions or other complications were noted.Renal tumor histology was as follows: 8 clear cell, 2 chromophobe,and 2 papillary RCC and 3 angiomyolipoma. The mean tumor sizewas 5.4 T 2.4 cm, and it was more frequently on the left side (67%).For detailed information on the localization and size of the tumors,refer to Table 3.

DCE-MRI of the Kidney

Functional AnalysisTable 4 provides detailed information on the derived func-

tional parameters. Figure 2 displays segmentation and assessment of1 exemplary kidney.

Functional analysis showed that results for single-slice evalu-ation were slightly higher than for whole kidney (approx. FP, 17.9%;VP, 11.6%; GFR, 7.8%). Nonetheless, data of VP and GFR were notsignificantly different. Only differences for FP were significant(256.1 T 104.1 vs 217.2 T 92.5 mL/100 mL/min) (P G 0.01).

Limits of AgreementFor detailed information, refer to Table 5 and Figure 5.

TABLE 2. Imaging Protocol

Index Series Plane TR TE FA

1 T2-HASTE Sagittal 1000 ms 91 ms 150-

2 T2-HASTE Coronal 1200 ms 91 ms 150-

3 T2-HASTE Transversal 900 ms 91 ms 150-

4 T1-fl2d Transversal 140 ms 2.46 ms 65-

5 T1-VIBE Coronal 3.2 ms 1.13 ms 11-

6 Truefisp Transversal 4.04 ms 2.02 ms 70-

YInjection:half body weight adapted dose of gadobutrol (Bayer Healthcare Pharmaceuticals)

7 TWIST Coronal 2.56 ms 0.85 ms 15-

8 BLADE Transversal 2300 ms 102 ms 140-

YInjection:full body weight adapted dose of gadobutrol (Bayer Healthcare Pharmaceuticals)

9 Fl3d Coronal 2.87 ms 1.06 ms 18-

10 T1-VIBE Coronal 3.20 ms 1.13 ms 11-

TR indicates repetition time; TE, echo time; FA, flip angle; HASTE, half Fourier-acquired single shot turbo spin echo; Fl2d, fast low-angle shot 2D; VIBE, volumeinterpolated breath hold examination; Truefisp, true fast imaging with steady precession; TWIST, time-resolved imaging with stochastic trajectories; BLADE, periodicallyrotated overlapping parallel lines with enhanced reconstruction (PROPELLER); Fl3d, fast-low angle shot 3D.

FIGURE 2. Exemplary whole-kidney segmentation. Morphological images (A) of a right kidney of a 68-year-old male patient 3 monthsafter partial nephrectomy of the contralateral kidney, leaving a minor tissue defect. Model-free deconvolution analysis provided FP maps(mL/100 mL/min) (B), serving as basis for ROI definition (C). The white framed central slices were chosen for single-slice evaluation.

Winter et al Investigative Radiology & Volume 49, Number 11, November 2014

722 www.investigativeradiology.com * 2014 Lippincott Williams & Wilkins


Bias and 95% limits of agreement of renal FP, VP, and GFRshowed good agreement. For VP and GFR, mean difference was smallerthan 10% and the range between the limits of agreement appeared small.Mean difference of FP (38.9 mL/100 mL/min) was about 15.2%, whichmatches with the results of functional analysis.

Goodness of FitBox-and-whisker plots of data yielded in the W

2 test aredisplayed in Figure 6.

The results demonstrate that the fit of the curves for whole-organ measurement (0.30% T 0.18%) was significantly better (P =0.0002) than for single slices (0.43% T 0.26%).

DCE-MRI of Renal Tumors

Functional AnalysisTable 4 provides detailed information on the derived functional

parameters. Figure 3 displays segmentation and assessment of 1 exem-plary tumor. Table 6 shows the functional parameters FP and VP withrespect to the histopathological classification of the tumors. Results be-tween single-slice and whole-tumor measurement were almost equiva-lent. We did not record significant differences for FP, VP, or PS.

Limits of AgreementFor detailed information of the Bland-Altman analysis, refer to

Table 5 and Figure 5.Bias and 95% limits of agreement of FP, VP, and PS demonstrated

high agreement.Mean differencewas smaller than 10% for all parameters.

Goodness of FitBox-and-whisker plots of datayielded in theW2 test are displayed in

Figure 6. The fit of the tumor enhancement curve was more precise whenmeasuring the whole mass (2.74% T 2.88%) instead of 1 single slice(4.05% T 4.79%). However, deviations were not significant (P = 0.34).

DISCUSSIONIn this DCE-MRI-study, we could show that whole-organ/tumor

measurement is more precise than the measurement derived from a sin-gle slice. In functional analysis, however, both kinds of measurementsprovide comparable values. To our knowledge, this study is the first tocompare the accuracy of the 2 mentioned DCE-MRI measurements.

Single Slice Versus Whole KidneyIn our study, single-slice and whole-kidney DCE-MRI perfusion

parameters yielded similar values. The only parameter that differed

FIGURE 3. Exemplary whole-tumor segmentation.Morphological images (A) of a right kidney of a 79 year-oldmale patient with clearcell renal cell carcinoma, before partial nephrectomy. Model-free deconvolution analysis provided FP maps (mL/100 mL/min) (B),serving as basis for ROI definition (C). The white framed central slices were chosen for single-slice evaluation.

FIGURE 4. Exemplary diagrams of 2-compartment models. Carried by FP, the contrast agent enters the tissue and distributes overplasma volume compartment with volume fraction (VP). A, Filtration model for renal calculation. A part is filtered out and carriedby tubular flow (FT = GFR) into the tubular compartment where it spreads over the tubular volume (VT). The outflow of the vascularand tubular compartments transports the contrast agent out of the tissue. B, Exchange model for tumor assessment. A fraction iscarried by a flow (FE) to an extravascular compartment (Ve), where it returns in the same proportion. The contrast agent leaves thetissue by vascular compartment outflow.




significantly in the functional analysis was FP, which commonly showsthe largest variability.27 Also, the standard deviation of the FP values waslarger using the single-slice approach, which is probably because of thefact that 3D measurement is more robust than the single-slice technique.The deviation might also be partly caused by coil inhomogeneity orpartial volume effects at the peripheral organ borders.28,29

Except for FP, Bland-Altman analysis showed a high agree-ment of the parameters.

With respect to the goodness of fit of the derived functional data tothe provided enhancement curves, the fit was better for whole-kidneyassessment, indicating that this technique is more precise. As 3D as-sessment covers the whole organ, it enables to measure the kidneyfunction without the need for choosing a representative slice. The latterconstitutes a potential source of error and particularly raises problems ofcoherency in follow-up examinations. In contrast, high resolutions andisotropic voxel sizes of 3D sequences allow for multiplanar re-constructions in all planes, thereby enabling side-by-side correlationswith isotropic T1 sequences.

However, 3D analysis is more time-consuming than single-sliceanalysis. Depending on the readers’ experience, a full covered analysisof the kidney takes up to 10 minutes, whereas the evaluation of singleslices only takes 2 to 3 minutes. Especially the definition of the ROIs,which requires the removal of the voxels outside the organ, is time-consuming. As this is a matter of practice, the evaluation of the wholeorgan becomes easier and quicker with growing experience.

With respect to the temporal constraints of DCE-MRI, ahigher temporal resolution can be achieved using the single-sliceimage acquisition technique. Previous studies that assessed the kid-ney function on single slices commonly used protocols with a tem-poral resolution of 1 second.9,30 In our study, we chose a lowertemporal resolution of 2.5 seconds for both single-slice and whole-organ assessment to ensure full comparability between the 2 tech-niques. Still, this temporal resolution appears reasonable as it hasbeen shown that even resolutions of up to 4 seconds can providereliable outcomes with an error rate of less than 10%.31

Apart from that, it is to be expected that technical developmentswill dramatically accelerate the acquisition of 3D protocols32 as well asthe postprocessing of 3D data sets, which is currently much more time-consuming than the evaluation of a single slice.33

Single Slice Versus Whole TumorIn our functional analysis of renal masses, single-slice and

whole-tumor DCE-MRI yielded comparable results without signifi-cant differences. Similar to the assessment of the whole kidney, thegoodness of fit was higher for whole-tumor analysis. In contrast tothe whole-organ analysis, however, the difference did not reach sta-tistical significance. This might be because of the fact that only arelatively small number of tumors were analyzed. In addition, forwhole-tumor assessment, usually less slices were needed than for thewhole kidney, lowering the probability of finding a significant dif-ference between the functional parameters of a single slice and thoseof the multiple slices in 3D assessment.

In addition to the evaluation of all tumors, we analyzed thefunctional data with respect to the histopathological subtype of clearcell RCC, chromophobe RCC, and papillary RCC. Likewise, theresults did not show any significant differences between single-sliceand whole-organ measurements. The analysis of the different tumorsubtypes showed that clear cell carcinoma had a higher perfusioncompared with the other subtypes, which is consistent to a previousstudy on DCE-MRI for characterization and differentiation of renalcell carcinoma.24 The higher perfusion of clear cell RCC might beuseful for the identification of this subtype.

When interpreting the similar results of the 2 different techniquesin our study, it must also be taken into account that we did not includepatients under antiangiogenic therapy. However, tumor inhomogeneityoften increases under the influence of this treatment.34 In these cases aswell as for asymmetric tumors, a 3D approach covering the whole tumorappears definitely preferable.35

TABLE 3. Tumor Characteristics

Tumor Histology Diameter Side Localization

1 Clear cell RCC 10.4 cm Left Lower pole

2 Chromophobe RCC 4.4 cm Left Lower pole

3 Angiomyolipoma 2.1 cm Left Upper pole

4 Papillary RCC 8.9 cm Right Upper pole

5 Chromophobe RCC 6.6 cm Left Upper pole

6 Papillary RCC 6.1 cm Left Lower pole

7 Angiomyolipoma 6.1 cm Right Lower pole

8 Clear cell RCC 7.9 cm Left Center


10 Clear cell RCC 5.3 cm Right Lower pole


12 Angiomyolipoma 2.6 cm Left Center

13 Clear cell RCC 3.3 cm Left Upper pole

14 Clear cell RCC 6.3 cm Right Upper pole

15 Clear cell RCC 3.9 cm Right Lower pole

Average

Clear cell RCC: 53.3%

5.4 cm

Left: 66.7% Upper pole: 33.3%

Chromophobe RCC:13.3%

Lower pole: 40%

Papillary RCC: 13.3% Right: 33.3% Center: 26.7%

Angiomyolipoma: 20%

Diameter measured in coronal orientation.

RCC indicates renal cell carcinoma.

TABLE 4. Data of Kidney Perfusion and Filtration/Tumor Perfusion and Permeability

Kidney Tumor

FP VP GFR FP VP PS

Single slice 256.1 T 104.1 28.0 T 9.8 15.2 T 7.8 112.8 T 81.1 18.8 T 10.7 3.3 T 2.1

Whole organ 217.2 T 92.5 25.1 T 10.2 14.1 T 7.5 112.7 T 84.5 18.7 T 10.6 3.4 T 2.5

P 0.01* 0.06 0.37 1.00 0.98 0.85

*P G 0.05.

Mean value, standard deviation, and significant paired t-test P values are shown.

FP indicates plasma flow (mL/100 mL/min); VP, plasma volume (mL/100 mL); GFR, glomerular filtration rate (mL/100 mL/min); PS, permeability-surface areaproduct (mL/100 mL/min).




FIGURE 5. Bland-Altman analysis. Diagrams of bias and agreement of measurements. Averages (x-axis) and differences (y-axis) ofsingle-slice versus whole-organ measurement are plotted. Except for FP, all parameters achieved good values for MD and no majordivergences between 95% LA appeared. MD corresponds to central blue broken lines. Higher and lower 95% LA each correspond tored dotted line. A, KidneyVFP (mL/100 mL/min): MD = 38.9, LA =j42.1 to 120; VP (mL/100 mL): MD = 2.9, LA =j3.3 to 9.2; GFR(mL/100 mL/min): MD = j0.8, LA = j8 to 6.5. B, TumorVFP (mL/100 mL/min): MD = 0.2, LA = j22.6 to 23.0; VP (mL/100 mL):MD = 0.1, LA = j4.3 to 4.5; PS (mL/100 mL/min): MD = j0.2, LA = j1.8 to 1.5. MD indicates mean difference;LA, 95% limits of agreement; GFR, glomerular filtration rate; PS, permeability-surface area product.

TABLE 5. Bias and Agreement of Bland Altman Analysis

Kidney Tumor

FP VP GFR FP VP PS

MD 38.9 2.9 j0.8 0.2 0.1 j0.2

SD 41.3 3.2 3.7 11.6 2.2 0.8

LA j42.1 to 120.0 j3.3 to 9.2 j8.0 to 6.5 j22.6 to 23.0 j4.4 to 4.5 j1.8 to 1.5

Calculated with GraphPad program (GraphPad Software, Inc).

FP indicates plasma flow (mL/100 mL/min); VP, plasma volume (mL/100 mL); GFR, glomerular filtration rate (mL/100 mL/min); PS, permeability-surface area product(mL/100 mL/min); MD, mean difference; SD, standard deviation; LA, 95% limits of agreement.




As the goodness of fit is part not only of DCE-MRI but also of otherfunctional MR techniques, our findings might, in principle, be transferableto other contrast-enhanced methods, particularly CT perfusion,36Y38 as wellas to contrast mediaYfree perfusion MRI techniques such as diffusion-weighted imaging,3,39,40 blood oxygenation levelYdependent imaging,41,42

ultrasound imaging,43,44 or arterial spin labeling perfusion in kidneys45Y47

and renal tumors.48 However, further studies are required to evaluate pos-sible advantages of 3D assessment for these functional techniques.

Our study has some limitations. First, we did not include a standardof reference. However, it has been shown that DCE-MRI of the kidneycorrelates very well with radionuclide methods, which is considered as thegold standard.49,50 Second, the tumor locations varied among the patients.We did not differentiate between anterior and posterior tumor localization,thus possibly neglecting the effects of coil inhomogeneity. Third, the sam-ple size of our study, especially of the tumors, was relatively small andheterogeneous. However, the results were already significant in this smallcohort, so that similar studies would be expected in larger studies.

Furthermore, we did not evaluate interreader or intrareader agreement.Principally, although not being the focus of our study, the influence ofdifferent regions on the results of MR renography is an interesting aspect,which has already been investigated in previous studies.51 In our study,ROIswere segmented semiautomatically and only voxels outside the organwere removedmanually.We think that differenceswould be relatively smalland probably not decisive for the final results. Finally, ROIs were definedby threshold, which could constitute a potential confounder by automati-cally selecting vital regions of the tumor and thus providing similar results.Still, we preferred semiautomatic ROI definition to a manual approach toreliably exclude necrotic tumor tissue.

To conclude, a 3-dimensional whole-organ/tumor analysis ismore precise than single-slice assessment in DCE-MRI of the kidney.Therefore, 3D measurement will be preferable in most patients whoundergo DCE-MRI. The potential use of whole-tumor DCE-MRI todepict the effects of antiangiogenic therapy has to be assessed in futurestudies.

FIGURE 6. Box-and-whisker plots. Chi-square test calculated the goodness of fit of derived functional data to the curveprovided by the 2-compartment-models. A, Mean results were lower for whole kidney (0.30% T 0.18%) than forsingle-slice evaluation (0.43% T 0.26%). Consequently, the curves for whole kidneys were significantly better fitted thanfor single-slice curves (P G 0.01). B, Tumor W

2 for whole tumor (2.74% T 2.88%) was smaller than for single slice(4.05% T 4.79%), but not significantly.

TABLE 6. Data of Tumor Perfusion With Respect to the Histopathological Classification

Clear Cell RCC Chromophobe RCC Papillary RCC

FP VP FP VP FP VP

Single slice 124.6 T 100.8 23.0 T 13.0 100.6 T 6.2 12.9 T 1.3 47.7 T 49.7 12.9 T 4.0

Whole tumor 123.9 T 104.6 23.3 T 12.5 92.4 T 16.9 12.4 T 1.8 45.2 T 49.9 11.2 T 2.5

P 0.99 0.96 0.58 0.78 0.96 0.66

Mean value, standard deviation, and significant paired t-test P values are shown.

RCC indicates renal cell carcinoma; FP, plasma flow (mL/100 mL/min); VP, plasma volume (mL/100 mL).




REFERENCES1. Winter JD, St Lawrence KS, Cheng HL. Quantification of renal perfusion: com-

parison of arterial spin labeling and dynamic contrast-enhancedMRI. J Magn ResonImaging. 2011;34:608Y615.

2. Tofts PS, Cutajar M, Mendichovszky IA, et al. Precise measurement of renalfiltration and vascular parameters using a two-compartment model for dynamiccontrast-enhanced MRI of the kidney gives realistic normal values. Eur Radiol.2012;22:1320Y1330.

3. Notohamiprodjo M, Reiser MF, Sourbron SP. Diffusion and perfusion of thekidney. Eur J Radiol. 2010;76:337Y347.

4. SadickM, SchockD,KraenzlinB, et al.Morphologic and dynamic renal imagingwithassessment of glomerular filtration rate in a pcy-mousemodel using a clinical 3.0 Teslascanner. Invest Radiol. 2009;44:469Y475.

5. Notohamiprodjo M, Sourbron S, Staehler M, et al. Measuring perfusion and per-meability in renal cell carcinomawith dynamic contrast-enhancedMRI: a pilot study.J Magn Reson Imaging. 2010;31:490Y501.

6. PalmowskiM,Schifferdecker I, ZwickS, et al. Tumor perfusion assessed by dynamiccontrast-enhanced MRI correlates to the grading of renal cell carcinoma: initial re-sults. Eur J Radiol. 2010;74:e176Ye180.

7. Notohamiprodjo M, Pedersen M, Glaser C, et al. Comparison of Gd-DTPA andGd-BOPTA for studying renal perfusion and filtration. J Magn Reson Imaging.2011;34:595Y607.

8. Braunagel M, Radler E, Ingrisch M, et al. Perfusion and permeability DCE-MRImeasurements in renal cell carcinoma and metastases: effect of ROI size and posi-tioning on inter- and intraobserver variability. Paper presented at: European Societyof Magnetic Resonance in Medicine and Biology (ESMRMB), Abstract 183,Lisbon, PT: October 4Y6, 2012.

9. Bokacheva L, Rusinek H, Zhang JL, et al. Assessment of renal function withdynamic contrast-enhanced MR imaging. Magn Reson Imaging Clin North Am.2008;16:597Y611, viii.

10. Cutajar M,Mendichovszky IA, Tofts PS, et al. The importance of AIF ROI selectionin DCE-MRI renography: reproducibility and variability of renal perfusion and fil-tration. Eur J Radiol. 2010;74:e154Ye160.

11. BuckleyDL, ShurrabAE, CheungCM, et al.Measurement of single kidney functionusing dynamic contrast-enhanced MRI: comparison of two models in human sub-jects. J Magn Reson Imaging. 2006;24:1117Y1123.

12. Michaely HJ, Sourbron S, Dietrich O, et al. Functional renal MR imaging: anoverview. Abdom Imaging. 2007;32:758Y771.

13. Song T, Laine AF, Chen Q, et al. Optimal k-space sampling for dynamic contrast-enhanced MRI with an application to MR renography. Magn Reson Med.2009;61:1242Y1248.

14. Renken NS, Krestin GP. Magnetic resonance imaging of the kidney. SeminUltrasound CT MR. 2005;26:153Y161.

15. Notohamiprodjo M, Glaser C, Horng A, et al. Dynamic 3D-MR-angiographyfor assessing rheumatoid disease of the handVa feasibility study. Eur J Radiol.2012;81:951Y956.

16. Shen W, Chen J, Gantz M, et al. A single MRI slice does not accurately predictvisceral and subcutaneous adipose tissue changes during weight loss. Obesity(Silver Spring, Md). 2012;20:2458Y2463.

17. So R, Sasai H, Matsuo T, et al. Multiple-slice magnetic resonance imaging candetect visceral adipose tissue reduction more accurately than single-slice im-aging. Eur J Clin Nutr. 2012;66:1351Y1355.

18. Attenberger UI, Ingrisch M, Dietrich O, et al. Time-resolved 3D pulmonary per-fusion MRI: comparison of different k-space acquisition strategies at 1.5 and 3 T.Invest Radiol. 2009;44:525Y531.

19. Lim RP, Shapiro M, Wang EY, et al. 3D time-resolved MR angiography (MRA) ofthe carotid arteries with time-resolved imaging with stochastic trajectories: compar-ison with 3D contrast-enhanced Bolus-Chase MRA and 3D time-of-flight MRA.AJNR Am J Neuroradiol. 2008;29:1847Y1854.

20. Nael K, Krishnam M, Ruehm SG, et al. Time-resolved MR angiography in theevaluation of central thoracic venous occlusive disease. AJR Am J Roentgenol.2009;192:1731Y1738.

21. Sourbron S, Biffar A, Ingrisch M, et al. PMI: platform for research in medicalimaging. Magn Reson Mater Phy. 2009;22:539.

22. IngrischM, Sourbron S. Tracer-kineticmodeling of dynamic contrast-enhancedMRIand CT: a primer. J Pharmacokinet Pharmacodyn. 2013;40:281Y300.

23. Luypaert R, Sourbron S, Makkat S, et al. Error estimation for perfusion parametersobtained using the two-compartment exchange model in dynamic contrast-enhancedMRI: a simulation study. Phys Med Biol. 2010;55:6431Y6443.

24. Notohamiprodjo M, Staehler M, Steiner N, et al. Combined diffusion-weighted,blood oxygen level-dependent, and dynamic contrast-enhanced MRI for characteri-zation and differentiation of renal cell carcinoma. Acad Radiol. 2013;20:685Y693.

25. Sourbron SP, Buckley DL. Tracer kinetic modelling in MRI: estimating perfusionand capillary permeability. Phys Med Biol. 2012;57:R1YR33.

26. Sourbron SP, Buckley DL. Classic models for dynamic contrast-enhancedMRI. NMR Biomed. 2013;26:1004Y1027.

27. Attenberger UI, Sourbron SP, Notohamiprodjo M, et al. MR-based semi-automatedquantification of renal functional parameters with a two-compartment modelVaninterobserver analysis. Eur J Radiol. 2008;65:59Y65.

28. GutierrezDR,Wells K, DiazMontesdeocaO, et al. Partial volume effects in dynamiccontrast magnetic resonance renal studies. Eur J Radiol. 2010;75:221Y229.

29. Chen JJ, Smith MR, Frayne R. The impact of partial-volume effects in dynamicsusceptibility contrast magnetic resonance perfusion imaging. J Magn ResonImaging. 2005;22:390Y399.

30. Michaely HJ, Schoenberg SO, Oesingmann N, et al. Renal artery stenosis:functional assessment with dynamic MR perfusion measurementsYfeasibilitystudy. Radiology. 2006;238:586Y596.

31. Michaely HJ, Sourbron SP, Buettner C, et al. Temporal constraints in renal perfusionimaging with a 2-compartment model. Invest Radiol. 2008;43:120Y128.

32. Feng L, Grimm R, Tobias Block K, et al. Golden-angle radial sparse parallel MRI:combination of compressed sensing, parallel imaging, and golden-angle radialsampling for fast and flexible dynamic volumetric MRI. Magn Reson Med. 2013.[Epub ahead of print].

33. Zollner FG,Weisser G, ReichM, et al. UMMperfusion: an open source software tooltowards quantitative MRI perfusion analysis in clinical routine. J Digit Imaging.2013;26:344Y352.

34. Murakami M, Zhao S, Zhao Y, et al. Evaluation of changes in the tumor microenvi-ronment after sorafenib therapy by sequential histology and 18F-fluoromisonidazolehypoxia imaging in renal cell carcinoma. Int J Oncol. 2012;41:1593Y1600.

35. CoenegrachtsK,BolsA,HaspeslaghM, et al. Prediction andmonitoring of treatmenteffect using T1-weighted dynamic contrast-enhancedmagnetic resonance imaging incolorectal liver metastases: potential of whole tumour ROI and selective ROI anal-ysis. Eur J Radiol. 2012;81:3870Y3876.

36. Reiner CS, Roessle M, Thiesler T, et al. Computed tomography perfusionimaging of renal cell carcinoma: systematic comparison with histopathologicalangiogenic and prognostic markers. Invest Radiol. 2013;48:183Y191.

37. Helck A, Wessely M, Notohamiprodjo M, et al. CT perfusion technique forassessment of early kidney allograft dysfunction: preliminary results. EurRadiol. 2013;23:2475Y2481.

38. Chen C, Liu Q, Hao Q, et al. Study of 320-slice dynamic volume CT perfusionin different pathologic types of kidney tumor: preliminary results. PloS One.2014;9:e85522.

39. Chandarana H, Kang SK, Wong S, et al. Diffusion-weighted intravoxel inco-herent motion imaging of renal tumors with histopathologic correlation. InvestRadiol. 2012;47:688Y696.

40. Lanzman RS, Ljimani A, Pentang G, et al. Kidney transplant: functional assessmentwith diffusion-tensor MR imaging at 3 T. Radiology. 2013;266:218Y225.

41. GloviczkiML,Glockner J,GomezSI, et al. Comparison of 1.5 and 3TBOLDMR tostudy oxygenation of kidney cortex and medulla in human renovascular disease.Invest Radiol. 2009;44:566Y571.

42. Ebrahimi B, Gloviczki M, Woollard JR, et al. Compartmental analysis of renalBOLD MRI data: introduction and validation. Invest Radiol. 2012;47:175Y182.

43. Helck A, D’Anastasi M, Notohamiprodjo M, et al. Multimodality imagingusing ultrasound image fusion in renal lesions. Clinical Hemorheol Microcirc.2012;50:79Y89.

44. HelckA, SommerWH,WesselyM, et al. Benefit of contrast enhanced ultrasound fordetection of ischaemic lesions and arterio venous fistulas in renal transplantsVafeasibility study. Clinical Hemorheol Microcirc. 2011;48:149Y160.

45. Wu WC, Su MY, Chang CC, et al. Renal perfusion 3-T MR imaging: a com-parative study of arterial spin labeling and dynamic contrast-enhanced tech-niques. Radiology. 2011;261:845Y853.

46. Heusch P, Wittsack HJ, Heusner T, et al. Correlation of biexponential diffusion pa-rameters with arterial spin-labeling perfusion MRI: results in transplanted kidneys.Invest Radiol. 2013;48:140Y144.

47. Heusch P, Wittsack HJ, Blondin D, et al. Functional evaluation of transplantedkidneys using arterial spin labeling MRI. J Magn Reson Imaging. 2013.

48. Lanzman RS, Robson PM, Sun MR, et al. Arterial spin-labeling MR imagingof renal masses: correlation with histopathologic findings. Radiology. 2012;265:799Y808.

49. Teh HS, Ang ES, Wong WC, et al. MR renography using a dynamic gradient-echo sequence and low-dose gadopentetate dimeglumine as an alternative toradionuclide renography. AJR Am J Roentgenol. 2003;181:441Y450.

50. Vivier PH, Storey P, Rusinek H, et al. Kidney function: glomerular filtrationrate measurement with MR renography in patients with cirrhosis. Radiology.2011;259:462Y470.

51. Goh V, Halligan S, Gharpuray A, et al. Quantitative assessment of colorectal cancertumor vascular parameters by using perfusion CT: influence of tumor region of in-terest. Radiology. 2008;247:726Y732.




single slice versus whole organ/tumorclinical-mri.com/wp-content/uploads/2014/11/...thus, 43 kidneys...

Documents