multimodal optical imaging of microvessel network convective oxygen transport dynamics

Multimodal optical imaging of microvessel networkconvective oxygen transport dynamics

Casey deDeugd, Mamta Wankhede, and Brian S. Sorg*J. Crayton Pruitt Family Department of Biomedical Engineering, University of Florida, Gainesville, Florida 32611, USA

*Corresponding author: [email protected]

Received 2 September 2008; revised 17 December 2008; accepted 26 January 2009;posted 27 January 2009 (Doc. ID 100736); published 2 March 2009

Convective oxygen transport bymicrovessels depends on several parameters, including red blood cell fluxand oxygen saturation. We demonstrate the use of intravital microscopy techniques to measure hemo-globin saturations, red blood cell fluxes and velocities, and microvessel cross-sectional areas in regions ofmicrovascular networks containing multiple vessels. With these methods, data can be obtained at highspatial and temporal resolution and correlations between oxygen transport and hemodynamic para-meters can be assessed. In vivo data are presented for a mouse mammary adenocarcinoma grown ina dorsal skinfold window chamber model. © 2009 Optical Society of America

OCIS codes: 180.0180, 170.2655, 170.3880.

1. Introduction

Oxygen is required by mammalian cells for normalaerobic metabolism to produce energy for a cellularfunction [1]. Capillaries are not the only site of oxy-gen exchange [2], as significant exchange also occursin precapillary arterioles [3] and postcapillary ve-nules [4]. The complexity of microvascular oxygentransport and exchange with tissue is such that tothis day the process is not completely understood[5,6]. There are a number of areas where a better un-derstanding of microvascular oxygen transport maylead to advances. For example, abnormal microcircu-latory function and oxygen transport is associatedwith several pathological conditions, including dia-betes [7,8], hypertension [9,10], hemorrhagic shock[11,12], wound healing [13,14], and solid tumors[15–17]. Studies of normal physiologic function, suchas functional activation of specific brain regions [18],can also benefit from further knowledge of microvas-cular oxygen transport function. Experimental stu-dies, theoretical models, and their combination willplay a role in these research areas.

Convective oxygen transport by the microcircula-tion can be mathematically described as a combina-tion of parameters including red blood cell (RBC)oxygen carrying capacity and saturation, and bloodflow measurements [6,19]. Hemoglobin saturation,hematocrit, and blood flow rate can be related to con-vective oxygen transport in a blood vessel per thefollowing equation by Secomb et al. [19]:

QoxðPO2Þ ¼ Qbl½HC0;RBCSHbðPO2

Þ þ αeffPO2�; ð1Þ

where QoxðPO2Þ is the rate of oxygen transfer

(cm3O2=s) and is a function of the blood partial pres-sure of oxygen (PO2

, mmHg); Qbl is the blood flowrate ðcm3=sÞ; H is the hematocrit (dimensionless);C0;RBC is the oxygen carrying capacity of a RBC,which is the amount of oxygen bound in a fully satu-rated RBC ðcm3O2=cm3Þ; SHb is fractional hemoglo-bin saturation (dimensionless), which is a function ofblood PO2

, and αeff is the effective solubility of oxygenin blood ðcm3 O2 cm−3 mmHg−1Þ, which is generallytaken as a constant. This equation includes theamount of dissolved oxygen in the plasma, whichis usually only a very small component of the amountof oxygen transported by the microcirculation rela-tive to the amount transported by RBCs at atmo-spheric pressures. The microvascular hematocrit

0003-6935/09/10D187-11$15.00/0© 2009 Optical Society of America

1 April 2009 / Vol. 48, No. 10 / APPLIED OPTICS D187

term in Eq. (1) can be obtained from the followingrelation [20,21]:

H ¼ VRBCFRBC

πR2vavg; ð2Þ

where VRBC is the average volume of RBCs (cubiccentimeters), FRBC is the RBC flux (number of RBCsper second), R is the radius of the blood vessel (cen-timeters) , and vavg is the average blood flow velocity(centimeters per second). A similar equation that de-scribes microvascular convective oxygen transportgiven by Pittman [6] that does not include dissolvedoxygen in the plasma is the following:

QoxðPO2Þ ¼ πR2vavg½Hb�SHbðPO2

ÞC0;Hb; ð3Þ

where ½Hb� is the concentration of hemoglobin(V=V ; cm3=cm3), and C0;Hb is the amount of oxygenbound in a fully saturated unit volume of hemoglobinðcm3 O2=cm3Þ. Note that, despite their having thesame units, there is a subtle yet important distinc-tion between C0;Hb in Eq. (3) and C0;RBC in Eq. (1):C0;RBC refers to a single RBC and is normalized tothe volume of the RBC, while C0;Hb refers to hemoglo-bin volume in the context of hemoglobin concentra-tion. If the dissolved oxygen term in Eq. (1) isneglected, then effectively Eqs. (1) and (3) are equiva-lent in that they both describe the volume of oxygenbeing transported in the blood vessel per unit time interms of the volume of hemoglobin being transportedper unit time and the amount of oxygen being carriedby the hemoglobin.A number of optical techniques capable of micro-

vascular oxygenation measurement have been devel-oped, including spectroscopic methods [22–27] andphosphorescence quenching techniques [23,28–31];however, measurement of microvascular convectiveoxygen transport requires knowledge of additionalhemodynamic and functional parameters besidesoxygenation. Researchers have employed intravitalmicroscopy imaging techniques to determine convec-tive oxygen transport in microvessel segments andbranches. In a notable example, Swain and Pittmanmeasured convective oxygen transport in individualmicrovessel branches in hamster cheek pouch retrac-tor muscles to identify longitudinal gradients of oxy-gen in precapillary arterioles, thus demonstratingdiffusion of oxygen out of these vessels into the sur-rounding tissue [32]. In this paper, we endeavored toidentify a particular combination of intravital micro-scopy optical techniques to simultaneously capturethe relevant parameters necessary to measure con-vective oxygen transport from a microvascular net-work region such that high spatial and temporalresolution and correlations of oxygen transport couldbe achieved across the network. Three different op-tical techniques were employed: wide-field spectro-scopic measurements for microvessel hemoglobinsaturation, wide-field fluorescence measurementsfor RBC flux and velocity, and confocal fluorescence

measurements for a microvessel cross-sectional area.The measured parameters were used with thepreviously described equations to calculate the con-vective oxygen transport throughout a microvesselnetwork. The abnormal oxygen transport in tumorsprovided the motivation for this research; thus thein vivo model chosen was tumors grown in mousedorsal skinfold window chambers. Furthermore, theadvantages and limitations of the method presentedin this study are discussed.

2. Materials and Methods

A. Tumor Cells

4TO7 mouse mammary adenocarcinoma cells, anonmetastatic subclone of the 4T1 cell line, were cul-tured as a monolayer in DMEM (Dulbecco’s ModifiedEagles’s Medium, Mediatech, Manassas, Virginia)with 10% fetal bovine serum (Mediatech, Manassas,Virginia). The tumor cells were a gift from MarkW. Dewhirst (Duke University, Durham, NorthCarolina). Cultures were used after one or two pas-sages from frozen stocks to ensure recovery from thethermal shock and a normal growth rate. The cellswere enzymatically dissociated from the flasks(BD Bioscience, San Jose, California) by using0.05% trypsin/EDTA (ethylenediaminetetraaceticacid, Mediatech, Manassas, Virginia) to preparesingle-cell suspensions. Cells were counted via a he-macytometer to determine the cell concentration.

B. Preparation of Fluorescent Red Blood Cells

RBCs were fluorescently labeled by using amodification of the procedure by Unthank et al.[33]. RBCs obtained from donor mice were labeledwith a 1mg=ml stock solution of Carbocyaninedye 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicar-bocyanine, 4-chlorobenzenesulfonate salt (DiD solid,Invitrogen, D-7757) dissolved in ethanol. DiDðexcitation 644nm=emission 665nmÞ was chosen asthe lipid membrane labeling solution because it is ananalog of the commonly used DiI dye ðex 549nm=em 565nmÞ, but with a markedly redshifted flu-orescence excitation and emission spectrum tominimize interference of RBC flux measurementswith hemoglobin saturation measurements ð500–575nmÞ. RBCs obtained by cardiac puncture froma donor mouse were washed twice via centrifugationand resuspension in phosphate buffered saline(PBS). The cells were labeled by adding 100 μl of cellsand 100 μl of DiD stock solution to 10ml of sterilePBS. The solution was incubated at room tempera-ture for 30min and agitated every 10min to ensuresuspension of the RBCs. After labeling, the RBCswere washed to remove unbound dye and resus-pended in PBS. Immediately prior to an imaging ses-sion, a 50 μl bolus of packed labeled cells in salinesolution (30% V=V) was administered to the mouseto be imaged by tail vein injection. An aliquot ofDiD labeled cells was saved for flow cytometric con-firmation of the efficacy of the labeling procedure.

D188 APPLIED OPTICS / Vol. 48, No. 10 / 1 April 2009

C. Animal Model and Imaging

All in vivo procedures were conducted in accordancewith a protocol approved by the University of FloridaInstitutional Animal Care and Use Committee. A to-tal of seven athymic ðnu=nuÞ female nude miceweighing at least 21 g (Charles River Laboratories,Raleigh, North Carolina) were surgically implantedwith a titanium window chamber under anesthesia(ketamine 100mg=kg IP and xylazine 10mg=kg IP)on the dorsal skin flap. Tumors were establishedat the time of window chamber surgery from a10 μL single-cell suspension of 5 × 103 to 10 × 103

4TO7 cells injected into the subcutaneous tissue im-mediately prior to placing a 12mm round glass cover-slip over the exposed area of the skin. Animals werehoused postsurgery in an environmental chambermaintained at 33°C and 50% humidity with free ac-cess to food and water and standard 12 h light–darkcycles. Experiments were conducted after the tumorhad been well established, with three of the five an-imals being imaged twice. Tumors suitable for ourexperiments were typically obtained 8–13 days afterimplantation.For spectral and RBC flux imaging, mice were

placed on the microscope stage on a heating pad tomaintain normal body temperature and were an-esthetized by 1%–2% isoflurane in air. In two experi-ments, an increase in convective oxygen transportwas induced during imaging by a change in breath-ing gases from room air to 100% oxygen.

D. Spectral Imaging

The spectral imaging system, image acquisition, andimage processing methods for hemoglobin saturationmeasurements were discussed in detail previously[26]. Briefly, one hemoglobin saturation image setcomprised 16 images acquired in the wavelengthrange of 500–575nm with an interval of 5nm. Auto-mated spectral image acquisition was performed byusing customized LabVIEW software (National In-struments, Austin, Texas). A Zeiss microscope (CarlZeiss, Inc., Thornwood, New York) was used as theimaging platform. For transillumination of windowchambers, a 100W tungsten halogen lamp was used.Images were obtained at 1380 × 1035 pixels and12 bit dynamic range by using a CCD camera ther-moelectrically cooled to −20°C (DVC Company,Austin Texas, Model 1412AM-T2-FW). The long-working-distance objectives used were 2:5× and 5×Fluars, 10× EC Plan-NeoFluar, and a 20× LD-Plan-NeoFluar (Carl Zeiss, Inc., Thornwood, New York). Aliquid crystal tunable filter (CRI, Cambridge, Massa-chusetts) with a 400–720nm transmission range anda 10nm nominal bandwidth placed in front of thecamera provided band-limited images. Images weresaved as 16 bit TIFF files. For image processing, all16 images per set were converted to double-precisionarrays in MATLAB (The Mathworks Inc., Natick,Massachusetts). Using the linear least squaresmethod proposed by Shonat et al. [23], we convertedimages on a pixelwise basis to relative values of

hemoglobin saturation. Hemoglobin saturation pseu-docolor maps of the microvessel networks were cre-ated from the spectral image data by a linear leastsquares regression fit of a model of the microvesselabsorbance to the data using pure oxyhemoglobinand deoxyhemoglobin reference spectra. Regions ofinterest for hemoglobin saturation measurementswere selected on the basis of the proximity of the mi-crovessel region to areas selected for RBC flux mea-surements. Customized LabVIEW software enabledautomated image acquisition using the specificationsfor camera exposure time and gain for each filterwavelength. Given that the liquid crystal tunable fil-ter transmits less at shorter wavelengths and moreat longer wavelengths, the exposure time for thecamera was controlled such that the full dynamicrange of the camera was utilized. The minimum ex-posure time used was 400ms for the longest wave-length, whereas the maximum exposure time usedwas 1400ms for the shortest wavelength, resultingin a typical acquisition time of approximately 16 sfor image acquisition, filter tuning, image transfer,and saving images on an external hard drive.

E. RBC Flux Imaging

Fluorescently labeled RBCs were imaged via stream-ing video using an Andor iXon electron multiplyingCCD camera (Andor Technology, South Windsor,Connecticut). A Cy5 filter set was used (ChromaTechnology Corp., Rockingham, Vermont, excitation640nmwith 20nm bandwidth, emission 680nmwith30nm bandwidth) in line with the illuminationsource of a Zeiss FluoArc mercury lamp. To optimizethe resolution of the fast-moving cells, data were ac-quired in kinetic acquisition mode with 2 × 2 binning,using an exposure time of 16:2ms, a shift speed of0:564 μs, and a frequency of ∼30Hz to ensure thatimages would be captured with sufficient temporalresolution. Internal triggering was used to spool datadirectly to the computer hard drive, and 20 s ofstreaming video was saved every minute for 1 h incoordination with the acquisition of spectral imagingdata sets. Each frame was saved as a tagged imageformat file (TIFF), resulting in a stack of 600 TIFFsper data point. RBC flux was determined by firstidentifying a region on a vessel of interest whereflowing RBCs appeared to be in good focus and thenmanually counting the number of labeled cells flow-ing past the designated location on the vessel overthe 20 s time interval of data acquisition for the timepoint. These measurements were converted to flux bycorrecting for the fraction of labeled RBCs versus un-labeled RBCs in the mouse blood. The fraction of la-beled RBCs was determined by flow cytometry of ablood sample obtained postimaging via retro-orbitalpuncture performed on the imaged mouse. The RBCflux was then calculated by using Eq. (4):

FRBC ¼ NRBC

t

�NRBC;fluorNRBC;total

� ; ð4Þ


where NRBC is the number of fluorescent RBCscounted flowing in the vessel, t is the time intervalin which NRBC was counted (20 s in this case), andNRBC;fluor andNRBC;total are the number of fluorescentand total RBCs, respectively, counted by flow cytome-try in the postexperiment blood sample.RBC velocity was determined from the time re-

quired for an individual RBC to travel a specific dis-tance in the blood vessel that passed through theregion of interest. The time interval was determinedfrom the frame rate of fluorescence video imaging.The mean velocity of 5–10 randomly selected RBCsfrom each time point was taken as vavg.

F. Laser Scanning Microscopy

After hyperspectral imaging sessions, animals weremoved to an Olympus IV100 laser scanning micro-scope (LSM). Before imaging, mice were anesthetizedusing injectable anesthesia (ketamine 100mg=kg IPand xylazine 10mg=kg IP), and placed on a heatingpad. A 4mg=ml stock solution of β-phycoerythrin (P-800, Invitrogen, Carlsbad, California) was diluted toapproximately 18.7% V=V in sterile saline in order toobtain a weight of 0:075mg per 100 μl bolus, which isthe weight/volume used in a protocol obtained fromBrizel et al. [20], and 50–100 μl of this solution wasdelivered via tail vein injection. β-phycoerythrin waschosen to illuminate the blood vessels because it is aplasma binding dye that has a very broad and brightemission spectrum (ex 542nm, em 550–700nm, withpeak λ ¼ 575nm). This was suitable for the lasersource that was available for the LSM, which hasa solid-state laser tuned to 561nm that was usedfor excitation. This wavelength closely coincides witha secondary excitation peak of β-phycoerythrin.Images of areas that had been analyzed via hyper-spectral imaging and RBC flux were used as a refer-ence to locate the same regions with the LSM. Oncethe regions were located visually, 3D automatedscanning was performed by acquiring x − y imagesof the 10× objective area and stepping through thez direction in 5–10 μm step sizes. While a maximalimage depth of about 300 μm could reasonably be ex-pected with the LSM system, imaging was performedonly to a depth of 50–100 μm which included the toptumor vessel layer in the window chamber. The mainreason for the limitation in image depth was thatdata acquisition at high axial spatial resolution in-creased imaging time, which increased the probabil-ity that a motion artifact that could ruin the data setmight occur during imaging. Despite the fact that thewindow chamber was firmly secured to the imagingplatform, the skin could still move within the windowchamber owing to a reflexive muscle twitch or a deepbreath by the animal. Additional restraints on ani-mal and skin movement during imaging could possi-bly enable imaging to a greater depth.Data obtained via LSM were originally saved to

the proprietary Olympus software as 16 bit TIFFfiles that were subsequently converted to 8 bit byusing ImageJ software. The 8 bit images were then

loaded into Image Surfer, a 3D image reconstructionfreeware program for visualization and analysis of3D images and data sets [34]. The software uses de-convolution techniques to provide volume renderingof image stacks and allows user manipulation of thevolume in the x, y, and z dimensions.

3. Results

Absorption from the fluorescent dye used to label theRBCs could potentially affect the calculation of he-moglobin saturation if the fraction of labeled cellsis high enough. In this case, the dye absorption wouldhave to be accounted for in the calculation. As shownin Fig. 1, the effect of absorption from the RBC fluor-escent labeling dye on calculations of hemoglobin sa-turation was determined for various fractions oflabeled RBCs with 100% oxygen saturation flowingin glass capillary tubes. To observe the effect, dyeabsorption was not accounted for in the hemoglobinsaturation calculation. It can be seen in Fig. 1 that, atRBC labeling fractions less than 5%, the effect of dyeabsorption is negligible, while larger labeled frac-tions cause progressively more interference. In thisstudy, the fraction of labeled RBCs in the mice wasin the range of 1%–3% as determined by flow cytome-try of blood samples taken from themice at the end ofeach experiment; so the labeled RBCs had a negligi-ble effect on the hemoglobin saturation calculation.

Figures 2–9 illustrate how the multimodality im-age data (Figs. 2–6) were used to obtain microvesselfunctional and hemodynamic data (hemoglobinsaturation and RBC flux in Figs. 7 and 8) for the cal-culation of microvessel convective oxygen transport(Fig. 9). A transmitted light image of an example tu-mor microvessel network is shown in Fig. 2, and acorresponding hemoglobin saturation map for the

Fig. 1. Plot of apparent hemoglobin saturation of different frac-tions of RBCs labeled with DiD. The absorption of the fluorescentdye is not taken into account in the hemoglobin saturation calcu-lation in order to observe the effect of the fluorescent dye absorp-tion. The data points are the mean� standard deviation of aregion of pixels in an image of RBCs with 100% hemoglobin satura-tion in a glass capillary tube. The standard deviation for the twolowest labeled fractions of RBCs is less than 1%.


network is shown in Fig. 3. In Fig. 4 is an image offlowing fluorescently labeled RBCs flowing in the mi-crovessel network. Data regarding dynamic changesin hemoglobin saturation, RBC flux, and blood flowwere derived fromdata sets with the images in Figs. 3and 4. Confocal image data (Fig. 5) were acquired onan Olympus IV100 system after 1 h of imaging of he-moglobin saturation and RBC flux. A 10× objectivewas used for imaging in both microscope systems.Since data for Figs. 3–5 were acquired in two differ-ent imaging systems, relocation of the animal was re-quired, but it was still possible to locate the originalimaged region. The confocal data were used to mea-sure the cross-sectional area of the microvessel re-gions of interest for analysis. In Fig. 6 is a 3Dreconstructed confocal image stack of the area im-aged in Figs. 2–5 showing a cross-sectional areaacross the dashed line in Fig. 5. The cross-sectionalareas marked in the figure are for the regions of in-terest marked 1 (481 μm2) and 2 (324 μm2) in Fig. 2.

Figures 7 and 8 show example plots of RBC flux andhemoglobin saturation over time for regions of inter-est 1 and 3 from Fig. 2. It can be seen from the figuresthat trends in the hemoglobin saturation dynamic

Fig. 2. Transmitted light image of the tumor microvessel net-work. White triangles represent regions of interest chosen for ana-lysis. A 10× objective was used for imaging (NA of 0.3, workingdistance of 5:5mm).

Fig. 3. (Color online) Hemoglobin saturation image of the tumormicrovessel network in Fig. 2. The pixels are colored according tothe hemoglobin saturation scale to the right of the figure. Thebackground is black. Regions of interest for analysis are indicatedin the figure.

Fig. 4. Fluorescence image of RBCs labeled with DiD flowing inthe tumor microvessel network. Images were captured with anelectron multiplying CCD camera at video rates (∼30Hz). 20 sof data were taken every minute of the hour-long imaging session,and cells were counted over this time increment and then con-verted to RBC flux measurements. The locations of the regionsof interest for analysis are indicated in the figure.

Fig. 5. (Color online) Confocal z-stack projection of the tumor mi-crovessel network in Fig. 2. A total of 49 images were acquired at5 μm intervals to a total depth of 245 μm by using an OlympusIV100 LSM. Regions of interest for analysis are indicated. Thedashed line indicates the location of the cross-sectional area inFig. 6.


changes in general are similar to trends in RBC fluxfluctuations. An example of the oxygen transportparameters that can be measured with this techni-que is demonstrated with the vessels marked withregion of interest numbers 1 and 2 in Fig. 2. Thetime-averaged RBC velocity for vessels 1 and 2, re-spectively, is 442 and 98 μm=s. Using an approximateRBC volume of 40 μm3 derived from mouse hemato-logical data [35] with measured RBC flux, vesselcross-sectional area, and RBC velocity data inEq. (2), the time averaged hematocrits for vessels 1and 2 are 0.14 and 0.08, respectively. Taking themaximal oxygen carrying capacity of RBC [C0;RBCin Eq. (1)] as 0:5 cm3 O2=cm3 [32], then the measureddata were used with the previously defined equa-tions to obtain the convective oxygen transport overtime in the two vessels as plotted in Fig. 9. Fourieranalysis was performed on the time series data inFig. 9, and the mean-subtracted normalized powerspectra of the convective oxygen transport data areshown in Fig. 10. The dominant fluctuation fre-quency is less than 0:2 cycles=min, which isconsistent with previously measured 4T1 tumor he-moglobin saturation fluctuation data obtained withour imaging system [25] and reported pO2 fluctua-

tions in a variety of tumor types measured withoxygen microelectrodes ([36–40]).

Perturbations to convective oxygen transport canbe measured with this technique. For the tumor net-work shown in Fig. 11, the mouse initially wasbreathing normal air during the imaging sessionand after 20 min was switched to 100% oxygenbreathing. The tumor microvessel network with indi-cated regions of interest and a graph of convectiveoxygen transport over time are shown, respectively,in Figs. 11 and 12. An increase in the averageconvective oxygen transport with 100% oxygen

Fig. 6. (Color online) Microvessel cross-sectional areas were mea-sured from confocal image stack data. The cross-sectional area wascalculated using the number of pixels in the vessel cross sectionand the dimensions of a pixel in the image plane. The dashed linein Fig. 5 indicates the location of the cross-sectional area.

Fig. 7. Plot of RBC flux versus hemoglobin saturation (HbSat) forregion of interest 1 in Fig. 2. The data points were acquired at1 min intervals for 1h.

Fig. 8. Plot of RBC flux versus hemoglobin saturation (HbSat) forregion of interest 3 in Fig. 2. The data points were acquired at1 min intervals for 1h.

Fig. 9. Convective oxygen transport over time for regions of inter-est 1 and 2 in Fig. 2. Note that the values for region of interest 2 areplotted as 10× their actual value for clarity. The oxygen transportin these vessels tends to fluctuate together, although vessel 1 istransporting about 10× more oxygen than vessel 2. It should benoted that there was difficulty in imaging RBCs at the 6 min timepoint for vessel 2; so the RBC flux was interpolated for that datapoint.


breathing was measured (see Table 1). Unlike ROI 1,the increase in average convective oxygen transportin ROI 2 was not statistically significant. This maybe a result of the increase in magnitude of the fluc-tuations that occurred in ROI 2 after oxygen breath-ing (air breathing standard deviation 1:0 cm3 O2=s,100% breathing standard deviation 1:8 cm3 O2=s). Anincrease in the magnitude of tumor pO2 fluctuationswith 100% oxygen breathing has been previously re-ported [37]. By way of comparison, the data in Fig. 9for regions of interest 1 and 2 in Figs. 2–6 were di-vided into time periods of 1–30 and 31–60min, andthe average convective oxygen transport for eachtime was period calculated. As is shown in Table 1,there was virtually no change in the average convec-tive oxygen transport in the two time periods whenthemouse breathed room air alone for the duration ofthe experiment.

4. Discussion

Convective oxygen transport in individual microves-sel branches has been measured using opticaltechniques. Swain and Pittman used microspectro-photometry to measure hemoglobin saturation andhemoglobin concentration, a dual detector cross-correlation technique to measure blood velocity, andvideo measurements of vessel diameter to calculatethe vessel cross-sectional area [32]. Using these mea-surements, the authors were able to calculate convec-tive oxygen transport in segments of hamster cheekpouch retractor muscle microvessels. In this study,we sought to expand upon previous techniques by de-veloping a method to enable high spatial and tempor-al resolution and correlations of convective oxygentransport to be achieved across an entire network re-gion rather than interrogating blood vessel segmentsindividually at different points in time. To achievethis goal, a spectral imaging technique developedpreviously [26] was employed to capture hemoglobinsaturation information from the microvessel net-work. As a wide-field imaging modality, this techni-que enables temporal and spatial correlation offluctuations of hemoglobin saturation across the net-work [25]. Wide-field fluorescence imaging was cho-sen for the same reason to obtain RBC flux andvelocity information across the microvessel network.Confocal imaging was used to obtain accurate micro-vessel cross-sectional areas rather than assuming acircular vessel cross section. Confocal imaging alsoenables more comprehensive measurements of mi-crovessel positions within the tissue and relativeto other branches in the network. It should benoted that intraluminal variations in hemoglobinsaturation and RBC flux or velocity are possible[41], and the technique described in this paper doesnot address this issue. Rather, the intent was to

Fig. 10. Normalized power spectra for the convective oxygentransport time series data in Fig. 9 obtained by Fourier analysis.The frequency scale is given in cycles per minute (cpm).

Fig. 11. Transmitted light image of the tumor microvessel net-work for Fig. 12. White triangles represent regions of interest cho-sen for analysis. Arrows indicate blood flow direction in the vesselsaround the regions of interest. A 10× objective was used for ima-ging (NA of 0.3, working distance of 5:5mm).

Fig. 12. Convective oxygen transport over time for regions of in-terest 1 and 2 in Fig. 11. The mouse initially started breathing nor-mal air and was switched to 100% oxygen breathing during theexperiment. The vertical dashed line indicates the point at which100% oxygen breathing was started.


make macroscopic observations across networkregions; thus the measured parameters can be con-sidered average or integrated values within themicrovessels.As can be seen from Fig. 1, the fraction of in vivo

labeled RBCs in these experiments was such that ab-sorption from the fluorescent dye did not have an ef-fect on the calculation of hemoglobin saturation. Ifthis were a concern, it would be relatively easy toaccount for the dye absorption by including it inthe regression equation for calculation of hemoglobinsaturation. It would be necessary to use a dye absorp-tion spectrum that was acquired on the spectralimaging system for this purpose. Figures 2–12 de-monstrate how multimodal optical imaging can beused to acquire imaging data on hemoglobin satura-tion, RBC flux, velocity, and hematocrit, and micro-vessel cross-sectional area that can be used tocalculate convective microvascular oxygen transportin a local network region, in this case a tumor micro-vessel network. Fluctuations in hemoglobin satura-tion in the microvessels were similar to changes inRBC flux, as can be seen in Figs. 7 and 8. Sponta-neous arteriolar vasomotion can cause fluctuationsin blood flow and has been documented in a varietyof normal tissues [42–44]. Acute fluctuations in tu-mor blood flow that contribute to tumor oxygen fluc-tuations have also been documented [36,38,45,46]and are believed to be caused by a combination of fac-tors, including tumor feeding vessel vasomotion, dy-namic changes in RBC rheology and flow resistance,and intussusceptive angiogenesis [36,47]. The corre-lation between RBC flux and hemoglobin saturationin Figs. 7 and 8 was expected, as previous measure-ments of RBC flux and perivascular oxygen tensionmade with microelectrodes have shown such a rela-tionship in tumor microvessels [45]. The measuredtime-averaged RBC velocity, flux, and microhemato-crit values in this study were comparable with thoseobtained for similarly sized microvessels in a ratmammary adenocarcinoma [20].The main advantage of the technique described in

this paper is the ability to perform simultaneousimaging and measurements of microvessel hemoglo-bin saturation and RBC flux across a microvesselnetwork region. This can enable analysis of complexconvective oxygen transport relationships among

vessels in a local network region. Our technique formeasurement of oxygen delivery may potentially becombined with other techniques to measure oxygendeposition or utilization in tissue to give a more com-plete picture of tissue oxygen supply and demand.For example, the fluorescent hypoxia marker EF5can highlight tissue areas that have been exposedto hypoxia [48], and this can be compared with theconvective oxygen transport in the supply vesselnetworks. The fluorescent respiratory metabolitesNADH (reduced nicotinamide adenine dinucleotide)and FAD (flavin adenine dinucleotide) may be usedto compute redox ratios [49] to provide a relativemeasure of tissue oxidative metabolism in responseto oxygen delivery. Phosphorescence lifetime imagingmay be utilized to measure tissue oxygen tension[23,29] to form a more complete picture of oxygen de-livery and deposition in tissue. The combination ofadditional techniques will increase the technicalcomplexity of experimental procedures and may besubject to various limitations. There are several re-search areas where a complete picture of tissue oxy-gen supply and demand may be beneficial. Theseinclude studies of tumor angiogenesis, treatmentwith antiangiogenic agents, and measurement ofchanges in tumor microvessel function [50], investi-gations of vascular pathologies that contribute to cer-ebral ischemia and brain injury [51], and studies ofimpaired oxygen delivery and hypoxia in various re-tinopathies due to pathological angiogenesis [52].

A limitation of the technique is the shallow depththat can be imaged by using spectral imaging for thehemoglobin saturation measurements. This will re-strict investigations to more superficial vessels inmost tissues with the exception of a few specificcases, such as mesentery tissue, which is relativelytransparent and 2D. Another constraint is that theRBC flux measurement technique places a limit onthe distribution of vessel diameters that can be mea-sured and the size of the field of view that can beused. It is difficult to accurately measure RBC fluxin larger vessels because some of the deeper labeledRBCs in a large-diameter vessel may be shielded byshallower RBCs in the vessel and therefore may notbe imaged. Also, larger vessels have larger microhe-matocrits, which makes it difficult to discriminate in-dividual labeled RBCs. The field of view is limited to

Table 1. Comparison of Tumor Microvessel Convective Oxygen Transport in Sequential Time Periods with Air Breathing Alone or Air Breathing and100% Oxygen Breathing

Average Convective Oxygen Transport ðcm3 O2=sÞ (mean� SEMa)

MicrovesselAir, 1–30 min Air, 1–20 min Air, 31–60 min 100% O2, 21–45 minFig. ROI

2 1 8:8 × 10−10 � 0:4 × 10−10b 8:9 × 10−10 � 0:4 × 10−10b

2 2 6:7 × 10−11 � 0:6 × 10−11b 6:8 × 10−11 � 0:5 × 10−11b

11 1 2:8 × 10−10 � 0:2 × 10−10c 3:8 × 10−10 � 0:2 × 10−10c

11 2 6:6 × 10−10 � 0:2 × 10−10b 7:3 × 10−10 � 0:4 × 10−10b

aStandard error of mean.bNot statistically different, t test.cStatistically significantly different, t test, p < 0:01.


a size such that single flowing labeled RBCs can beresolved and imaged reliably. If microvessel dia-meters changed during imaging owing to a changein vascular muscle tone, then the cross-sectionalareasmeasured with confocal microscopymay not re-present the actual value at the previously measuredtime points during spectral and fluorescence ima-ging. However, in tumors, microvessels are abnormalwith little to no smooth muscle or innervation [53]; sothis occurrence is less likely, but in normal tissuesspontaneous vasomotion is common [42–44]. Errorsin the microvessel cross-sectional area can affect cal-culations of microvessel hematocrit and convectiveoxygen transport. However, if Eq. (1) is used for cal-culation of convective oxygen transport with theplasma oxygen transport term neglected, then the ef-fects of microvessel diameter changes can be avoidedif the product of the microvessel cross-sectional areaand average RBC velocity (πR2Vavg) is used to calcu-late volumetric blood flow (Qbl) in Eq. (1), as is effec-tively done in Eq. (3), so the πR2Vavg term will cancelthe denominator in the hematocrit term [H;see Eq. (2)].Two-photon microscopy (TPM) has been used to

measure microvascular oxygenation (pO2) with phos-phorescence lifetime imaging [54], blood velocity[55], and microvessel morphology [56]. TPM also hasthe potential to measure hemoglobin saturation [57].An appropriate combination of these methods mayenable measurement of convective oxygen transportin microvessel networks in 3D with greater imagingdepth than the technique described in this paper. Thescanning beam nature of TPM can limit the spatialand temporal correlations of dynamic changes in con-vective oxygen transport across the network, depend-ing on the time scale of the dynamic events beingmeasured and the image acquisition speed of the in-strument. Amore severe limitation is that TPMmea-surements of blood velocity require scanning of theimaging laser beam parallel to the axis of blood flowin each individual measured microvessel segment,which can be time consuming and impractical inthe case of networks with complex geometries. Wide-field phosphorescence lifetime imaging has beencombined with imaging of fluorescently labeledRBCs [58], thus enabling simultaneous acquisitionof microvessel oxygenation and RBC velocity or fluxacross a network. A problem with phosphorescencelifetime imaging is that it can be difficult to usefor measurements of microvascular convective oxy-gen transport. Recall from Section 1 that convectiveoxygen transport is a measurement of the amount ofoxygen transported by blood flow, which predomi-nantly occurs as oxygen bound to hemoglobin inRBCs and can be determined from measurementsof hemoglobin saturation (percent of maximal hemo-globin oxygen carrying capacity) [6,19]. Phosphores-cence lifetime imaging measures the amount ofdiffusing oxygen ðpO2Þ available to interact with aphosphorescent dye [6]. While hemoglobin satura-tion can be related to ambient pO2 through the

hemoglobin–oxygen dissociation curve, the shapeof the curve is dependent on the local pH and pCO2[1,5]; so unless these factors are somehow known, in-accurate calculations of hemoglobin saturation maybe obtained. This is especially true in tumor micro-environments, which are known to have pathologi-cally acidotic pH values [30,59]. Spectral imaginghas been combined with laser speckle contrast ima-ging for simultaneous wide-field imaging of micro-vascular hemoglobin saturation and blood flow [22].However, laser speckle contrast provides only rela-tive measurements of perfusion [60]. Techniques forabsolute velocity measurements with laser speckleand Doppler techniques are in development but cur-rently not reliable enough for routine use with tissue[61]. Several recent papers provide evidence thatoptical coherence tomography (OCT) may have thepotential to make 3D measurements of microvesselconvective oxygen transport from endogenous sig-nals with a single imaging modality. Doppler OCTcan provide blood velocity information [62,63], andblood flow can be mapped in 3D microvessel net-works [64]. Blood flow measurements with OCT canbe made with flow rates measured in terms of abso-lute blood volumes [65]. Measurements of hemoglo-bin saturation in blood vessels using OCT are indevelopment [66,67], and a recent in vivo study de-monstrated that precise hematocrit measurementusing OCT is feasible [68]. These techniques have notyet been combined to demonstrate convective oxygentransport measurements with OCT. The spatial reso-lution and relatively low hemoglobin absorption inthe wavelength range typically used for OCT mayrestrict such measurements to vessels with largerdiameters and hematocrits.

In summary, we have developed a multimodalityoptical imaging technique to measure microvascularconvective oxygen transport with high spatial andtemporal resolution across a microvessel network re-gion. We demonstrated this technique in vivo in amouse window chamber tumor model. This techni-que may be useful for investigations of complex oxy-gen transport dynamics in microvessel networks.

C. deDeugd and M. Wankhede contributed equallyto this work.

References1. A. C. Guyton, Textbook of Medical Physiology

(W. B. Saunders, 1991).2. B. R. Duling and R. M. Berne, “Longitudinal gradients in peri-

arteriolar oxygen tension,” Circ. Res. 27, 669–678 (1970).3. B. R. Duling, “Microvascular responses to alterations in oxy-

gen tension,” Circ. Res. 31, 481–489 (1972).4. R. D. Shonat and P. C. Johnson, “Oxygen tension gradients

and heterogeneity in venous microcirculation: a phosphores-cence quenching study,” Am. J. Physiol. Heart Circ. Physiol.272, H2233–H2240 (1997).

5. A. G. Tsai, P. C. Johnson, and M. Intaglietta, “Oxygen gradi-ents in themicrocirculation,” Physiol. Rev. 83, 933–963 (2003).

6. R. N. Pittman, “Oxygen transport and exchange in the micro-circulation,” Microcirculation 12, 59–70 (2005).


7. R. L. Greenman, S. Panasyuk, X. Wang, T. E. Lyons, T. Dinh,L. Longoria, J. M. Giurini, J. Freeman, L. Khaodhiar, andA. Veves, “Early changes in the skin microcirculationand muscle metabolism of the diabetic foot,” Lancet 366,1711–1717 (2005).

8. P. E. James, D. Lang, T. Tufnell-Barret, A. B. Milsom, andM. P. Frenneaux, “Vasorelaxation by red blood cells and im-pairment in diabetes: reduced nitric oxide and oxygen deliveryby glycated hemoglobin,” Circ. Res. 94, 976–983 (2004).

9. M. A. Boegehold, “Microvascular structure and function insalt-sensitive hypertension,” Microcirculation 9, 225–241(2002).

10. J. J. Mourad and M. Laville, “Is hypertension a tissue perfu-sion disorder? Implications for renal and myocardial perfu-sion,” J. Hypertens. 24, S10–S16 (2006).

11. P. Cabrales, A. G. Tsai, and M. Intaglietta, “Is resuscitationfrom hemorrhagic shock limited by blood oxygen-carrying ca-pacity or blood viscosity?” Shock Waves 27, 380–389 (2007).

12. L. N. Torres, I. P. Torres Filho, R. W. Barbee, M. H. Tiba,K. R. Ward, and R. N. Pittman, “Systemic responses to pro-longed hemorrhagic hypotension,”Am. J. Physiol. Heart Circ.Physiol. 286, H1811–H1820 (2004).

13. F. Gottrup, “Oxygen in wound healing and infection,” World J.Surg. 28, 312–315 (2004).

14. P. Melis, M. L. Noorlander, A. J. van der Kleij, C. J. vanNoorden, and C. M. van der Horst, “Oxygenation and micro-circulation during skin stretching in undermined and nonun-dermined skin,” Plast. Reconstr. Surg. 112, 1295–1301(2003).

15. M. W. Dewhirst, “Concepts of oxygen transport at the micro-circulatory level,” Semin. Radiat. Oncol. 8, 143–150 (1998).

16. R. K. Jain, “Determinents of tumor blood flow: a review,” Can-cer Res. 48, 2641–2658 (1988).

17. P. Vaupel and L. Harrison, “Tumor hypoxia: causative factors,compensatory mechanisms, and cellular response,”Oncologist9 (Suppl. 5), 4–9 (2004).

18. T. A. Woolsey, C. M. Rovainen, S. B. Cox, M. H. Henegar,G. E. Liang, D. Liu, Y. E. Moskalenko, J. Sui, and L.Wei, “Neu-ronal units linked to microvascular modules in cerebral cor-tex: response elements for imaging the brain,” Cereb.Cortex 6, 647–660 (1996).

19. T. W. Secomb, R. Hsu, E. Y. H. Park, and M. W. Dewhirst,“Green’s function methods for analysis of oxygen delivery totissue by microvascular networks,” Ann. Biomed. Eng. 32,1519–1529 (2004).

20. D. M. Brizel, B. Klitzman, J. M. Cook, J. Edwards, G. Rosner,and M. W. Dewhirst, “A comparison of tumor and normal tis-sue microvascular hematocrits and red cell fluxes in a rat win-dow chamber model,” Int. J. Radiat. Oncol. Biol. Phys. 25,269–276 (1993).

21. B. Klitzman and B. R. Duling, “Microvascular hematocrit andred cell flow in resting and contracting striated muscle,” Am.J. Physiol. 237, H481–H490 (1979).

22. A. K. Dunn, A. Devor, H. Bolay, M. L. Andermann,M. A. Moskowitz, A. M. Dale, and D. A. Boas, “Simultaneousimaging of total cerebral hemoglobin concentration, oxygena-tion, and blood flow during functional activation,” Opt. Lett.28, 28–30 (2003).

23. R. D. Shonat, E. S. Wachman, W.-H. Niu, A. P. Koretsky, andD. L. Farkas, “Near-simultaneous hemoglobin saturation andoxygen tension maps in mouse brain using an AOTF micro-scope,” Biophys. J. 73, 1223–1231 (1997).

24. B. Styp-Rekowska, N. M. Disassa, B. Reglin, L. Ulm,H. Kuppe, T. W. Secomb, and A. R. Pries, “An imaging spectro-scopy approach for measurement of oxygen saturation andhematocrit during intravital microscopy,” Microcirculation14, 207–221 (2007).

25. B. S. Sorg, M. E. Hardee, N. Agarwal, B. J. Moeller, andM. W. Dewhirst, “Spectral imaging facilitates visualizationand measurements of unstable and abnormal microvascularoxygentransport in tumors,”J.Biomed.Opt.13, 014026 (2008).

26. B. S. Sorg, B. J. Moeller, O. Donovan, Y. Cao, andM. W. Dewhirst, “Hyperspectral imaging of hemoglobinsaturation in tumor microvasculature and tumor hypoxiadevelopment,” J Biomed. Opt. 10, 044004 (2005).

27. M. L. Ellsworth, R. N. Pittman, and C. G. Ellis, “Measurementof hemoglobin oxygen saturation in capillaries,” Am. J. Phy-siol. 252, H1031–H1040 (1987).

28. H. Kerger, G. Groth, K. Kalenka, P. Vajkoczy, A. G. Tsai, andM. Intaglietta, “pO2 measurements by phosphorescencequenching: characteristics and applications of an automatedsystem,” Microvasc. Res. 65, 32–38 (2003).

29. L. S. Ziemer, W. M. F. Lee, S. A. Vinogradov, C. Sehgal, andD. F. Wilson, “Oxygen distribution in murine tumors: charac-terization using oxygen-dependent quenching of phosphores-cence,” J. Appl. Physiol. 98, 1503–1510 (2005).

30. D. Fukumura, L. Xu, Y. Chen, T. Gohongi, B. Seed, andR. K. Jain, “Hypoxia and acidosis independently up-regulatevascular endothelial growth factor transcription in braintumors in vivo,” Cancer Res. 61, 6020–6024 (2001).

31. A. S. Golub and R. N. Pittman, “Erythrocyte-associated tran-sients in PO2 revealed in capillaries of rat mesentery,” Am. J.Physiol. Heart Circ. Physiol. 288, H2735–H2743 (2005).

32. D. P. Swain and R. N. Pittman, “Oxygen exchange in the mi-crocirculation of hamster retractor muscle,” Am. J. Physiol.Heart Circ. Physiol. 256, H247–H255 (1989).

33. J. L. Unthank, J. M. Lash, J. C. Nixon, R. A. Sidner, andH. G. Bohlen, “Evaluation of carbocyanine-labeled erythro-cytes for microvascular measurements,” Microvasc. Res. 45,193–210 (1993).

34. D. Feng, D. Marshburn, F. Jen, R. J. Weinberg, R. M. Taylor II,and A. Burette, “Stepping into the third dimension,” J. Neu-rosci. 27, 12757–12760 (2007).

35. M. T. Fallon, “Rats andmice,” inHandbook of Rodent and Rab-bit Medicine, K. Laber-Laird, M. M. Swindle, and P. Flecknell,eds. (Elsevier Science, 1996), pp. 1–38.

36. R. D. Braun, J. L. Lanzen, and M. W. Dewhirst, “Fourier ana-lysis of fluctuations of oxygen tension and blood flow inR3230Ac tumors and muscle in rats,” Am. J. Physiol. 277,H551–H568 (1999).

37. L. I. Cardenas-Navia, D. Yu, R. D. Braun, D. M. Brizel,T. W. Secomb, andM.W. Dewhirst, “Tumor-dependent kineticsof partial pressure of oxygen flucutations during air and oxy-gen breathing,” Cancer Res. 64, 6010–6017 (2004).

38. J. Lanzen, R. D. Braun, B. Klitzman, D. Brizel, T. W. Secomb,and M. W. Dewhirst, “Direct demonstration of instabilities inoxygen concentrations within the extravascular compartmentof an experimental tumor,” Cancer Res. 66, 2219–2223 (2006).

39. K. G. Brurberg, H. K. Skogmo, B. A. Graff, D. R. Olsen, andE. K. Rofstad, “Fluctuations in pO2 in poorly and well-oxygenated spontaneous canine tumors before and duringfractionated radiation therapy,” Radiother. Oncol. 77, 220–226 (2005).

40. K. G. Brurberg, M. Thuen, E.-B. M. Ruud, and E. K. Rofstad,“Fluctuations in pO2 in irradiated human melanoma xeno-grafts,” Radiat. Res. 165, 16–25 (2006).

41. A. Vadapalli, R. N. Pittman, and A. S. Popel, “Estimating oxy-gen transport resistance of the microvascular wall,” Am. J.Physiol. Heart Circ. Physiol. 279, 657–671 (2000).

42. W. G. Hundley, G. J. Renaldo, J. E. Levasseur, and H. A.Kontos, “Vasomotion in cerebral microcirculation of awakerabbits,” Am. J. Physiol. 254, H67–H71 (1988).

43. C. E. Riva, C. J. Pournaras, C. L. Poitry-Yamate, andB. L. Petrig, “Rhythmic changes in velocity, volume, and flow


of blood in the optic nerve head tissue,” Microvasc. Res. 40,36–45 (1990).

44. S. Bertuglia, A. Colantuoni, G. Coppini, and M. Intaglieta,“Hypoxia- or hyperoxia-induced changes in arteriolar vasomo-tion in skeletal muscle microcirculation,” Am. J. Physiol. 260,H362–H372 (1991).

45. H. Kimura, R. D. Braun, E. T. Ong, R. Hsu, T. W. Secomb,D. Papahadjopoulos, K. Hong, and M. W. Dewhirst, “Fluctua-tions in red cell flux in tumor microvessels can lead to tran-sient hypoxia and reoxygenation in tumor parenchyma,”Cancer Res. 56, 5522–5528 (1996).

46. J. L. Lanzen, R. D. Braun, A. L. Ong, and M. W. Dewhirst,“Variability in blood flow and pO2 in tumors in response tocarbogen breathing,” Int. J. Radiat. Oncol. Biol. Phys. 42,855–859 (1998).

47. M. W. Dewhirst, Y. Cao, and B. J. Moeller, “Cycling hypoxiaand free radicals regulate angiogenesis and radiotherapyresponse,” Nat. Rev. Cancer 8, 425–437 (2008).

48. E. M. Lord, L. Harwell, and C. J. Koch, “Detection of hypoxiccells by monoclonal antibody recognizing 2-nitroimidazoleadducts,” Cancer Res. 53, 5721–5726 (1993).

49. M. C. Skala, K. M. Riching, A. Gendron-Fitzpatrick,J. Eickhoff, K. W. Eliceiri, J. G. White, and N. Ramanujam,“In vivo multiphoton microscopy of NADH and FAD redoxstates, fluorescence lifetimes, and cellular morphology in pre-cancerous epithelia,” Proc. Natl. Acad. Sci. USA 104, 19494–19499 (2007).

50. R. K. Jain, “Normalization of tumor microvasculature: anemerging concept in antiangiogenic therapy,” Science 307,58–62 (2005).

51. F. Adhami, G. Liao, Y. M. Morozov, A. Schloemer,V. J. Schmithorst, J. N. Lorenz, R. S. Dunn, C. V. Vorhees,M. Wills-Karp, J. L. Degen, R. J. Davis, N. Mizushima,P. Rakic, B. J. Dardzinski, S. K. Holland, F. R. Sharp, andC.-Y. Kuan, “Cerebral ischemia-hypoxia induces intravascularcoagulation and autophagy,” Am. J. Pathol. 169, 566–583(2006).

52. R. F. Gariano and T. W. Gardner, “Retinal angiogenesis indevelopment and disease,” Nature 438, 960–966 (2005).

53. M.W. Dewhirst, “Angiogenesis and blood flow in solid tumors,”inDrug Resistance in Oncology, B. A. Teicher, ed. (Marcel Dek-ker, 1993), pp. 3–23.

54. A. D. Estrada, A. Ponticorvo, T. N. Ford, and A. K. Dunn, “Mi-crovascular oxygen quantification using two-photon micro-scopy,” Opt. Lett. 33, 1038–1040 (2008).

55. C. B. Schaffer, B. Friedman, N. Nishimura, L. F. Schroeder,P. S. Tsai, F. F. Ebner, P. D. Lyden, and D. Kleinfeld, “Two-photon imaging of cortical surface microvessels reveals a ro-bust redistribution in blood flow after vascular occlusion,”PLoS Biol. 4(2), e22 (2006).

56. E. B. Brown, R. B. Campbell, Y. Tsuzuki, L. Xu, P. Carmeliet,D. Fukumura, and R. K. Jain, “In vivo measurement of gene

expression, angiogenesis and physiological function in tumorsusing multiphoton laser scanning microscopy,” Nat. Med. 7,864–868 (2001).

57. D. Fu, T. Ye, T. E. Matthews, B. J. Chen, G. Yurtserver, andW. S.Warren, “High-resolution in vivo imaging of blood vesselswithout labeling,” Opt. Lett. 32, 2641–2643 (2007).

58. K. Tsukada, E. Sekizuka, C. Oshio, K. Tsujioka, andH. Minamitani, “Red blood cell velocity and oxygen tensionmeasurement in cerebral microvessels by double-wave-length photoexcitation,” J. Appl. Physiol. 96, 1561–1568(2004).

59. P. Vaupel, “Tumor microenvironmental physiology and its im-plications for radiation oncology,” Semin. Radiat. Oncol. 14,198–206 (2004).

60. J. D. Briers, “Laser Doppler, speckle and related techniquesfor blood perfusion mapping and imaging,” Physiol. Meas.22, R35–R66 (2001).

61. V. Rajan, B. Varghese, T. G. van Leeuwen, andW. Steenbergen, “Review of methodological developmentsin laser Doppler flowmetry,” Lasers Med. Sci. (31 Jan.2008).

62. C. Hitzenberger, P. Trost, P.-W. Lo, and Q. Zhou, “Three-dimensional imaging of the human retina by high-speed opti-cal coherence tomography,” Opt. Express 11, 2753–2761(2003).

63. J. A. Izatt, M. D. Kulkarni, S. Yazdanfar, J. K. Barton, andA. J. Welch, “In vivo bidirectional color Doppler flow imagingof picoliter blood volumes using optical coherence tomogra-phy,” Opt. Lett. 22, 1439–1441 (1997).

64. R. Michaely, A. H. Bachmann, M. L. Villiger, C. Blatter,T. Lasser, and R. A. Leitgeb, “Vectorial reconstruction of ret-inal blood flow in three dimensions measured with high reso-lution resonant Doppler Fourier domain optical coherencetomography,” J. Biomed. Opt. 12, 041213 (2007).

65. Y. Wang, B. A. Bower, J. A. Izatt, O. Tan, and D. Huang, “Invivo total retinal blood flow measurement by Fourier domainDoppler optical coherence tomography,” J. Biomed. Opt. 12,041215 (2007).

66. L. Kagemann, G. Wollstein, M. Wojtkowski, H. Ishikawa,K. A. Townsend, M. L. Gabriele, V. J. Srinivasan,J. G. Fujimoto, and J. S. Schuman, “Spectral oximetry as-sessedwith high-speed ultra-high resolution optical coherencetomography,” J. Biomed. Opt. 12, 041212 (2007).

67. C.-W. Lu, C.-K. Lee, M.-T. Tsai, Y.-M. Wang, and C. C. Yang,“Measurement of the hemoglobin oxygen saturation level withspectroscopic spectral-domain optical coherence tomography,”Opt. Lett. 33, 416–418 (2008).

68. N. V. Iftimia, D. X. Hammer, C. E. Bigelow, D. I. Rosen,T. Ustun, A. A. Ferrante, D. Vu, and R. D. Ferguson, “Towardnoninvasive measurement of blood hematocrit using spectraldomain low coherence interferometry and retinal tracking,”Opt. Express 14, 3377–3388 (2006).


multimodal optical imaging of microvessel network convective oxygen transport dynamics

Documents