radiology paper

Quing Zhu, PhDEdward B. Cronin, MDAllen A. Currier, MDHugh S. Vine, MDMinming Huang, PhDNanGuang Chen, PhDChen Xu, MS

Published online10.1148/radiol.2371041236

Radiology 2005; 237:57–66

Abbreviation:FWHM � full width at half maximum

1 From the Bioengineering Program,Electrical and Computer EngineeringDepartment, University of Connecti-cut, 371 Fairfield Rd, U2157, Storrs,CT 06269-1157 (Q.Z., M.H., N.C.,C.X.); and Department of Radiology,Hartford Hospital, Hartford, Conn(E.B.C., A.A.C., H.S.V.). Received July15, 2004; revision requested Septem-ber 22; revision received November 4;accepted December 14. Supported bythe Department of Defense, ArmyBreast Cancer Program (DAMD17-00-1-0217); the Donaghue Foundation;and the National Institutes of Health(R01EB002136). Address correspon-dence to Q.Z. (e-mail: [email protected]).

Authors stated no financial relation-ship to disclose.

Author contributions:Guarantor of integrity of entire study,Q.Z.; study concepts/study design ordata acquisition or data analysis/inter-pretation, all authors; manuscriptdrafting or manuscript revision for im-portant intellectual content, Q.Z.,E.B.C.; manuscript final version ap-proval, all authors; literature research,Q.Z.; clinical and experimental stud-ies, all authors; statistical analysis,Q.Z.; manuscript editing, Q.Z., E.B.C.© RSNA, 2005

Benign versus MalignantBreast Masses: OpticalDifferentiation withUS-guided Optical ImagingReconstruction1

PURPOSE: To investigate prospectively the feasibility of using optical tomographywith ultrasonographic (US) localization to differentiate malignant from benignbreast masses and to compare optical tomography with color Doppler US.

MATERIALS AND METHODS: The study was approved by the local internal reviewboard committee and by the Human Subjects Research Review Board of ArmyMedical Research and Materiel Command. Signed informed consent was obtained,and the study was HIPAA compliant. Between May 2003 and March 2004, 65consecutive women (mean age, 51 years; age range, 24–80 years) with 81 breastlesions underwent US-guided biopsy and were scanned with a combined imager.The hand-held probe, which consisted of a centrally located US transducer sur-rounded by near-infrared sensors, was used to simultaneously acquire coregisteredUS images and optical data. The lesion location obtained at US was used to guideoptical imaging reconstruction. Light absorption was measured at two wavelengths.From these measurements, tumor angiogenesis was assessed on the basis of calcu-lated total hemoglobin concentration. A Student t distribution was used to calculatethe statistical significance of mean maximum and mean average hemoglobin con-centrations obtained in malignant and benign lesion groups, and P � .001 wasconsidered to indicate a statistically significant difference.

RESULTS: Biopsy results revealed eight early stage invasive carcinomas (malignantgroup) and 73 benign lesions (benign group). The mean maximum and meanaverage hemoglobin concentrations in the malignant group were 122 �mol/L �26.8 (� standard deviation) and 88 �mol/L � 24.5, respectively. The mean maxi-mum and mean average hemoglobin concentrations in the benign group were 55�mol/L � 24.8 and 38 �mol/L � 17.4, respectively. Both the maximum andaverage total hemoglobin concentrations were significantly higher in the malignantgroup compared with the benign group (P � .001). When a maximum hemoglobinconcentration of 95 �mol/L was used as the threshold value, the sensitivity, speci-ficity, positive predictive value, and negative predictive value of optical tomographywere 100%, 96%, 73%, and 100%, respectively, and the sensitivity, specificity,positive predictive value, and negative predictive value of color Doppler US were63%, 69%, 19%, and 94%, respectively.

CONCLUSION: Findings indicate that optical tomography with US localization isfeasible for differentiating benign and early stage malignant breast lesions.© RSNA, 2005

Large numbers of breast biopsies are performed that yield benign results (1). The use ofultrasonography (US) as an adjunct to conventional mammography decreases the numberof benign biopsy results by enabling reliable differentiation of simple cysts from solidlesions (2–4). The US features of a solid lesion alone, however, are frequently not reliableenough to obviate biopsy (5,6).

57

Ra

dio

logy

Tumor angiogenesis is known to becritical for the autonomous growth andspread of breast cancers (7,8). Tumor an-giogenesis is a complex process that in-volves both the incorporation of existinghost blood vessels into the tumor and thecreation of tumor microvessels. This pro-cess is moderated by means of tumor an-giogenesis factors (9). In principle, thealtered hemodynamics that accompanytumor angiogenesis provide a basis fordiscriminating between malignant andbenign breast masses at color Doppler US(10). The diagnostic value of color Dopp-ler US in obviating biopsy, however, hasbeen limited (11,12).

Optical tomography, a new techniquethat employs diffused light in the near-infrared spectrum, provides functionalimages of tumor angiogenesis and tumorhypoxia. This technique has recentlybeen investigated in several clinical pilotstudies (13–19). When a single opticalwavelength is used, the optical absorp-tion, which is related to tumor angiogen-esis, can be measured. When multiple ap-propriately selected wavelengths areused, the optical absorptions at thesewavelengths can be measured, and theproportions of oxyhemoglobin and de-oxyhemoglobin can be calculated. Totalhemoglobin concentration and tumorhypoxia can be calculated from oxyhe-moglobin and deoxyhemoglobin distri-butions, which are highly correlated withlesion malignancy (20). Because the in-tense scattering that is caused by the tis-sue limits the imaging resolution and le-sion localization, optical tomographyhas not been widely used in clinical stud-ies. Coregistration of optical tomographywith US or magnetic resonance (MR) im-aging has shown promise in overcomingthese problems (17,19,21). The use of aflexible light guide that consists of opti-cal fibers makes optical imaging compat-ible with many other imaging modalitiesand allows for simultaneous imagingwith identical geometric conditions. In-formation on lesion structure that is pro-vided by other modalities can be used toassist optical imaging reconstruction andthereby reduce location uncertainty andimprove the quantification accuracy oflight.

A prior study on optical tomographydemonstrated that, by using the a prioriinformation on lesion location providedby coregistered US, the total hemoglobinconcentration in two early stage invasivecancers was twofold higher than that of asmall group of benign lesions (19). Thus,the purpose of our study was to investi-gate prospectively the feasibility of using

optical tomography with US localizationto differentiate malignant from benignbreast masses and to compare optical to-mography with color Doppler US.

MATERIALS AND METHODS

Patients and US

The study protocol was approved bythe local internal review board commit-tee and by the Human Subjects ResearchReview Board of Army Medical Researchand Materiel Command. Signed in-formed consent was obtained from allpatients. The study was compliant withthe Health Insurance Portability and Ac-countability Act. From May 2003 toMarch 2004, eligible patients were ini-tially referred for US-guided biopsy, andthe lesions were identified with US at thetime of the study. Patients with lesionsthat were not visible at US were not stud-ied. Three such patients, however, wereincluded in the study because the loca-tion of the lesions was well estimated ateither conventional mammography orMR imaging by one of the authors(E.B.C., A.A.C., or H.S.V.). The final studygroup consisted of 65 consecutive women(mean age, 51 years; age range, 24–80years) with 81 lesions. Eight additionalpatients were excluded from the finalstudy group. One patient did not have alesion that was visible at US after carefulexamination, and the optical data of twopatients were not recorded appropriatelyowing to equipment failure. Two patientshad lesions that were located close todark nipples, and optical images showednipple artifacts located at nipple-shadow-ing positions seen at US. Three patientshad small dense breasts, and the poorprobe-to-tissue contact produced imageartifacts.

US examinations were performed withspatial image compounding and a15-MHz linear transducer (15L8 Sequoia;Acuson, Mountain View, Calif). On thebasis of US, conventional mammo-graphic, and/or MR imaging findings,one of the authors (E.B.C., A.A.C., orH.S.V., all certified radiologists with 6–20years experience in US of the breast)scored each lesion before biopsy by usingthe Breast Imaging Reporting and DataSystem, or BI-RADS, classification. Cate-gory 1 lesions were classified as normal,category 2 as benign, category 3 as prob-ably benign, category 4 as suspicious formalignancy, and category 5 as highly sus-picious for malignancy. For color Dopp-ler US measurements, any persistentcolor Doppler signals were considered to

represent blood vessels. The location ofthe vessels with respect to the lesion wasdocumented as either peripheral to thelesion, inside the lesion, or both.

Near-Infrared System and OpticalImaging Method

The hand-held hybrid probe consistedof a commercially available US trans-ducer, which was located in the middleof the probe, and near-infrared source-detector light guides (optical fibers),which were distributed at the peripheryof the probe (Fig 1). The technical aspectsof the near-infrared imager have beenpreviously described in detail (22).Briefly, the imager consisted of 12 pairs ofdual wavelength (780-nm and 830-nm) la-ser diodes that were used as light sources,and the outputs of the diodes were coupledto the probe through optical fibers. On thereceiving side, eight photomultiplier tubeswere used to detect diffusely scattered lightfrom the tissue, and eight optical fiberswere used to couple the detected light tothe photomultiplier tubes. The outputs ofthe laser diodes were amplitude modulatedat 140 MHz, and the detector outputs weredemodulated to 20 kHz. Eight detectionsignals and one reference signal were si-multaneously amplified, sampled, and ac-quired into a personal computer (Dimen-sion XPS B800; Dell, Round Rock, Tex).The entire data acquisition took about 3–4seconds, which was fast enough to obtaindata from all patients. For each patient, USimages and optical measurements weresimultaneously acquired before biopsy,which was performed at multiple loca-tions, including the lesion site, a normal

Figure 1. Picture of combined probe and fre-quency domain near-infrared imager (arrow).Commercial US probe is located at center ofcombined probe, and optical source and detec-tor fibers are distributed at periphery of probe.Because diffused photons launched from asource and received by a detector travel in acurved path, lesions underneath US probe canbe imaged with high sensitivity.

58 � Radiology � October 2005 Zhu et al

Ra

dio

logy

symmetric region of the affected breast,and a normal region of the contralateralbreast in the same quadrant as the lesion.Two normal areas—that is, one area ofthe affected breast and one area of thecontralateral breast—were chosen to en-sure that one of the normal areas couldprovide a better reference for tissue ho-mogeneity. The final reference data setwas selected on the basis of the linearfitting result of the received amplitudeand phase profiles. The scattered fieldthat was used for optical imaging recon-struction was calculated as the differencebetween measurements obtained at thelesion site and those obtained at the ref-erence site.

One author (Q.Z.) performed opticalimaging. Details of the dual-mesh opticalimaging reconstruction algorithm havebeen described previously (21,23). Briefly,the near-infrared reconstruction uses USlocalization of lesions. The reconstructionalgorithm then segments the imaging vol-ume into a finer grid in the lesion regionidentified at US and into a coarser grid innonlesion regions. To account for the pos-sibility of larger angiogenesis extension inlesions identified at US, we used a muchlarger region of interest for mapping le-sions on the finer grid. For all optical im-ages, a 0.5 � 0.5 � 0.5-cm imaging grid wasused for the lesion, and a 1.5 � 1.5 �1.0-cm grid was used for the background.The maximum and average total hemoglo-bin concentrations of the lesion were mea-sured. Averages were computed inside thelesion, with the total hemoglobin concen-tration calculated within 50% of the max-

imum value (ie, within the full width athalf maximum [FWHM] region) (see Ap-pendix for additional details).

Because the hand-held probe could beeasily rotated or translated, at least threecoregistered US and near-infrared datasets were acquired at the lesion location,and the corresponding optical absorp-tion maps, as well as the distribution ofthe total hemoglobin concentration,were reconstructed by using coregisteredUS.

Statistical Analysis

For statistical analysis, sensitivity wascalculated as TP/(TP � FN), specificity asTN/(TN � FP), positive predictive valueas TP/(TP � FP), and negative predictivevalue as TN/(TN � FN), where TP repre-sents the number of true-positive find-ings, TN represents the number of true-negative findings, FP represents the num-ber of false-positive findings, and FNrepresents the number of false-negativefindings. Information obtained at biopsywas used as the reference standard. A Stu-dent t distribution was used to calculatethe statistical significance of the meanmaximum and mean average hemoglo-bin concentrations obtained in the ma-lignant group and the benign group.Commercially available statistical soft-ware (SAS, version 8.2; SAS, Cary, NC)was used for calculations, and P � .001was considered to indicate a statisticallysignificant difference.

RESULTS

Biopsy results for malignant and benigngroups are indicated in Tables 1 and Tables2. The malignant group included eightearly stage invasive carcinomas, and thebenign group included 20 fibroadenomas,15 cases of fibrocystic change, eight casesof fibrosis, eight other solid benign lesions,21 complex cysts, and one case of hyper-plasia. No significant differences weredemonstrated among findings in the be-nign group, but a more than twofoldhigher total hemoglobin concentration(Fig 2) was found in the malignant group(mean maximum [� standard deviation],122 �mol/L � 26.8; mean average, 88�mol/L � 24.5, as measured within FWHMregion) compared with the benign group(mean maximum, 55 �mol/L � 24.8;mean average, 38 �mol/L � 17.4, as mea-sured within FWHM region). Both themaximum and average total hemoglobinlevels were significantly higher in the ma-lignant group compared with the benigngroup (P � .001). When a maximum he-moglobin concentration of 95 �mol/L wasused as a threshold to separate malignantlesions from benign lesions, the near-infra-red data of two fibroadenomas and onecase of minimal nonatypical hyperplasiaand fibrosis (which was classified as fibro-sis) were above the threshold value (Fig 3).The sensitivity, specificity, positive predic-tive value, and negative predictive value ofoptical tomography were 100% (eight ofeight lesions), 96% (70 of 73 lesions), 73%(eight of 11 lesions), and 100% (70 of 70

TABLE 1Characteristics of Malignant Lesions as Determined at Biopsy, US, and Optical Tomography

Biopsy Finding and Lesion Diameter (cm)BI-RADS

ScoreAppearance of BloodFlow at Doppler US

Total Hemoglobin Concentration(�mol/L)

Maximum Average*

Intraductal and invasive mammary duct carcinoma0.6 (n � 1) 4 Inside and peripheral 147.9 109.5

Intraductal and invasive mammary duct carcinoma2.2 (n � 1) Known† Peripheral 103.8 69.7

Invasive duct carcinoma�1.0 (n � 2) 5 Peripheral 151.8 � 29.7 116.4 � 11.7

Infiltrating lobular carcinoma0.9 (n � 1) Known† No blood flow 102.1 68.7

Invasive mammary duct carcinoma and ductal carcinoma in situ1.1 (n � 1) 5 Peripheral 97.5 67.7

Infiltrating ductal carcinoma and ductal carcinoma in situ1.0 (n � 1) 4 No blood flow 100.0 68.4

Infiltrating mammary ductal carcinoma0.9 (n � 1) 5 No blood flow 108.5 78.7

Average 4.6 � 0.5 … 121.6 � 26.8 88.4 � 24.5

Note.—Data are presented � standard deviation.* Data were calculated within FWHM region.†Cancer was already designed in these two cases.

Volume 237 � Number 1 Benign versus Malignant Breast Masses � 59

Ra

dio

logy

lesions), respectively. There were threefalse-positive lesions (Table 3). When amean hemoglobin concentration of 66�mol/L was used as a threshold to separatemalignant lesions from benign lesions, thesame sensitivity, specificity, positive pre-dictive value, and negative predictive valuewere achieved.

Optical tomography provided highersensitivity and higher specificity thancolor Doppler US. Although blood flowwas demonstrated at color Doppler US infive of eight malignant lesions, bloodflow was also present in six (30%) of 20fibroadenomas, six (40%) of 15 fibrocys-tic changes, two (25%) of eight cases of

fibrosis, four (50%) of eight other solidbenign lesions, and four (19%) of 21complex cysts. The sensitivity, specific-ity, positive predictive value, and nega-tive predictive value of color Doppler USwere 63% (five of eight lesions), 69% (48of 70 lesions), 19% (five of 27 lesions),and 94% (48 of 51 lesions), respectively.

TABLE 2Characteristics of Benign Lesions as Determined at Biopsy, US, and Optical Tomography

Biopsy Finding andLesion Diameter (cm)

BI-RADSScore

Appearance of BloodFlow at Doppler US

Total Hemoglobin Concentration (�mol/L)

Maximum Average*

Fibroadenoma1.0 � 0.6 (n � 2) 4.0 � 0.0 Inside and peripheral 55.7 � 38.2 37.7 � 28.40.9 � 0.3 (n � 4) 4.0 � 0.0 Peripheral 54.9 � 9.6 38.7 � 7.40.8 � 0.4 (n � 14) 3.8 � 0.6 No blood flow 70.5 � 27.2 49.5 � 19.4

Average 3.9 � 0.5 … 65.9 � 25.5 46.1 � 18.3Fibrocystic change

0.6 � 0.2 (n � 5) 3.8 � 0.4 Peripheral 43.3 � 14.5 29.6 � 9.70.8 � 0.5 (n � 7) 3.9 � 0.4 No blood flow 56.5 � 14.0 40.3 � 10.40.8 � 0.3 (n � 2) 4.0 � 0.0 NA† 76.7 � 4.9 51.1 � 11.40.6 (n � 1) 4 Inside 68.7 48.5

Average 3.9 � 0.4 … 55.6 � 16.6 38.9 � 11.9Fibrosis

0.8 (n � 1) 4 Inside and peripheral 81.2 57.80.4 (n � 1) 4 Peripheral 49.0 31.01.0 � 0.8 (n � 5) 4.0 � 0.0 No blood flow 68.4 � 37.9 47.3 � 26.1NA (n � 1) 4 NA† 54.0 36.3

Average 4.0 � 0.0 … 66.2 � 31.2 45.2 � 21.3Tubular adenoma

0.4 (n � 1) 4 Inside and peripheral 94.1 65.2Intraductal papilloma

1.5 (n � 1) 4 Peripheral 38.3 26.6Tubular adenoma

0.4 (n � 1) 4 Peripheral 66.2 45.8Normal tissue

1.0 � 0.4 (n � 4) 3.5 � 0.6 No blood flow 42.8 � 23.2 29.3 � 14.2Ductal and stomal cells

0.3 (n � 1) 4 Inside 72.0 53.1Average 3.8 � 0.5 … 55.2 � 25.2 38.5 � 17.2Complex cyst

0.7 (n � 1) 4 Inside and peripheral 37.0 23.90.5 � 0.2 (n � 3) 2.7 � 0.6 Peripheral 46.9 � 9.6 32.5 � 7.80.8 � 0.4 (n � 17) 2.9 � 0.6 No blood flow 39.8 � 23.3 27.5 � 15.9

Average 3.0 � 0.7 … 40.7 � 21.3 28.0 � 14.6Hyperplasia

0.6 (n � 1) 3 No blood flow 56.2 38.0

Note.—Data are presented � standard deviation. NA � not applicable.* Data were calculated within FWHM region.† Lesions were not visible at US. Lesions location was estimated with MR imaging or conventional mammography.

TABLE 3Lesion Characteristics and Total Hemoglobin Concentration of Three False-Positive Findings

Biopsy Findings and Lesion Diameter(cm)

BI-RADSScore

Appearance of BloodFlow at Doppler US

Total Hemoglobin Concentration(�mol/L)

Maximum Average*

Fibroadenoma1.91 � 0.85 4 No blood flow 117.8 81.3

Fibroadenoma1.50 � 0.70 4 No blood flow 119.0 85.2

Minimal duct hyperplasia and fibrosis3.40 � 1.56 4 No blood flow 129.9 89.2

* Data were calculated within the FWHM region.


Ra

dio

logy

Because diffused light probes microvesseldensity rather than relatively large bloodvessels (as is the case with color DopplerUS), the sensitivity and specificity of op-tical tomography are expected to bemuch higher than those of color DopplerUS.

Typical examples of malignant lesionsare shown in Figures 4 and Figures 5;

examples of benign lesions are given inFigures 6–8.

DISCUSSION

US is frequently used as an adjunct toconventional mammography in differen-tiating simple cysts from solid lesions. US

also plays an important role in guidinginterventional procedures, such as needleaspiration, core-needle biopsy, and prebi-opsy needle localization (24). Previously,screening US has also been recom-mended for patients with dense breasts(25,26). We believe that our technique,which incorporates optical tomographyand US localization, has demonstratedgreat potential to noninvasively distin-guish malignant from benign breastmasses and thereby potentially reducethe number of benign biopsy results. In aprevious study (19), results obtainedfrom two invasive early stage cancers and17 benign lesions demonstrated that ma-lignant cancers measuring 1 cm in diam-eter have an average maximum total he-moglobin concentration of 119 �mol/L,whereas benign lesions have an averagemaximum total hemoglobin concentra-tion of 67 �mol/L. Thus, a nearly twofoldgreater total hemoglobin concentrationwas observed. Our study of eight earlystage cancers and 73 benign lesions ob-tained at different clinical sites furtherdemonstrates the diagnostic potential ofthis technique. In our current study, in-vasive cancers revealed a more than two-fold greater total hemoglobin concentra-tion. No false-negative results werefound. Because our samples were limited,a larger prospective clinical trial isneeded to validate these results.

In addition to the elevated hemoglo-bin concentration, information on thedistribution of hemoglobin—that is, onthe morphologic features of the le-sion—is important in reaching an appro-priate diagnosis. In general, the hemo-globin distribution of early stage invasivecancers appears isolated, and the lesion iswell resolved; the hemoglobin distribu-tion of solid benign lesions, however, ap-pears more diffused. For larger complexcysts, a ring-shaped structure is some-times observed around the lesions be-cause of the substantial water content.Thus, cross-referencing the hemoglobinlevel and distribution with the results ofstandard imaging techniques will likelyyield the most accurate diagnosis.

One potential limitation of optical to-mography is the difficulty of imaging oflesions that are located close to dark nip-ple-areolar tissue. Because the light ab-sorption of dark skin is high, the lightperturbations that are attributed to thelesions may be secondary to the pertur-bations that are caused by the nipple-areolar complex. Therefore, the opticalimaging algorithm may not reconstructthese lesions correctly. In our study, twocases were excluded from analysis be-

Figure 2. Bar graph demonstrates average maximum total hemoglobin concentration obtainedin seven groups. Group 1 represents carcinoma (121.6 �mol/L � 26.8); group 2, fibroadenoma(65.9 �mol/L � 25.5); group 3, fibrocystic changes (55.6 �mol/L � 16.6); group 4, fibrosis (66.2�mol/L � 31.2); group 5, other solid benign lesions (55.2 �mol/L � 25.2); group 6, complex cysts(40.7 �mol/L � 21.3); and group 7, hyperplasia (56.2 �mol/L � 0.0). Horizontal axis representsgroup number, and vertical axis represents maximum total hemoglobin concentration in micro-moles per liter. Error bars indicate standard deviations.

Figure 3. Scatter plot shows maximum total hemoglobin concen-tration for carcinoma (small ■ ), fibroadenoma (�), fibrocysticchanges (Œ), fibrosis (F), other solid benign lesions (short bar), com-plex cysts (long bar), and hyperplasia (large ■ ). Total hemoglobinconcentration is measured on vertical axis in micromoles per liter.


Ra

dio

logy

cause of this problem. A lighter-colorednipple-areolar complex did not appear tocause a problem. Also, lesions that werelocated more than 2 cm away from a darknipple did not cause artifacts.

Another potential problem is the poorprobe-to-tissue contact that is madewhile examining small dense breasts.Three cases were excluded from analysisbecause of this problem. Good probe-to-

tissue contact could be achieved by de-veloping probes of different sizes that aresuitable for smaller breasts. Currently, wehave designed an optical switch that al-lows the selection of different probes thatare suitable for the size of the examinedbreast.

Three benign lesions, including two fi-broadenomas, one case of minimal non-atypical duct hyperplasia, and one case of

fibrosis, showed a high total hemoglobinconcentration that suggested high micro-vessel density. The high total hemoglobinconcentration of fibroadenomas may beexplained by the results of a pilot study onvascular features of fibroadenoma (27). Inthis study (27), two different groups offibroadenomas were recognized. The firstgroup showed low microvascular perme-ability and a high extracellular volume

Figure 4. US image, optical absorption maps, and hemoglobin concentration maps of suspicious lesion at 3 o’clock position in left breast of45-year-old woman. (a) US image demonstrates ill-defined, slightly heterogeneous, and mildly hypoechoic mass (arrow) measuring 6 mm indiameter. Color Doppler US revealed large blood vessels both inside and at periphery of lesion. Pathologic analysis revealed intraductal and invasivemammary duct carcinoma (nuclear grade 3, histologic grade 3). (b, c) Reconstructed optical absorption maps obtained at (b) 780 nm and (c) 830nm. Optical absorption values range from 0 to 0.3 cm�1. The first section (Slice #1 in b and c) is a 9 � 9-cm spatial x-y image obtained at a depthof 0.5 cm, as measured from skin surface, and the last section (Slice #7 in b and c) is a spatial x-y image obtained 3.5 cm toward chest wall. Spacingbetween sections is 0.5 cm in propagation direction. Lesion was well resolved in section two (Slice #2). (d) Total hemoglobin concentration mapcomputed from absorption maps of b and c. Hemoglobin map revealed an isolated high concentration mass in section two (Slice #2). Maximumhemoglobin concentration was 149.9 �mol/L, and average hemoglobin concentration was 121.1 �mol/L, as calculated within FWHM region.Average total hemoglobin concentration of background areas outside FWHM region was 29.0 �mol/L. Vertical scale presents total hemoglobinconcentration in micromoles and ranges from 0 to 150 �mol/L.


Ra

dio

logy

fraction. At histopathologic analysis, le-sions with low vascular permeability hada lower density of small vessels. The sec-ond group showed higher microvascularpermeability and a lower extracellularvolume fraction. Lesions with higher mi-crovascular permeability had a higherdensity of small vessels than lesions inthe first group. In our study, the patientwith minimal nonatypical duct hyper-plasia and fibrosis showed significant en-hancement at MR imaging, suggestingincreased vascularity in the area of thelesion. This patient had undergone exci-

sional biopsy at the same location 2 yearsearlier, with similar pathologic findings.It was therefore unlikely that a clinicallyimportant lesion was missed during therecent core biopsy. The residual angio-genesis resulting from changes associatedwith the normal healing process may be apossible reason for the enhanced vascu-larity.

We do not envision that our techniquewill be used as a screening tool becauselesions must be visible at US to map thelesion and background regions for dual-mesh optical imaging reconstruction. If

lesion locations can be accurately esti-mated by using other imaging modali-ties, such as MR imaging, our techniquecould be expanded for lesion character-ization by using those modalities, as well.More precise depth information, how-ever, is required if data obtained withother modalities are not coregisteredwith the near-infrared data. The three re-ported cases of lesions seen at MR imag-ing or conventional mammography werereconstructed with accurate a prioriknowledge of lesion location.

Because of intense light scattering, op-tical tomography alone has not been

Figure 5. US image and hemoglobin concentration maps of infil-trating lobular carcinoma (nuclear grade 2–3) located at 12 o’clockposition in right breast of 37-year-old woman. (a) US image demon-strates lesion (arrow) measuring 1 cm in diameter. No blood vessels orblood flow were seen at color Doppler US. (b) Total hemoglobinconcentration map computed from absorption maps (not shown)obtained at 780 nm and 830 nm. Hemoglobin map reveals isolatedhigh concentration mass in sections three (slice #3) and four (slice #4).Maximum hemoglobin concentration was 102.1 �mol/L, and averagehemoglobin concentration was 68.7 �mol/L, as measured withinFWHM region. Vertical scale presents total hemoglobin concentra-tion in micromoles and ranges from 0 to 150 �mol/L.

Figure 6. US image and hemoglobin concentration maps of suspicious6-mm lesion located at 7 o’clock position in right breast of same patientas Figure 5. (a) US image demonstrates lesion (arrow). No blood vesselsor blood flow were seen at color Doppler US. Biopsy result revealedbenign nonproliferative fibroadipose breast tissue. (b) Total hemoglobinconcentration map computed from absorption maps (not shown) ob-tained at 780 nm and 830 nm. No resolvable lesion was found inhemoglobin map, and maximum total hemoglobin concentration wasonly 24.2 �mol/L. Vertical scale represents total hemoglobin concentra-tion in micromoles and ranges from 0 to 150 �mol/L.


Ra

dio

logy

widely used in clinical studies. The datain the published literature have been lim-ited to feasibility studies or case reports.In addition, the systems that are used toacquire optical tomographic data varyconsiderably. Most of the optical systemsuse transmission geometry, in which ei-ther the light sources and detectors aredeployed around the examined breast insingle or multiple rings (13,16) or thelight sources are deployed on one side ofthe breast and the detectors are deployed

on the opposite side with the breast com-pressed in between (17,18). It is clear thatwithout a priori knowledge of lesion lo-cations from other imaging modalities,no one has achieved significant and con-sistent improvements in accurate light

quantification and, therefore, in the dif-ferentiation of benign from malignantprocesses. Only two examples that wererelated to our study were found in theliterature (15,28). In one of the studies(15), the authors reported an average to-

Figure 7. US image and hemoglobin concentration maps of suspi-cious 1.9 � 1.1-cm lesion at 9 o’clock position in right breast of a75-year-old woman. (a) US image demonstrates lesion with solidcomponent adjacent to cyst (arrow). Large blood vessels were seen atperiphery of lesion during color Doppler US. Biopsy result revealedintraductal papilloma, with no evidence of atypical cells or malig-nancy. (b) Total hemoglobin concentration map computed fromabsorption maps (not shown) obtained at 780 nm and 830 nm.Vertical scale presents total hemoglobin concentration in micromolesand ranges from 0 to 150 �mol/L. Maximum total hemoglobin con-centration was 38.2 �mol/L, and average total hemoglobin concen-tration was 26.3 �mol/L, as measured within FWHM region. Totalhemoglobin distribution was diffused rather than localized, as seen incases of malignancy.

Figure 8. US image and hemoglobin concentration maps of 1.3 �1.0-cm lesion, simple cyst, and complex cyst located between 3o’clock and 4 o’clock positions in left breast of 59-year-old woman.(a) US image demonstrates simple cyst with adjacent complex cyst orsolid nodule (arrow). Complex cyst was aspirated completely, and noresidual abnormality was noted. (b) Total hemoglobin concentrationmap indicates low hemoglobin concentration at central location ofcyst (arrow) close to origin in section 2 (slice #2). Light absorptionmaps (not shown) indicated low light absorption. Incomplete ringwith higher light absorption and high hemoglobin concentrationwas observed surrounding cyst. Although measured total hemoglobinconcentration was highest in complex cyst owing to higher absorp-tion ring, distribution was obviously different from malignant cases(maximum total hemoglobin concentration, 94 �mol/L; average totalhemoglobin concentration, 62.8 �mol/L, as measured within FWHMregion). Cysts, in general, have low light absorption (and thereforelow hemoglobin concentration) owing to low water absorption inwavelength range studied. The higher absorption ring was likelycaused by the cyst wall. This example suggests that both hemoglobindistribution and threshold level must be evaluated for accurate diag-nosis of suspicious lesions.


Ra

dio

logy

tal hemoglobin concentration of 35�mol/L in a 2-cm ductal carcinoma insitu by using near-infrared measurementsalone. After reprocessing the same near-infrared data with an approximate lesiondepth obtained from a separate US im-age, the average total hemoglobin con-centration was increased to 67 �mol/L(28). This value was closer to the averageof 88 �mol/L, which was obtained ineight malignant lesions in our study.

In principle, deoxygenated and oxy-genated hemoglobin distributions can beestimated from absorption distributionsobtained at wavelengths of 780 nm and830 nm. The estimated deoxygenatedand oxygenated hemoglobin distribu-tions, however, were not robust becausethe negative coefficients in the computa-tion formulas (see Appendix) could leadto negative deoxygenated and/or oxy-genated hemoglobin values in some im-aging voxels. The total hemoglobin dis-tribution is more robust because of thepositive coefficients in the computationformula. Therefore, no deoxygenated oroxygenated hemoglobin distribution wascomputed alone.

In principle, the distribution of oxygensaturation can be estimated from twowavelengths (29). For the two wave-lengths (780 nm and 830 nm) that wereused in this study, the estimated oxygensaturation in normal background tissuewas not robust and could exceed thephysiologic range. Recently, we added a660-nm wavelength to the near-infraredsystem and found that the backgroundtissue oxygen saturation can be esti-mated more reliably from measurementsof 660 nm and 830 nm.

The reported malignant lesions weregenerally early stage cancers that wereapproximately 1 cm in diameter (exceptfor one that measured 2.2 cm in diame-ter). For larger cancers, the angiogenesisdistributions are complex and heteroge-neous (29). This heterogeneous distribu-tion is dependent on angiogenic factors(9) and is related to the incorporation ofexisting host blood vessels into tumorand the creation of tumor microvessels.Blood flow through these tumor vessels isheterogeneous. Some areas may havehigh flow, whereas others may haveslower flow and develop necrosis (30).Tumors with relatively poor blood perfu-sion may not receive adequate delivery ofsystemic therapy. This lack of perfusionmay be a factor in some patients whohave poor response to chemotherapytreatment (31,32). Optical tomographyfor the imaging of larger tumors may be avaluable technique for mapping tumor

angiogenesis and evaluating vascularchanges, as well as tumor hypoxia, dur-ing chemotherapy (29). This type of in-formation could also prove invaluable inmonitoring chemotherapy responses tobreast cancer treatment and assessingtreatment efficacy.

APPENDIX

A modified Born approximation was usedto relate the scattered field Usd (rsi, rdi, �) tothe light absorption variations �a

(r�) atwavelength in each volume element r� ofthe two regions within the sample. For scat-tered field Usd (rsi, rdi, �), � is the modula-tion frequency and i is the index of thesource detector pair for the optical sourceand detector positions rs and rd, respec-tively. The matrix form of image recon-struction is given by

�Usd M�1 � �WL, WB M�N�ML, MB N�1T ,

where WL and WB are weight matrices forthe lesion and background regions, respec-tively, N is the total number of imagingvoxels, T is the matrix transpose, and M isthe total number of measurements. ML andMB are the total absorption distributions ofthe lesion and background regions, respec-tively, which are calculated as

�ML � ��1L

�a�r��d3r�, . . . �

NL

�a�r��d3r��

and

�MB � ��1B

�a�r��d3r�, . . .�

NB

�a�r��d3r��,

where d3r� is the volume element. Weightmatrices are calculated on the basis of back-ground absorption �� a

and reduced scatter-ing �� s

measurements, which are obtainedfrom the normal contralateral breast. In-stead of reconstructing the �a

distributiondirectly (as is done in the standard Bornapproximation), the total absorption distri-bution M is reconstructed, and the total isdivided by the different voxel sizes of thelesion and background tissue to obtain a�a

distribution. By choosing a finer gridfor the lesion tissue and a coarser grid forthe background tissue, we can maintain thetotal number of voxels with unknown opti-cal absorption on the same scale as the totalmeasurements. As a result, the inverse prob-lem is less well defined. In addition, becausethe absorption coefficient of the lesion ishigher than that of the background tissue,the total absorption of the lesion over asmaller voxel is, in general, on the samescale as the total absorption of the back-ground tissue over a larger voxel. Therefore,

the matrix [ML, MB] is appropriately scaledfor inversion.

Because the major chromophores in thewavelength range studied were deoxygen-ated and oxygenated hemoglobin, we esti-mated deoxygenated hemoglobin (deoxyHb)and oxygenated hemoglobin (oxyHb) con-centrations at each imaging voxel by invert-ing the following equations voxel by voxel as

� �a1�r��

�a2�r�� εHb

1 , εHbO2

1

�εHb2 , εHbO2

2 � � � deoxyHb�r��oxyHb�r��

and

� deoxyHb�r��oxyHb�r��

�1 � εHbO2

2 , �εHbO2

1

�εHb2 , εHb

1 � � � �a1�r��

�a2�r�� ,

where �a1(r�) and �a

2(r�) are the absorptioncoefficients obtained at imaging voxel r�,and wavelengths 1 and 2 correspond to780 nm and 830 nm, respectively. Extinc-tion coefficients (�) for deoxygenated andoxygenated hemoglobin are given in theliterature (33). The total hemoglobin con-centration totalHb(r�), which is the sum ofdeoxyHb(r�) and oxyHb(r�), can be calcu-lated as follows:

totalHb�r�� 1

��εHbO2

2 � εHb2 ��a

1�r��

� �εHb1 � εHbO2

1 ��a2�r��

The perturbations at each wavelengththat were used to calculate absorption mapswere normalized as Usd (rsi, rdi, �) � [UL (rsi,rdi, �) � UN (rsi, rdi, �)] � [UB (rsi, rdi, �)/UN (rsi,rdi, �)], where UL (rsi, rdi, �) and UN (rsi, rdi, �)are the optical measurements obtained atthe lesion region and the normal region ofthe contralateral breast, respectively, and UB

(rsi, rdi, �) was the incident field that wascalculated by using the measured back-ground optical properties. This procedurecancels the unknown system gains associ-ated with different sources, detectors, andelectronic channels.

Acknowledgments: Roxanne P. Rotondaro,coordinator of the Partnership for Breast Careof Hartford Hospital, is greatly acknowledgedfor coordinating this research project. MarthaAhlquist, coordinator of the Clinical ResearchOffice of Hartford Hospital, is thanked for ob-taining patient consent. The US technologistsof the radiology department of Hartford Hos-pital are also thanked for their extremely help-ful assistance with US data acquisition and pa-tient scanning.

References1. Poplack SP, Tosteson AN, Grove MR, Wells

WA, Carney PA. Mammography in 53,803


Ra

dio

logy

women from the New Hampshire Mam-mography Network. Radiology 2000;217:832–840.

2. Hilton SV, Leopold GR, Olson LK, WillsonSA. Real-time breast sonography: applica-tion in 300 consecutive patients. AJR Am JRoentgenol 1986;147:479–486.

3. Rubin E, Miller VE, Berland LL, Han SY,Koehler RE, Stanley RJ. Hand-held real-time breast sonography. AJR Am J Roent-genol 1985;144:623–627.

4. Jackson VP. The current role of ultrasonog-raphy in breast imaging. Radiol Clin NorthAm 1995;33:1161–1170.

5. Venta LA, Dudiak CM, Salomon CG, FlisakME. Sonographic evaluation of the breast.RadioGraphics 1994;14:29–50.

6. Rahbar G, Sie AC, Hansen GC, et al. Benignversus malignant solid masses: US differen-tiation. Radiology 1999;213:889–894.

7. Schor AM, Schor SL. Tumour angiogenesis.J Pathol 1983;141:385–413.

8. Folkman J, Watson K, Ingber D, HanahanD. Induction of angiogenesis during thetransition from hyperplasia to neoplasia.Nature 1989;339:58–61.

9. Vaupel P, Kallinowski F, Okunieff P. Bloodflow, oxygen and nutrient supply, andmetabolic microenvironment of humantumors: a review. Cancer Res 1989;49:6449–6465.

10. Taylor KJW, Ramos I, Carter D, Morse SS,Snower D, Fortune K. Correlation of Dopp-ler US tumor signals with neovascularmorphologic features. Radiology 1988;166:57–62.

11. Wilkens TH, Burke BJ, Cancelada DA, JatoiI. Evaluation of palpable breast masseswith color Doppler sonography and grayscale imaging. J Ultrasound Med 1998;17:109–115.

12. Reinikainen H, Rissanen T, Paivansalo M,Paakko E, Iauhiainen J, Suramo I. B-mode,power Doppler and contrast-enhancedpower Doppler ultrasonography in the di-agnosis of breast tumors. Acta Radiol 2001;42(1):106–113.

13. Pogue BW, Poplack SP, McBride TO, et al.Quantitative hemoglobin tomographywith diffuse near-infrared spectroscopy:pilot results in the breast. Radiology 2001;218:261–266.

14. Franceschini MA, Moesta KT, Fantini S, etal. Frequency-domain techniques enhanceoptical mammography: initial clinical re-sults. Proc Natl Acad Sci U S A 1997;94:6468–6473.

15. Tromberg BJ, Shah N, Lanning R, et al.Non-invasive in vivo characterization ofbreast tumors using photon migrationspectroscopy. Neoplasia 2000;2(1-2):26–40.

16. Jiang H, Xu Y, Iftimia N, et al. Three-di-mensional optical tomographic imagingof breast in a human subject. IEEE TransMed Imaging 2001;20(12):1334–1340.

17. Ntziachristos V, Yodh AG, Schnall M,Chance B. Concurrent MRI and diffuse op-tical tomography of breast after indocya-nine green enhancement. Proc Natl AcadSci U S A 2000;97(6):2767–2772.

18. Grosenick D, Moesta KT, Wabnitz H, et al.Time-domain optical mammography: ini-tial clinical results on detection and char-acterization of breast tumors. Appl Opt2003;42(16):3170–3186.

19. Zhu Q, Huang MM, Chen NG, et al. Ultra-sound-guided optical tomographic imag-ing of malignant and benign breast le-sions. Neoplasia 2003;5(5):379–388.

20. Brown JM. The hypoxic cell. Cancer Res1999;59(23):5863–5870.

21. Zhu Q, Chen NG, Kurtzman SH. Imagingtumor angiogenesis by the use of com-bined near infrared diffusive light and ul-trasound. Opt Lett 2003;28(5):337–339.

22. Chen NG, Guo PY, Yan SK, Piao DQ, ZhuQ. Simultaneous near infrared diffusivelight and ultrasound imaging. Appl Opt2001;40(34):6367–6380.

23. Huang M, Zhu Q. A dual-mesh optical to-mography reconstruction method withdepth correction using a priori ultrasoundinformation. Appl Opt 2004;43(8):1654–1662.

24. Fornage BD, Coan JD, David CL. Ultra-sound-guided needle biopsy of the breastand other interventional procedures. Ra-diol Clin North Am 1992;30:167–185.

25. Bassett LW, Ysrael M, Gold RH, Ysrael C.Usefulness of mammography and sonog-raphy in women less than 35 years of age.Radiology 1991;180:831–835.

26. Kolb TM, Lichy J, Newhouse JH. Occultcancer in women with dense breast: detec-tion with screening US—diagnostic yieldand tumor characteristics. Radiology 1998;207:191–199.

27. Weinstein D, Strano S, Cohen P, Fields S,Gomori JM, Degani H. Breast fibroade-noma: mapping of pathophysiologic fea-tures with three-time-point, contrast-en-hanced MR imaging—pilot study. Radiol-ogy 1999;210:233–240.

28. Holboke M, Tromberg B, Li X, et al. Three-dimensional diffuse optical mammogra-phy with ultrasound localization in hu-man subject. J Biomed Opt 2000;5(2):237–247.

29. Zhu Q, Kurtzma SH, Hegde P, et al. Utiliz-ing optical tomography with ultrasoundlocalization to image heterogeneous he-moglobin distribution in large breast can-cers. Neoplasia 2005;7(3):263–270.

30. Mankoff DA, Dunnwald LK, Gralow JR, etal. Blood flow and metabolism in locallyadvanced breast cancer: relationship to re-sponse to therapy. J Nucl Med 2002;43(4):500–509.

31. Jain RK. Hemodynamic and transport bar-riers to the treatment of solid tumors. Int JRadiat Biol 1991;60:85–100.

32. Weidner N, Semple JP, Welch WR, Folk-man J. Tumor angiogenesis and metasta-sis: correlation in invasive breast carci-noma. N Engl J Med 1991;324(1):1–8.

33. Cope M. The application of near infraredspectrscopy to non-invasive monitoring ofcerebral oxygenation in the newborn in-fant [PhD dissertation]. London, England:University College, 1991.


Ra

dio

logy

radiology paper

Documents