polarization-sensitive multimodal imaging for detecting ......integrated systems and technologies...

Integrated Systems and Technologies

Polarization-Sensitive Multimodal Imaging for DetectingBreast Cancer

Rakesh Patel1, Ashraf Khan2, Robert Quinlan2, and Anna N. Yaroslavsky1

AbstractIntraoperative delineation of breast cancer is a significant problem in surgical oncology. A reliable method for

demarcation of malignant breast tissue during surgery would reduce the re-excision rate due to positive margins.We present a novel method of identifying breast cancer margins using combined dye-enhanced wide-fieldfluorescence polarization imaging for en face cancer margins and polarization-sensitive (PS) optical coherencetomography (OCT) for cross-sectional evaluation. Tumor specimens were collected following breast surgery,stainedwithmethylene blue, and imaged.Wide-fieldfluorescence polarization imageswere excited at 640 nmandregistered between 660 and 750 nm. Standard and PS OCT images were acquired using a commercial 1,310-nmswept-source system. The imaging results were validated against histopathology. Statistically significant higherfluorescence polarization of cancer as compared with both normal and fibrocystic tumor tissue was measured inall the samples. Fluorescence polarization delineated lateral breast cancer margins with contrast superior to thatprovided by OCT. However, OCT complemented fluorescence polarization imaging by facilitating cross-sectionalinspection of tissue. PS OCT yielded higher contrast between cancer and connective tissue, as compared withstandard OCT. Combined PS OCT and fluorescence polarization imaging shows promise for intraoperativedelineation of breast cancer. Cancer Res; 74(17); 4685–93. �2014 AACR.

IntroductionWomen have a 1 in 8 lifetime risk of having invasive breast

cancer, which is the second leading cause of cancer death inwomen (1). Sixty to seventy percent of patients with breastcancer undergo lumpectomy or breast-conserving surgery(BCS) for surgical resection of the tumor and the rest havemastectomy (2, 3). Alternative noninvasive treatments are alsobeing developed such as MRI-guided focused ultrasound sur-gery (4, 5) but BCS presently remains the preferred method oftreatment. Currently, breast cancer resection margins aredefined several days postoperatively by pathologists analyzingthe gross resection specimen macroscopically and represen-tative histopathology sections of the tissue microscopically,thus rendering a final diagnosis. However, it is not practical toevaluate the entire tumor margin with this method, as thou-sands of histologic tissue sections would have to be examinedto fully evaluate the surface of a breast specimen measuring1 cm in length (6). Therefore, pathologists render a diagnosis byanalyzing the regions of interest selected from macroscopicevaluation, which may lead to sampling error. In addition,pathologic tissue processing takes days to complete and in 20%

to 60% of cases additional tumor is found at the margins (7, 8).Many patients require additional surgery, which results inhigher treatment cost, greater morbidity, infection risk,delayed adjuvant therapy, poor cosmetic outcomes, and apatient's loss of confidence in the adequacy of the treatment(6). Therefore, defining cancer margins intraoperatively is animportant step for lowering the re-excision rate. Alternativepathology approaches such as frozen section (FS) pathology (9)and Touch Prep (TP) cytology (10) have been proposed.However, performance limitations (11) have prevented main-stream adoption of such methods. For example, TP cannotdistinguish between in situ and invasive cancers (12). Similarly,FS processing is time consuming and may damage tissuecompromising definitive permanent histopathology (13, 14).Further, TP and FS are not widely available and not routinelyused even at high volume centers due to the need for an onsitepathologist (8). Therefore, surgeons performing BCS have anurgent need for a method to observe the entire cancer marginintraoperatively in real time.

Optical techniques such as wide-field polarization imaging(WFPI; refs. 15–19) and optical coherence tomography (OCT;refs. 20–22) can potentially perform rapid and accurate eval-uation of the entire surgical margin intraoperatively. Previ-ously, our laboratory has shown promising results in delineat-ing cancer in ex vivo samples of breast ductal carcinomas usinga combination of WFPI and standard OCT (16). In the study,wide-field imaging was used to accurately evaluate lateralextent of the tumor of the tissue, whereas standard OCTprovided high-resolution information on the depth of thelesion. Standard OCT was capable of distinguishing adiposetissue and cancer, but it was difficult to discern tumor on the

1University of Massachusetts, Lowell, 1 University Ave., Lowell, Massa-chusetts. 2University ofMassachusettsMedical School andUMassMemo-rial Medical Center, Worchester, Massachusetts.

Corresponding Author: Anna N. Yaroslavsky, University of Massachu-setts Lowell, 1 University Avenue, Lowell, MA 01854. Phone: 978-934-1350; Fax: 978-934-1305; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-13-2411

�2014 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 4685

on August 14, 2021. © 2014 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst June 23, 2014; DOI: 10.1158/0008-5472.CAN-13-2411

http://cancerres.aacrjournals.org/

background of connective tissue. To remedy this deficiency, inthis study, we investigated the combination of wide-fieldfluorescence polarization and polarization-sensitive OCT (PSOCT). PS OCT is capable of examining birefringence differ-ences within the tissue sample (23). As the birefringenceexhibited by fibrous tissue is higher than that of tumors, PSOCT can be expected to enhance contrast of cancer on thefibrous background (16), thus improving the results of tumordelineation as compared with standard OCT (24–26). In thisarticle, we present en face fluorescence polarization and cross-sectional PSOCT images of breast tumors and compare opticalimages with histopathology. In addition, we evaluated thevalues of fluorescence polarization in invasive ductal carcino-ma (IDC), benign fibrocystic breast disease or benign fibro-cystic change (FCC), and normal breast tissue.

Materials and MethodsContrast agent

Methylene blue (MB) was used as a fluorescent contrastagent for WFPI. MB is currently used for detecting breastcancer in sentinel node (27–30). It is approved for human useand does not affect OCT imaging because the MB absorptionpeaks (618 nmand 668 nm) do not overlapwith thewavelength(1,325 nm) used for OCT. Furthermore, this dye has beenshown to have similar staining pattern to hematoxylin andeosin (H&E) in histopathology (31–33), which simplifies mor-phologic analysis of the images by clinical personnel. Com-mercially available MB (MB 1% injection, USP, AmericanRegent Laboratories, Inc.) was purchased for this study. Forstaining, MB was diluted to a concentration of 0.05 mg/mLwith Dulbecco Phosphate Buffered Saline solution (DPBS 1�,pH 7.4, Mediatech).

Sample preparation and handlingExcess tissue specimens for the study were obtained from

UniversityofMassachusettsMedicalCenter followingBCSunderan Institutional Review Board-approved protocol. In total, wehave investigated 16 specimens from16 patients. Information onthe specimens is summarized in the firstfive columns of Table 1.The specimens for the study were obtained by a pathologistbefore fixation in formalin and after performing gross exami-nation. The tissues included some tumor and a rim of adjoiningresidual breast tissue, which could contain normal tissue. Thespecimenswere placed in vials with saline solution (DPBS pH 7.4from Mediatech Inc.) and immediately transported to UMASSLowell for imaging. Imaging team was not informed of thediagnosis of the specimens. Of the 16 specimens, five wereobtained from mastectomy and 11 from needle localizationlumpectomy specimens either for a preoperative diagnosis ofcarcinoma or in four cases for a diagnosis of fibrocystic changewith atypical ductal hyperplasia. Of the five fibrocystic changesamples, four were from lumpectomies for benign condition.The fifth sample was preoperatively diagnosed as ductal carci-noma in situ. However, after histopathologic examination of theimaged tissue, it was confirmed to be benign fibrocystic change.

Upon receiving samples at the Advanced BiophotonicsLaboratory, the samples were stained in a 0.05 mg/mL MBsolution for 2 minutes. The excess stain rinsed off and thetissues were imagedwith theWFPI system and PSOCT system.Following imaging, the specimens were fixed in formalin andhistopathology sections were processed from approximatelythe same en face and vertical planes that were imaged.

OCT imaging and data processingA commercially available Thorlabs swept-source system

(OCS1300SS) with a PS add-on module (PSOCT-1300) was

Table 1. Study subjects

Mean fluorescencepolarization � SD

Sample # Gender AgeLateral dimensions,(mm � mm)

Diagnosis,grade

Surgicaltreatment Benign Malignant

1 F 72 6.7 � 8.1 IDC, 1 Lumpectomy 0.01 � 0.01 0.04 � 0.012 F 44 7.4 � 9.9 IDC, 1 Lumpectomy 0.01 � 0.01 0.04 � 0.013 F 71 7.2 � 8.7 IDC, 1 Lumpectomy 0.02 � 0.01 0.05 � 0.024 F 49 7.9 � 10.5 IDC,1 Mastectomy 0.02 � 0.01 0.04 � 0.015 F 48 8.8 � 9.8 IDC, 2 Lumpectomy 0.03 � 0.02 0.06 � 0.026 F 73 8.6 � 9.3 IDC, 2 Lumpectomy 0.03 � 0.02 0.08 � 0.027 F 49 8.9 � 9.4 IDC, 2 Mastectomy 0.03 � 0.02 0.07 � 0.028 F 55 9.2 � 9.4 IDC, 2 Lumpectomy 0.03 � 0.02 0.08 � 0.029 F 44 7.9 � 9.7 IDC, 3a Mastectomy 0.03 � 0.01 0.09 � 0.03

10 F 72 7.5 � 7.7 IDC, 2b Mastectomy 0.03 � 0.02 0.07 � 0.0211 F 72 7.3 � 8.3 IDC, 2 Mastectomy 0.03 � 0.02 0.07 � 0.0212 F 72 6.6 � 9.9 FCC Lumpectomy 0.02 � 0.0113 F 67 11.8 � 12.9 FCC Lumpectomy 0.01 � 0.0114 F 61 6.5 � 8.5 FCCc Lumpectomy 0.01 � 0.0115 F 38 9.8 � 12.4 FCC Lumpectomy 0.03 � 0.0116 F 53 6.1 � 10.2 FCC Lumpectomy 0.01 � 0.01

aWith squamous differentiation; bwith lobular growth; cpreoperatively diagnosed as ductal carcinoma in situ.

Patel et al.

Cancer Res; 74(17) September 1, 2014 Cancer Research4686




used for OCT imaging in this study. The system incorporated a1,325-nm high-speed, frequency swept laser to illuminate thesample. The two orthogonal polarization states were detectedusing a fiber polarization beam splitter. The interferencesignals were fed to two balanced detectors measuring thevertical and horizontal polarization states, which were usedto calculate birefringence-induced phase retardation imagesalong with standard OCT images. The OCT system provided alateral resolution of 25 mm and an axial resolution of 9 mm to12 mmwith amaximum imaging depth of 1.5 to 2mm in breasttissue and a scanning rate of 25 fps. The data acquisition andprocessing was handled using Thorlabs-integrated softwarepackage. The software allowed for automatic acquisition ofmultiple cross-sectional images of a given volume, which couldbe processed into a three-dimensional (3D) volume block.

Wide-field imaging and data processingThe illuminationwas provided by a xenon arc lamp (Lambda

LS; Sutter) combined with a 640-nm bandpass filter, with a fullwidth at half maximum of 10 nm. The lamp was attached to acustom-built ring light guide, which provided homogeneousillumination of the sample. Images were acquired using a 0.5�Rodenstock lens coupled to a CCD camera (CoolSnap cf2;Roper Scientific). Linear polarizing filters (Meadowlark Optics)were integrated into the ring light guide in the pathways of thelight incident on the sample and light collected by the camera.The analyzing polarizer was calibrated for two positions,parallel or perpendicular to the polarization of the incidentlight, which allowed for rapid sequential acquisition of co- andcross-polarized images. Data acquisition and processing wasperformed using Metamorph software (Molecular Devices,Inc.). Fluorescence polarization images were registeredbetween 660 nm and 750 nm using a bandpass filter (660AELP;Omega Optical) placed in front of the CCD camera. Thecalibration factor (G ¼ 0.98) was measured to account for abias in the detection of the two polarization states in themanner described by Lakowicz (34). The system provided alateral resolution of 12mmandafield of viewof 2.2 cm� 1.6 cm.Fluorescence polarization images were calculated pixel-by-pixel using the formula

fpli ¼ Ifco � G � IfcrIfco þ G � Ifcr ; ð1Þ

where fpli is the fluorescence polarization image, Ifco is thefluorescence co-polarized image, and Ifcross is the fluorescencecross-polarized image.

HistopathologyBoth en face and vertical histopathology sections were

processed from each of the imaged tissue samples. Afterimaging, the fixed tissues were embedded horizontally inparaffin, and several 5-mm thick slices were cut using amicrotome. The sections were cut from the approximateplanes that were imaged. Once cut, the tissue sections weretransferred to glass slides and stained with H&E. After the enface histopathology had been processed, the paraffin tissueblocks were melted, the tissue was reoriented and embeddedfor cross-sectional slices to be cut. These vertical sections werecut from the approximate locations of regions of interest

defined from OCT and PS OCT images. The cross-sectionalsections were then transferred to glass slides and stained asdescribed for en face sections. All histopathology slides weredigitized using a Zeiss Axioscope microscope (Zeiss; 5� objec-tive lens, numerical aperture, 0.13). The optical images werecorrelated with corresponding digitized histopathology.

Evaluation of tumor contrast in optical imagesContrast of tumor with respect to normal tissue was

assessed quantitatively for wide-field fluorescence polariza-tion, standardOCT, andPSOCT images. For correlating opticaland histologic images and for evaluating cancer contrast in theoptical images, the tumor and normal areas were outlined inthe optical images of each specimen, exactly as they present inhistopathology. This procedure was described in detail else-where (15, 16, 35, 36). In short, cancerous and normal regionswere outlined by a pathologist in digitized histopathologyslides. Because of the preparation of paraffin-embedded his-topathology, sections may not preserve the shape and size ofthe imaged tissue block. To correct for this artifact, digitizedhistopathology slides were overlaid onto the optical images.Then affine, projective, or polynomial transformations wereapplied so that similar structures in the optical images coin-cided with corresponding structures in histopathology. Aftercorrection, the regions corresponding to cancer and normalbreast tissue in histopathology were outlined in the opticalimages.

Mean fluorescence polarization values were determined foreach specimen frommean pixel values of the entire tumor andnormal areas in the fluorescence polarization images. Thus,obtainedmean fluorescence polarization values for cancer andnormal tissue were averaged across all samples.

To compare contrast of the standard OCT and PS OCTimages, the images were normalized by the maximum pixelvalue. Normalized mean pixel values for tumor and normalareas were calculated for each specimen. Then obtained meanpixel values for cancer and normal tissue were averaged acrossall specimens.

Tumor contrast was calculated using the formula

C ¼ pxlT � pxlN jjpxlT þ pxlN

; ð2Þwhere C is the contrast, pxlT is the mean pixel value for tumorareas, and pxlN is the mean pixel value for normal areas,including adipose and fibrous connective tissue. Contrastvalue of 1 indicates maximum contrast, whereas 0 indicatesno contrast.

Statistical analysisThe differences in tumor and normal values were evaluated

statistically for wide-field fluorescence polarization, standardOCT, and PS OCT images using a one-tailed Student t test fortwo independent populations. In all cases, we tested thealternative hypothesis that the mean pixel value averaged overtumor regions in each sample was significantly different fromthe mean pixel value averaged over all normal regions in thesamples. The P value results were calculated for each modalityand considered significant at P < 0.005.

Multimodal Polarization Imaging for Breast Cancer Detection

www.aacrjournals.org Cancer Res; 74(17) September 1, 2014 4687




Results/DiscussionIn total, 11 IDCs and five benign fibrocystic breast tumors

from 16 patients were imaged for this study. The patient agesranged from 38 to 73 years. The sample sizes varied approx-imately between 6.1 to 12.9 mm in the lateral dimensions and 3to 7mm in depth. The respective data are summarized in Table1. Histologic analysis performed by the study pathologist afteroptical imaging established that all cancer specimens includedtumor and normal regions. Fibrocystic tumor specimens didnot contain normal tissue.

Example en face images of a representative IDC sample arepresented in Fig. 1. IDC shows architecturally abnormal pro-liferation of ducts, with ill-defined margins and variableamount of fibrocollagenous or adipose stroma. In this carci-noma sample, the tumor presents on the background offibrocollagenous connective tissue as seen in histopathology(Fig. 1A) and residual benign breast tissue seen in the bottomthird of the tissue section is predominantly adipose (fat) tissue.Cancer-affected area is bright in the fluorescence polarizationimage (Fig. 1B). Comparison of the tumor boundary in fluo-rescence polarization image and in histopathology (Fig. 1A)demonstrates good correlation. In Fig. 1B, the tumor is clearlydistinguished from the surrounding normal tissue, as fluores-cence polarization rapidly decreases upon transition into thenormal tissue. The en face standard OCT and PS OCT imagesare presented in Fig. 1C and D, respectively. Tumor-affectedarea, as well as connective tissue, seems equally bright in thestandard OCT image (Fig. 1C) and can hardly be distinguished.In contrast, in the PS OCT image (Fig. 1D), tumor is dark,

whereas the fibrous tissue is bright, which confirms ourhypothesis that birefringence imaging may help in delineatingtumor and connective tissue. Comparison of the en face opticalimages, including fluorescence polarization (Fig. 1B), standardOCT (Fig. 1C), and PS OCT (Fig. 1D), with histopathology(Fig. 1A) demonstrates that fluorescence polarization deline-ates the tumor margins more accurately as compared withstandard OCT and PS OCT. In particular, neither of the OCTmodalities highlights cancer in the upper left corner of thespecimen. For this particular sample, the contrast for tumor toadipose tissue was measured to be 0.65 in the fluorescencepolarization image (Fig. 1B), 0.14 and 0.03 in the standard OCT(Fig. 1C) and PS OCT (Fig. 1D) en face images, respectively.Moreover, it is hard to discern fibrous tissue surrounded bythe tumor in Fig. 1C. However, the standard OCT imagein Fig. 1C demonstrates remarkable morphologic detail. Inparticular, honeycomb structure of adipose cells is very prom-inent. PS OCT (Fig. 1D) provides complimentary informationby highlighting the morphology of connective tissue. Thecontrast for tumor to fibrous connective tissue was 0.34 inthe fluorescence polarization image (Fig. 1A), 0.22 and 0.40 inthe standardOCT (Fig. 1C) andPSOCT (Fig. 1D) en face images,respectively.

Cross-sectional OCT images of the IDC specimen shownin Fig. 1 are presented in Fig. 2B–C. The approximate locationof the cross-sectional images is indicated in the en facehistopathology image displayed in Fig. 1A. Cross-sectionalhistopathology (Fig. 2A) shows an area of tumor and connec-tive tissue surrounded by adipose tissue on either side. Sim-ilarly to the en face standardOCT image in Fig. 1, standardOCTshows good contrast of adipose tissue in the cross-sectionpictured (Fig. 2B). The standard OCT image correlates wellwith histopathology but does not differentiate tumor on thebackground of connective tissue. PS OCT (Fig. 2C) emphasizesan area of connective tissue that is not clearly delineated in thestandard OCT image (Fig. 2B), although it does not clearlydistinguish adipose tissue.

A representative benign fibrocystic breast disease case isshown in Figs. 3 and 4. Fibrocystic change (FCC) presents as acombination of features, including stromal fibrosis, cysticdilatation of ducts, apocrine metaplasia, and adenosis. En facehistologic and optical images are displayed in Fig. 3 withcorresponding cross-sectional images shown in Fig. 4. Inhistopathology (Fig. 3A), the pink area seen above the fattybreast tissue shows FCC, which mainly includes stromalfibrosis and cystic dilatation of the ducts. Smaller foci includelobules that have some minimal degree of adenosis. Fluores-cence polarization does not highlight any areas in the sample.Thus, the benign lesion does not exhibit significant fluores-cence polarization.

As compared with the ductal carcinoma sample presentedin Figs. 1 and 2, the contrast in fluorescence polarization imageof FCC is poor (Fig. 3A). FCC exhibited fluorescence polariza-tion of 0.02. This value is significantly lower as compared withthe average fluorescence polarization of the tumor area in theductal carcinoma specimen presented above, which was mea-sured as 0.09. Standard OCT (Fig. 3B) and PS OCT (Fig. 3C) alsodisplay low contrast except for small pockets of adipose tissue

HistologyFluorescencepolarization

Standard OCT PS OCT180º

0º

Ph

ase

reta

rdat

ion

A B

C D

Figure 1. En face images of a representative sample with invasive ductalcarcinoma. A, histopathology. B, wide-field fluorescence polarization. C,standard OCT. D, PS OCT. Red line in A, approximate location of cross-sectional images shown in Fig. 2 (bar, 1 mm).

Patel et al.





in between areas of dense fibrous tissue. Neither of the OCTimages of FCC highlights the lesion.Cross-sectional standard OCT (Fig. 4B) shows no distinction

between fibrous tissue and benign ducts and lobules, asindicated in histopathology. There is no adipose tissue presentin this cross-section and as such the standard OCT image lackscontrast. In contrast, PS OCT (Fig. 4C) yields higher signal inareas corresponding to fibrous connective tissue, whereas thebenign ducts and lobules exhibit a lower signal. PS OCThighlights dense connective tissue within FCC sample. It alsoemphasizes the stromal fibrosis within the tissue except for theregions containing breast ducts and lobules with adenosis.The pixel values of optical images for cancer, fibrocystic

tumor, and normal areas averaged over all samples are sum-marized in Fig. 5. For fluorescence polarization images, theseaveraged pixel values represent mean fluorescence polariza-tion. The mean fluorescence polarization of normal, fibrocys-tic, and cancerous areas measured over all samples in thisstudy was found to be 0.03� 0.01, 0.02� 0.01, and 0.06� 0.02,respectively. The fluorescence polarization values found fornormal and cancer tissue in this study are in good accordancewith previously measured values from our earlier study (15).Higher fluorescence polarization of cancer with respect tobenign tissue was measured in all the samples. There could beseveral explanations of these results. In our previous work, wehave ruled out the hypothesis that differences in elastic scat-tering from tumor and normal tissue lead to enhanced fluo-rescence polarization of breast cancer (13). Restricted rota-tional motion as well as reduced fluorescence lifetime of thedye molecules in cancer cells can result in higher fluorescencepolarization (34). Restricted rotational motion of dye mole-cules could be caused by several factors such as binding of thedye molecules to large cell organelles and higher intracellularviscosity. The work is currently under way to investigate the

origins of higher fluorescence polarization in breast cancer ascompared with normal tissue and benign pathology.

Fluorescence polarization values averaged over the cancerand normal areas for each sample are presented in columnsseven and eight of Table 1. Mean fluorescence polarizationvalues for normal tissues varied between 0.01 and 0.03, whereasmean fluorescence polarization of cancer was in the rangebetween 0.04 and 0.09. Even though the highest mean fluores-cence polarization, 0.03, of normal tissue seems to be close tothe lowestmean fluorescence polarization exhibited by cancer,0.04, it can be seen from the table that the samples with lowerfluorescence polarization of normal tissue exhibit lower fluo-rescence polarization values of cancer and similarly sampleswith higher normal tissue fluorescence polarization exhibithigher cancer fluorescence polarization. These variationsbetween the samples are most likely caused by the differencesin the background optical properties of the specimens, and inparticular, by differences in the scattering coefficient. It canalso be appreciated that higher grade cancer tissue demon-strated overall higher fluorescence polarization. Further stud-ies are required to investigate the apparent correlationbetween the stage of cancer and fluorescence polarizationvalues exhibited by tissue.

Averaged over all samples, fluorescence polarization exhib-ited by breast cancer and FCC is 0.06 � 0.02 and 0.02 � 0.01,respectively. The gradient of fluorescence polarization at theboundary between cancer and benign tissue is high in all thespecimens investigated. Thisfinding indicates thatfluorescencepolarization may be capable of differentiating between malig-nant and benign disease and accurately delineating cancer. It isknown that dense breast tissue as a result of fibrocystic changecan cause enhanced contrast onMRmammography, leading tofalse positives (37). Furthermore, with commonly used currentbreast cancer screening tests, conventional mammography,

Adipose

Tumor

Connective

Bar = 1 mm

0º 180ºPhase retardation

A

B

C

Histology

Standard OCT

Polarization OCT

Figure 2. Cross-sectional images of the invasive ductal carcinoma sample presented in Fig. 1. The images were acquired from the approximate locationindicated in Fig. 1A (red line). A, the histopathology cross section of IDC sample outlining areas of tumor, adipose, and connective tissue. Correspondingstandard OCT and PS OCT sections are shown in B and C, respectively (bar, 1 mm).






and/or clinical breast examination, false-positive rates over a10-year period have been reported as high as 31.7% (38).Therefore, if benign diseases could be reliably diagnosed pre-operatively using "optical biopsy," it could prevent unnecessaryresection of nonmalignant tissue or surgery all together forthese patients.

Normalized average pixel values of normal and cancerousareas, measured from OCT and PS OCT images, are summa-rized in Fig. 6. Averaged pixel values for normal and cancertissue for standard OCT were 0.13 � 0.09 and 0.29 � 0.07,respectively. For PSOCT, the normalized pixel value for normalareaswas 0.26� 0.08 and 0.29� 0.08 for cancer. The contrast ofcancer in the fluorescence polarization images was found to be0.39 � 0.08, whereas in standard OCT and PS OCT images, itwas 0.39 � 0.15 and 0.24 � 0.09. Fluorescence polarizationshowed a significant difference between tumor and normal(P < 0.00001). Standard OCT differences between tumor andnormal were also significant (P < 0.0001). Interestingly, eventhough PS OCT reliably delineated cancer on the backgroundof birefringent connective tissue, it revealed no significantdifference in contrast of cancer as compared with normaltissue (fibrous and adipose) averaged across all samples inthis study. This could be explained by low birefringenceexhibited by cancer and normal adipose tissue. It should also

be noted that contrast of the tumor in the image is not the onlyimportant factor in evaluating the performance of themethod.The ability of the technique to present morphology of tissuewith adequate resolution should also be considered as equallyimportant. In this respect, all the techniques investigated,including fluorescence polarization, standard OCT, and PSOCT, are complimentary and demonstrated good performancein discerning the morphology of tumor, adipose, and connec-tive tissues, respectively.

A PS combination of wide-field and OCT imaging wouldallow for more rapid and accurate evaluation of tumor mar-gins, as compared with the OCT imaging alone. The OCTsystem is capable of simultaneously acquiring standard andPS OCT cross-sectional images in approximately 40 millise-conds. The en face imaging of 3 cm2 sample surface will takemore than 30 seconds. A combination system could use wide-field fluorescence polarization to acquire en face images,rapidly acquired within 1 second, and then use this image toguide cross-sectional OCT inspection of suspicious areas.

In summary, we have shown that fluorescence polarizationwide-field imaging is suitable for delineating lateral extent ofthe breast cancer margins from en face images with contrastsuperior to that provided by OCT. However, OCT imaging cancomplement wide-field fluorescence polarization by enabling

HistologyFluorescencepolarization

Standard OCT PS OCT

Adipose

Stromalfibrosis

Benignbreast duct

Lobules w/adenosis

Bar = 1 mm

180º

0º

Pha

se r

etar

datio

n

A B

C D

Figure 3. En face images of arepresentative samplewith FCC. A,histopathology. B, wide-fieldfluorescence polarization. C,standardOCT. D, PSOCT. Red lineinA, approximate locationof cross-sectional images shown in Fig. 4(bar, 1 mm).

Patel et al.





high-resolution cross-sectional imaging and thus facilitatinginspection of the close margins. In our previous publication(16), we have reported that delineating tumor on the back-ground of the fibrous tissue using standard OCT may bechallenging and we proposed to use PS OCT to emphasizebirefringence properties of the connective tissue and thusimprove the contrast of the tumor margins. In this study, wehave confirmed that PS OCT is capable of differentiatingcancer from fibrous tissue yielding the higher contrast of0.40, as compared with 0.22 in the standard OCT. Fluorescencepolarization has also demonstrated the capability to differen-tiate benign tissue from cancer. Thus, this technologymay alsoprove to be useful as an optical biopsy tool to help avoidunnecessary surgical removal of benign tissue, such as FCC.The results of our pilot trial suggest that presented combinedtechnology has potential to detect microscopic nests of ductal

carcinoma in situ. The resolution, afforded by the device(� 12 mm), and high contrast of fluorescence polarizationimages point toward the feasibility of successful outcome.However, larger trials are required for conclusive proof of thishypothesis. Further clinical studies will focus on determiningthe sensitivity and specificity of this combined technology fordelineating different types of breast cancer.

In conclusion, this study demonstrates the potential ofcombining wide-field fluorescence polarization imaging andOCT, standard and PS, for rapid and reliable evaluation ofbreast cancer margins. The proposed approach can be used toimage the excised tissue or the surgical wound in situ. In

0.5

0.4

0.3

0.2

0.1

0PS OCT Standard OCT

Normal Malignant

Nor

mal

ized

pix

el v

alue

Figure 6. Normalized standard OCT and PS OCT pixel values for normaland malignant tissue averaged over all samples. Bars, SD above themean.

Bar = 1 mm

Stromalfibrosis

Benignbreast duct

Lobules w/adenosis

0º 180ºPhase retardation

Histology

A

B

C

Standard OCT

Polarization OCT

Figure 4. Cross-sectional images of the FCC sample presented in Fig. 3. The images were acquired from the approximate location indicated in Fig. 3A (redline). A, the histopathology cross sectionof FCCsample outlining areasof stromal fibrosis andbenignducts and lobules. CorrespondingstandardOCTandPSOCT sections are shown in B and C, respectively (bar, 1 mm).

0.08

0.07

0.06

0.05

0.04

0.03

0.02

0.01

0.00FCC Normal Malignant

Flu

ores

cenc

e po

lariz

atio

n

Figure 5. Averaged fluorescence polarization values for FCC, normal, andmalignant tissue. Bars, SD above the mean.






general, the twoapproaches should yield equivalent results andboth can be implemented in clinical practice. For in vivoimplementation, it will be required to rinse the surgical bedwith aqueous solution ofMB,which should be safe, as currentlyMB is used clinically in much higher concentration for map-ping sentinel nodes. Both methods, i.e., fluorescence polariza-tion andmultimodal OCT, aremature technologies that can bereadily implemented in a single instrument. Further develop-ment of this combined technology may enable intraoperativeevaluation of tissue and serve to lower positive margin andreoccurrence rates.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: A. Khan, A.N. Yaroslavsky

Development of methodology: A. Khan, A.N. YaroslavskyAcquisition of data (provided animals, acquired and managedpatients, provided facilities, etc.): R. Patel, A. Khan, R. Quinlan,A.N. YaroslavskyAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): R. Patel, A. Khan, A.N. YaroslavskyWriting, review, and/or revision of the manuscript: R. Patel, A. Khan, A.N.YaroslavskyStudy supervision: A. Khan, R. Quinlan, A.N. Yaroslavsky

AcknowledgmentsThe authors thankAnjulDavis and Javier Jurado fromThorlabs for lending the

PS OCT system and technical assistance.

The costs of publication of this article were defrayed in part by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received August 21, 2013; revised June 4, 2014; accepted June 5, 2014;published OnlineFirst June 23, 2014.

References1. American Cancer Society [Internet]. Atlanta: American Cancer Society,

Inc. c2011 [updated 2014; cited 2014]. Available from: http://www.cancer.org/research/cancerfactsstatistics/breast-cancer-facts-figures

2. McGuire KP, Santillan AA, Kaur P, Meade T, Parbhoo J, Mathias M,et al. Are mastectomies on the rise? A 13-year trend analysis of theselection of mastectomy versus breast conservation therapy in 5865patients. Ann Surg Oncol 2009;16:2682–90.

3. BalchCM, Jacobs LK.Mastectomies on the rise for breast cancer: "thetide is changing". Ann Surg Oncol 2009;16:2669–72.

4. Gombos EC, Kacher DF, Furusawa H, Namba K. Breast focusedultrasound surgery with magnetic resonance guidance. Top MagnReson Imaging 2006;17:181–8.

5. Hirose M, Kacher DF, Smith DN, Kaelin CM, Jolesz FA. Feasibility ofMR imaging–guided breast lumpectomy for malignant tumors in a0.5-T open-configuration MR imaging system. Acad Radiol 2002;9:933–41.

6. Gould EW, Robinson PG. The pathologist's examination of the "lump-ectomy"–the pathologists' viewof surgicalmargins. Semin SurgOncol1992;8:129–35.

7. Dick AW, Sorbero MS, Ahrendt GM, Hayman JA, Gold HT, SchiffhauerL, et al. Comparative effectiveness of ductal carcinoma in situ man-agement and the roles of margins and surgeons. J Nat Cancer Inst2011;103:92–104.

8. Wilke LG, Brown JQ, Bydlon TM, Kennedy SA, Richards LM, JunkerMK, et al. Rapid noninvasive optical imaging of tissue composition inbreast tumor margins. Am J Surg 2009;198:566–74.

9. Gal AA, Cagle PT. The 100-year anniversary of the description of thefrozen section procedure. JAMA 2005;294:3135–7.

10. Valdes EK, Boolbol SK, Cohen J-M, Feldman SM. Intra-operativetouch preparation cytology; does it have a role in re-excisionlumpectomy? Ann Surg Oncol 2007;14:1045–50.

11. Weinberg E, Cox C, Dupont E, White L, Ebert M, Greenberg H, et al.Local recurrence in lumpectomy patients after imprint cytologymarginevaluation. Am J Surg 2004;188:349–54.

12. Harigopal M, Chhieng DC. Breast cytology: current issues and futuredirections. Open Breast Cancer J 2010;81–9.

13. Cend�an JC, Coco D, Copeland EM. Accuracy of intraoperative frozen-section analysis of breast cancer lumpectomy-bedmargins. J AmCollSurg 2005;201:194–8.

14. D'Halluin F, Tas P, Rouquette S, Bendavid C, Foucher F, Meshba H,et al. Intra-operative touch preparation cytology following lumpectomyfor breast cancer: a series of 400 procedures. Breast (Edinburgh,Scotland) 2009;18:248–53.

15. Patel R, Khan A, Wirth D, Kamionek M, Quinlan R, Yaroslavsky AN.Multimodal optical imaging for detecting breast cancer. J Biomed Opt2012;17:066008.

16. Patel R, Khan A, Kamionek M, Kandil D, Quinlan R, Yaroslavsky AN.Delineating breast ductal carcinoma using combined dye-enhancedwide-field polarization imaging and optical coherence tomography.J Biophotonics 2013;6:679–86.

17. Jacques SL, Ramella-Roman JC, Lee K. Imaging skin pathology withpolarized light. J Biomed Opt. 2002;7:329–40.

18. Demos SG, Alfano RR. Optical polarization imaging. Appl Opt OSA1997;36:150.

19. Jacques SL, Roman JR, Lee K. Imaging superficial tissues withpolarized light. Lasers Surg Med 2000;26:119–29.

20. McLaughlin RA, Quirk BC, Curatolo A, Kirk RW, Scolaro L, Lorenser D,et al. Imaging of breast cancer with optical coherence tomographyneedle probes: feasibility and initial results. IEEE J Selected Topics inQuantum Electronics 2012;18:1184–91.

21. Boppart SA, Bouma BE, Pitris C, Tearney GJ, Southern JF, BrezinskiME, et al. Intraoperative assessment of microsurgery with three-dimensional optical coherence tomography. Radiology 1998;208:81–6.

22. Nguyen FT, Zysk AM, Chaney EJ, Kotynek JG, Oliphant UJ, BellafioreFJ, et al. Intraoperative evaluation of breast tumormargins with opticalcoherence tomography. Cancer Res 2009;69:8790–6.

23. De Boer JF, Milner TE, van Gemert MJC, Nelson JS. Two-dimensionalbirefringence imaging in biological tissue by polarization-sensitiveoptical coherence tomography. Opt Lett OSA 1997;22:934.

24. Yang Y, Wu L, Feng Y, Wang RK. Observations of birefringence intissues from optic-fibre-based optical coherence tomography. MeasSci Technol 2003;14:41–6.

25. Pircher M, Goetzinger E, Leitgeb R, Hitzenberger CK. Three dimen-sional polarization sensitive OCT of human skin in vivo. Opt Express2004;12:3236.

26. G€otzinger E, Pircher M, Sticker M, Fercher AF, Hitzenberger CK.Measurement and imaging of birefringent properties of the humancornea with phase-resolved, polarization-sensitive optical coherencetomography. J Biomed Opt 2004;9:94–102.

27. Nour A. Efficacy of methylene blue dye in localization of sentinel lymphnode in breast cancer patients. Breast J 2004;10:388–91.

28. Varghese P, Abdel-Rahman AT, Akberali S, Mostafa A, GattusoJM, Carpenter R. Methylene blue dye–a safe and effectivealternative for sentinel lymph node localization. Breast J 2008;14:61–7.

29. Simmons R, Thevarajah S, Brennan MB, Christos P, Osborne M.Methylene blue dye as an alternative to isosulfan blue dye for sentinellymph node localization. Ann Surg Oncol 2003;10:242–7.

30. Blessing WD, Stolier AJ, Teng SC, Bolton JS, Fuhrman GM. A com-parison of methylene blue and lymphazurin in breast cancer sentinelnode mapping. Am J Surg 2002;184:341–5.

Patel et al.





31. Park J, Mroz P, Hamblin MR, Yaroslavsky AN. Dye-enhanced multi-modal confocal microscopy for noninvasive detection of skin cancersin mouse models. J Biomed Opt 2010;15:026023.

32. Al-Arashi MY, Salomatina E, Yaroslavsky AN. Multimodal confocalmicroscopy for diagnosing nonmelanoma skin cancers. Lasers SurgMed 2007;39:696–705.

33. Yaroslavsky AN, Barbosa J, Neel V, DiMarzio C, Anderson RR. Com-bining multispectral polarized light imaging and confocal microscopyfor localization of nonmelanoma skin cancer. J Biomed Opt 2005;10:14011.

34. Lakowicz JR. Principles of fluorescence spectroscopy. New York:Kluwer Academic/Plenum Publishers; 1999. p. 298–9.

35. Yaroslavsky AN, Neel V, Anderson RR. Fluorescence polarization imag-ing fordelineatingnonmelanomaskincancers.OptLett2004;29:2010–2.

36. Yaroslavsky AN, Salomatina EV, Neel V, Anderson R, Flotte T. Fluo-rescence polarization of tetracycline derivatives as a technique formapping nonmelanoma skin cancers. J Biomed Opt 2007;12:014005.

37. VanGoethemM,Schelfout K,DijckmansL, VanDer Auwera JC,WeylerJ, Verslegers I, et al. MRmammography in the pre-operative staging ofbreast cancer in patients with dense breast tissue: comparison withmammography and ultrasound. Eur Radiol 2004;14:809–16.

38. Elmore JG,BartonMB,Moceri VM,PolkS,ArenaPJ, FletcherSW.Ten-year risk of false positive screening mammograms and clinical breastexaminations. N Engl J Med 1998;338:1089–96.






2014;74:4685-4693. Published OnlineFirst June 23, 2014.Cancer Res Rakesh Patel, Ashraf Khan, Robert Quinlan, et al. CancerPolarization-Sensitive Multimodal Imaging for Detecting Breast

Updated version

10.1158/0008-5472.CAN-13-2411doi:

Access the most recent version of this article at:

Cited articles

http://cancerres.aacrjournals.org/content/74/17/4685.full#ref-list-1

This article cites 34 articles, 1 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/74/17/4685.full#related-urls

This article has been cited by 2 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/74/17/4685To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-13-2411

http://cancerres.aacrjournals.org/content/74/17/4685.full#ref-list-1

http://cancerres.aacrjournals.org/content/74/17/4685.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/74/17/4685


polarization-sensitive multimodal imaging for detecting ......integrated systems and technologies...

Documents