12
Diffuse optical tomography of breast cancer during neoadjuvant chemotherapy: A case study with comparison to MRI Regine Choe, a! Alper Corlu, Kijoon Lee, Turgut Durduran, Soren D. Konecky, and Monika Grosicka-Koptyra Department of Physics and Astronomy, University of Pennsylvania, 209 South 33rd Street, Philadelphia, Pennsylvania 19104-6396 Simon R. Arridge Department of Computer Science, University College London, London, WC1E 6BT, United Kingdom Brian J. Czerniecki and Douglas L. Fraker Department of Surgery, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, Pennsylvania 19104 Angela DeMichele Department of Medicine (Heme/Onc), Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, Pennsylvania 19104 Britton Chance Department of Biochemistry and Biophysics, University of Pennsylvania, 3620 Hamilton Walk, Philadelphia, Pennsylvania 19104 Mark A. Rosen Department of Radiology, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, Pennsylvania 19104 Arjun G. Yodh Department of Physics and Astronomy, University of Pennsylvania, 209 South 33rd Street, Philadelphia, Pennsylvania 19104-6396 sReceived 17 November 2004; revised 13 January 2005; accepted for publication 14 January 2005; published 29 March 2005d We employ diffuse optical tomography sDOTd to track treatment progress in a female subject presenting with locally advanced invasive carcinoma of the breast during neoadjuvant chemo- therapy. Three-dimensional images of total hemoglobin concentration and scattering identified the tumor. Our measurements reveal tumor shrinkage during the course of chemotherapy, in reasonable agreement with magnetic resonance images of the same subject. A decrease in total hemoglobin concentration contrast between tumor and normal tissue was also observed over time. The results demonstrate the potential of DOT for measuring physiological parameters of breast lesions during chemotherapy. © 2005 American Association of Physicists in Medicine. fDOI: 10.1118/1.1869612g Key words: diffuse optical tomography, neoadjuvant chemotherapy, breast cancer, near infrared imaging, breast imaging I. INTRODUCTION Neoadjuvant chemotherapy si.e., preoperative chemotherapyd is an important therapeutic approach for women with locally advanced breast cancer sLABCd. LABC generally refers to lesions larger than 5 cm that may or may not have spread beyond the breast and axillary lymph nodes. If the patient responds to neoadjuvant chemotherapy, the size of the pri- mary tumor decreases, facilitating better control through sur- gery while potentially eradicating micrometastatic disease. 1 Neoadjuvant chemotherapy enables a higher percentage of patients to undergo breast conservation therapy without negatively impacting local recurrence rates or long-term out- come when compared with adjuvant chemotherapy. 2 Addi- tionally, the neoadjuvant setting provides a means to monitor the effectiveness of chemotherapy by observing its effects on the primary tumor in vivo. Information about tumor response during chemotherapy may be useful for treatment optimization. Physical examina- tion, as well as mammographic and ultrasonographic evalu- ations sometimes have limited utility for assessing tumor re- sponse due to chemotherapy-induced fibrosis. 3–5 While magnetic resonance imaging sMRId has proved useful for defining the extent of residual disease when compared with pathology, 6,7 and while dynamic contrast-enhanced magnetic resonance imaging sDCE-MRId has demonstrated ability to monitor tumor size and vascularity during neoadjuvant che- motherapy using gadolinium contrast agents, 7–9 the high cost and invasiveness of these methods render them impractical for frequent se.g., dailyd monitoring. Frequent monitoring of vascularity is important. Indeed a reduction of tumor angio- genesis from neoadjuvant chemotherapy in combination with hormone therapy has been confirmed by pathologic 1128 1128 Med. Phys. 32 4, April 2005 0094-2405/2005/324/1128/12/$22.50 © 2005 Am. Assoc. Phys. Med.

Diffuse optical tomography of breast cancer during neoadjuvant

  • Upload
    others

  • View
    5

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Diffuse optical tomography of breast cancer during neoadjuvant

Diffuse optical tomography of breast cancer during neoadjuvantchemotherapy: A case study with comparison to MRI

Regine Choe,a! Alper Corlu, Kijoon Lee, Turgut Durduran,Soren D. Konecky, and Monika Grosicka-KoptyraDepartment of Physics and Astronomy, University of Pennsylvania, 209 South 33rd Street, Philadelphia,Pennsylvania 19104-6396

Simon R. ArridgeDepartment of Computer Science, University College London, London, WC1E 6BT, United Kingdom

Brian J. Czerniecki and Douglas L. FrakerDepartment of Surgery, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia,Pennsylvania 19104

Angela DeMicheleDepartment of Medicine (Heme/Onc), Hospital of the University of Pennsylvania, 3400 Spruce Street,Philadelphia, Pennsylvania 19104

Britton ChanceDepartment of Biochemistry and Biophysics, University of Pennsylvania, 3620 Hamilton Walk, Philadelphia,Pennsylvania 19104

Mark A. RosenDepartment of Radiology, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia,Pennsylvania 19104

Arjun G. YodhDepartment of Physics and Astronomy, University of Pennsylvania, 209 South 33rd Street, Philadelphia,Pennsylvania 19104-6396

sReceived 17 November 2004; revised 13 January 2005; accepted for publication 14 January 2005;published 29 March 2005d

We employ diffuse optical tomographysDOTd to track treatment progress in a female subjectpresenting with locally advanced invasive carcinoma of the breast during neoadjuvant chemo-therapy. Three-dimensional images of total hemoglobin concentration and scattering identified thetumor. Our measurements reveal tumor shrinkage during the course of chemotherapy, in reasonableagreement with magnetic resonance images of the same subject. A decrease in total hemoglobinconcentration contrast between tumor and normal tissue was also observed over time. The resultsdemonstrate the potential of DOT for measuring physiological parameters of breast lesions duringchemotherapy. ©2005 American Association of Physicists in Medicine.fDOI: 10.1118/1.1869612g

Key words: diffuse optical tomography, neoadjuvant chemotherapy, breast cancer, near infraredimaging, breast imaging

yallyoreatien

prsurse.gehouout-

nitots o

rapyina-valu-r re-

rwithetictoche-tcticalofgio-with

I. INTRODUCTION

Neoadjuvant chemotherapysi.e., preoperative chemotherapdis an important therapeutic approach for women with locadvanced breast cancersLABCd. LABC generally refers tlesions larger than 5 cm that may or may not have spbeyond the breast and axillary lymph nodes. If the paresponds to neoadjuvant chemotherapy, the size of themary tumor decreases, facilitating better control throughgery while potentially eradicating micrometastatic disea1

Neoadjuvant chemotherapy enables a higher percentapatients to undergo breast conservation therapy witnegatively impacting local recurrence rates or long-termcome when compared with adjuvant chemotherapy.2 Addi-tionally, the neoadjuvant setting provides a means to mothe effectiveness of chemotherapy by observing its effec

the primary tumorin vivo.

1128 Med. Phys. 32 „4…, April 2005 0094-2405/2005/32 „4…/

dti--

oft

rn

Information about tumor response during chemothemay be useful for treatment optimization. Physical examtion, as well as mammographic and ultrasonographic eations sometimes have limited utility for assessing tumosponse due to chemotherapy-induced fibrosis.3–5 Whilemagnetic resonance imagingsMRId has proved useful fodefining the extent of residual disease when comparedpathology,6,7 and while dynamic contrast-enhanced magnresonance imagingsDCE-MRId has demonstrated abilitymonitor tumor size and vascularity during neoadjuvantmotherapy using gadolinium contrast agents,7–9 the high cosand invasiveness of these methods render them imprafor frequentse.g., dailyd monitoring. Frequent monitoringvascularity is important. Indeed a reduction of tumor angenesis from neoadjuvant chemotherapy in combination

hormone therapy has been confirmed by pathologic

11281128/12/$22.50 © 2005 Am. Assoc. Phys. Med.

Page 2: Diffuse optical tomography of breast cancer during neoadjuvant

odseo-

terdi-in-

tionre-

res

by

mue cation

obinde

lied

ter-ceny

ob-

-andeas

yy

Ththe

rtiesabo

ge-to

-s thehichuti

entain

awen

ofreeofree

mo-

glo-se oflumewith

. 1.char-tomscy-

ils of

sub-box.

weenthe

reastfluidThe

tinkn

, andated,ntohichgter-

ent.sioning in

9 FDCCD

1129 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1129

analysis.10 Therefore, new noninvasive imaging methwith capabilities for monitoring vascularity responses to nadjuvant chemotherapy are desirable.

Diffuse optical tomographysDOTd11 is a relatively newimaging modality with potential to follow therapy at shortime intervals. DOT provides physiological informationrectly related to tumor vascularity and oxygenation, withexpensive, nonionizing, and noninvasive instrumentaDOT utilizes light sources in the near-infrared spectralgion s600–1000 nmd wherein the major tissue chromophoare oxy-hemoglobinsHbO2d, deoxy-hemoglobinsHbd, water,and lipid. In this regime, photon transport is dominatedscattering, and light can penetrate centimetersse.g., 10 cmdbelow the tissue surface. Using optical measurements attiple source-detector positions on the tissue surfaces, onreadily reconstruct the internal distribution of the absorpcoefficientsmad and the reduced scattering coefficientsms8d inthree dimensions. Physiological images of total hemoglconcentration, blood oxygenation, water, and lipids arerived from this information. Thus far DOT has been appin breast cancer imaging,12–23 brain functional imaging,25,27

stroke detection,24,26 muscle functional study,28,29 photody-namic therapy,30,31 and radiation therapy monitoring.68

Not surprisingly, use of DOT for detection and characization of breast cancer has been of great reinterest.14,17–22,32–41The functional information provided bDOT is complimentary to the morphological informationtained by traditional breast imaging modalitiessi.e., x-raymammogram, ultrasound, MRId. For example, clinical studies employing three-dimensional DOT have identifiedquantified total hemoglobin concentration in human brcancer,18,32 correlating well with histologic parametersse.g.,blood vessel densityd.14 Jakubowskiet al.41 have recentldemonstrated the capability of diffuse opticalspectroscopsi.e., as opposed to diffuse opticalimagingd for monitoringneoadjuvant chemotherapy in a breast cancer patient.important paper introduced a new clinical application tofield. However, quantification of breast cancer propefrom spectroscopic data alone requires assumptionstissues se.g., homogeneous media, etc.d which are oftenquestionable.42 In addition, the remission measurementometry used in their experiments is primarily sensitivepalpable, near-surface tumors.

The potential of diffuse opticalimaging as a chemotherapy monitoring tool has not yet been explored, and isubject of this contribution. We present a case study wdemonstrates the feasibility of this approach. We havelized a four-wavelength near-infrared hybrid DOT instrumwith continuous wave transmission and frequency-domremission detection for breast imaging.12 Our subject hadlocally advanced invasive ductal carcinoma, and underfour cycles of adriamycin plus cytoxan and four cyclestaxotere prior to surgery. DCE-MRI were performed at thtime points throughout the therapy. After completionadriamycin cycles, we tracked the subject with DOT at thtime points. Three-dimensional DOT images of total he

globin concentration, oxygenation, and scattering were re

Medical Physics, Vol. 32, No. 4, April 2005

.

l-n

-

t

t

is

ut

-

t

constructed. We found that tumor volume and total hemobin tumor-to-normal contrast decreased over the courneoadjuvant chemotherapy. Furthermore, tumor vochanges measured by DOT showed good correlationDCE-MRI measurements of the same subject.

II. METHODS

A. DOT instrumentation

A schematic of the DOT instrument is shown in FigThis parallel-plane DOT system has been extensivelyacterized for breast cancer imaging using tissue phanand normal breast.12 The hybrid system combines frequendomain sFDd remission and continuous-wavesCWd trans-mission detection. We briefly summarize the salient detathis instrumentation below.

The table was designed to accommodate the femaleject in prone position with her breasts inside the breastA breast is typically centered and softly compressed beta movable compression plate and a viewing window withcompression distance ranging from 5.5 to 7.5 cm. The bbox was coated black and designed to hold a matchingwhich has optical properties similar to human tissue.matching fluid was composed of Liposyn IIIs30%, AbbotLaboratories, Chicago, ILd as a scattering agent and IndiasBlack India 4415, Sanford, Bellwood, ILd as an absorptioagent.

Four laser diode sources, operating at 690, 750, 786830 nm, respectively, were sinusoidally intensity modulat 70 MHz. A series of optical switchessDiCon Fiber OpticsRichmond, CAd were used to direct the light sources idifferent source positions on the compression plate wcontained 45 fibers arranged in a 935 grid as shown in Fi1sbd. Nine 3 mm diameter detection fiber bundles were in

FIG. 1. Schematic of parallel plate diffuse optical tomography instrumsad Frequency-domainsFDd remission and continuous-wave transmismeasurements were performed simultaneously on a female subject lyprone position.sbd The source plate contains 45 source positions anddetectors. 984 detection points with 3 mm spacing were selected fromdata for image reconstruction.

-laced in a 333 grid on the compression plate and connected

Page 3: Diffuse optical tomography of breast cancer during neoadjuvant

yneasemeartiethe

340assion70

derarci-

ofof

d

asekudyRI

r Rerap

-nd.

ing-

eor

at

igh-tedperad-

thein-

fash-agit-ne.rmal,

inisto-

wasy of

tu-t

zedthewas

n byingnure-d on

er-kthehest-dary.eeks

e-

o-

ughginge-andmea-

reDOT

nd,the

r to

1130 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1130

to the frequency-domain detection module. Homodtechniques43 were utilized to extract the amplitude and phof the detected remission signal. The frequency-domainsurements provided accurate quantification of bulk propeof the human tissue and matching fluid, thus improvinginitial guess used for the image reconstruction.

A lens-coupled 16-bit charge-coupled devicesCCDd cam-era sRoper Scientific, Trenton, NJ, VersArray:1300F, 131300 pixeld system with an antireflection coated glviewing window was used for collecting CW transmissdata. 232 on-chip binning of the CCD gave 53400 pixels over a detection area of 15.639.0 cm.

B. Neoadjuvant chemotherapy and MRI protocol

A 35-year-old premenopausal Caucasian female unwent neoadjuvant chemotherapy for invasive ductal cnoma in her left breast at the Hospital of the UniversityPennsylvaniasPhiladelphia, PAd. The therapy consistedfour cycles of doxorubicinsadriamycin, 60 mg/m2d pluscyclophos-phamidescytoxan, 600 mg/m2, regimen denoteas ACd followed by docetaxelstaxotere, 100 mg/m2, regi-men denoted asTd. Herein we will refer to the treatment“AC” followed by “T.” Each cycle was taken at three weintervals. She participated in the MRI research stCALGB150007/150012: “Contrast Enhanced Breast Mand Correlative Science Studies to Characterize Tumosponse in Patients Undergoing Neoadjuvant Chemothefor Locally Advanced Breast CancersI-SPYd” sPI:L. Esserman, MD, MBAd. The timing diagram for chemotherapy aimaging measurementssMRI and DOTd is provided in Fig2.

DCE-MRI measurements were performed at the followtime points: one week before chemotherapysprechemotherapyd, week 12 following completion ofAC, but prior toinitiation of taxoteresTd therapy, and week 23 following thcompletion of taxotere therapy, but prior to surgical tumremovalsmastectomyd. MRI of the breast was performed1.5 T sGeneral Electric, Signa, Milwaukee, WId using in-house sagittal compression receiver coils.44 At each timepoint, the DCE-MRI measurement consisted of sagittal hresolution thin section three-dimensional T1-weighspoiled gradient echo imaging of the affected breastformed before, and twice following, bolus intravenousministration of 0.1 mmol/kg gadolinium diamidesOmnis-

FIG. 2. Neoadjuvant chemotherapy timing diagram. Four cycles ofACsadriamycin+cytoxand therapy were followed by four cycles of taxotetherapy, and then by a mastectomy. Arrows indicate timing of MRI andmeasurements.

can©, Nycomed, Princeton, NJd.

Medical Physics, Vol. 32, No. 4, April 2005

-s

-

-y

-

The patient underwent simple mastectomy followingeighth chemotherapy cycle. The surgical sample wasspected grossly and margins were inked in a standardion. The sample was then oriented anatomically, and stally sectioned to correspond to the MRI imaging plaGross inspection revealed an indurated area of abnofirm tissue in the superior breast measuring 1.5 cmanterior-posterior dimension. This area was evaluated hlogically with Hematoxylin and Eosin staining.

C. DOT protocol

Before each DOT measurement, informed consentobtained from the patient in accordance with UniversitPennsylvania Institutional Review Board. Based on themor location identified by palpationfFig. 3sadg, the breasposition with respect to the viewing window was optimifFig. 3sbdg. Then a soft compression was applied to holdbreast in a stable position. The compression distancefixed at 7.5 cm. A snapshot of the breast outline was takethe CCD camera before filling the box with the matchfluid fFig. 3sbdg. After filling, the diffuse optical image scawas conducted for 12 min. After human subject measments, reference optical measurements were performethe box filled completely with matching fluid. For the refence measurements, a siliconesRTV12 with carbon blacand TiO2d block was placed on top of the box to extenddiffuse medium in a manner analogous to the subject’s cwall, and thus avoid signal saturation due to the air bounDOT measurements were performed at 10, 14, and 19 wafter the firstAC cycle ssee Fig. 2d. These time points corrspond to after the fourthAC sfourth chemotherapyd, the firsttaxoteresfifth chemotherapyd, and the third taxotere chemtherapy cyclessseventh chemotherapyd, respectively.

D. MRI data analysis

Tumor measurements by MRI were performed throanalysis of enhancing tumor pixels. Subtraction imaspost-gadolinium minus pregadoliniumd was employed as rquired. Maximum intensity projections in the sagittalaxial planes were obtained to facilitate accurate tumor

FIG. 3. Tumor location.sad According to x-ray mammogram, ultrasouand MRI, the primary tumor was located at the 12 o’clock position inleft breast.sbd Photo of breast outline in caudal-cranial viewsfoot to headviewd. The location of the tumor was approximated by palpation prioDOT measurement and is indicated by a circle.

surement in three orthogonal dimensions. A radiologist with

Page 4: Diffuse optical tomography of breast cancer during neoadjuvant

ogog a

et os.

bu-ata

fth,

sitere-

ntifts.

taredffu-ton:

-

r-ssue

etinc-g

s in

tri-

alpec-

theness isunc-

elring

i-

ments,ingatedized

thised byulatethe

g.on-

torhile

oc-enta-houtas ain

twoereents

ct tol.

1131 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1131

breast MRI experience measured the tumor in three orthnal planes. Tumor volume was then computed assuminellipsoid shape.

E. DOT transillumination

The sequential nature of data acquisition produces a stwo-dimensionals2Dd continuous wave CCD intensity mapWe define the detected intensity at detection positionr d dueto source numbers to beIss,r dd. A two-dimensionaltransil-lumination picture is constructed by summing the contrition of all sources, and normalizing with reference dI0ss,r dd, i.e.,

Tsr dd = − logS os

Ns Iss,r dd

os

Ns I0ss,r ddD . s1d

Figure 4 shows transillumination pictures after fourth, fiand seventh chemotherapy cycle.

The transillumination picture offers a quick compolook at the attenuation level of the raw breast data withspect to the reference data. It also enabled us to idesurface features which may correlate with image artifac

F. DOT data analysis: 3D reconstruction

Three-dimensionals3Dd images are derived from the dausing the diffusion model. The propagation of near-infralight in biological tissue can be modeled by a photon dision equation.45–47For the continuous wave case, the photransport description simplifies to the following equation

¹ · fDsr ,ld ¹ Fsr ,ldg − masr ,ldFsr ,ld = − Ssr ,ld. s2d

Here r denotes position in space,Dsr ,ld>f1/3ms8sr ,ldgsRef. 48d is the light diffusion coefficient,ms8sr ,ld is thereduced scattering coefficient,masr ,ld is the absorption co

2

FIG. 4. Transillumination of breast at 830 nm normalized with respereference measurement, offering a composite view of attenuation leve

efficient, Fsr ,ld is the photon fluence ratesfW/m gd, and

Medical Physics, Vol. 32, No. 4, April 2005

-n

f

y

Ssr ,ld is the isotropic light source termsfW/m3gd withwavelengthl. The absorption coefficientmasr ,ld and re-duced scattering coefficientms8sr ,ld are the two most impotant parameters connecting the diffusion model to tiphysiology via the relations: masr ,ld=ol=1

L elsldClsr d+ma

backgroundsr ,ld and ms8sr ,ld=Asr dl−bsr d. Here we assumL chromophores contribute to absorption, each with extion coefficientelsld and concentrationClsr d. The scatterinprefactorAsr d and scattering powerbsr d give informationabout the size and density distribution of scatterertissue.49,50

For image reconstruction we start with some initial disbution of Clsr d, Asr d, and bsr d, and then solve Eq.s2d forFsr ,ld using a finite elementsFEMd based numericsolver.51 We next update these parameters using a multistral inverse approach52,53 until good agreement betweencalculatedFcsr ,ld and measuredFmsr ,ld fluence rates othe boundaries is obtained. Specifically, the update proctreated as an optimization problem where an objective ftion

x2 =1

2on=1

N

oj=1

S

oi=1

Mj FlogSFmsr i,r j,lndFm

0 sr i,r j,lndFc

0sr i,r j,lndFcsr i,r j,lndDG2

, s3d

is minimized iteratively using a gradient-based schem54

suited to the multispectral method.52,53 Here N is the totanumber of measurement wavelengths,S is the total numbeof sources, andMj is the number of detectors correspondto sourcej above the noise floor.r i is the ith detector postion and r j is the j th source position.Fm

0 and Fc0 are the

measured and calculated fluence for reference measurerespectively, with the breast box entirely filled with matchfluid. The objective function is then modified to incorporspatially variant regularization12,55 with an envelope-guideapproach56 where regularization parameters are optimwithin the overall iterative process.57

The reconstruction volume was a 16 cm37.5 cm316 cm region, extending into the chestwall area. Involume, a finite element mesh with 58087 nodes was usthe finite element method based forward solver to calcFc. Fm was constructed by sampling and smoothingCCD data on a 41324 grid ftotal 984 detection points, Fi1sbdg with 3 mm spacing for each source. Hemoglobin ccentrationsCHbsr d, CHbO2

sr d, and the scattering prefacAsr d were chosen as unknowns to be reconstructed wother variables such as water concentrationCH2Osr d and lipidconcentrationClipidsr d and scattering power,bsr d were fixedas described below.

Since our particular imaging geometry involves spacecupied by both breast and matching fluid, image segmtion of the breast and matching fluid was used througour calculations. The breast region was approximatedthree-dimensionals3Dd half-ellipsoid based on its outlinethe photograph taken prior to the measurement scanfFig.3sbdg. To assign background and initial values to theregions, bulk optical properties of the matching fluid wderived from frequency-domain reference measurem

made on the box when it was completely filled with match-
Page 5: Diffuse optical tomography of breast cancer during neoadjuvant

romt suthetral

rands

dmeaatch

ionn

l re-

thea 1w

ach

y ofSin

regue in

ion

o-.RI

rma-rdi-alingnnotom-rans-n-

aridth

fined

r

theto

entersthisOTthescal-t ac-dingasticsion

holdaram-

ges.ceide is

ans-o the

1132 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1132

ing fluid. The breast optical properties were derived ffrequency-domain measurements in contact with breasface. A homogeneous semi-infinite analytic solution offrequency-domain diffusion equation with a multispecapproach58 was utilized to fit directly for the bulkCHb,CHbO2

, CH2O, A, andb in each region. The scattering powebwas allowed to take on different values in the breastmatching fluid, respectively.ma

backgroundinside the breast wathen fixed as a combination of 15% watersestimated fromfrequency-domain measurement of bulkCH2Od and 57% lipidabsorptionsfrom literature59–61d. ma

backgroundfor the matchingfluid region was fixed at the fitted bulkma of the matchingfluid. For the initial guess,CHb, CHbO2

, andA were assigneto the breast and the matching fluid based on the bulksurements, e.g., zero hemoglobin concentration in the ming fluid region. After the reconstruction ofCHbsr d, CHbO2

sr d,and Asr d, 3D images of total hemoglobin concentratfTHCsr d=CHbsr d+CHbO2

sr dg, blood oxygenation saturatiofStO2sr d=CHbO2

sr d /THCsr dg, and scattering fms8sr ,ld=Asr dl−bsr dg were constructed.

G. DOT image orientation

Figure 5 shows the orientation of three-dimensionaconstructed DOT images. A series of slices along they axisare arranged from left to right, from the source plane todetector plane, respectively. Each slice represents311 cm image in thex-z plane, in a caudal-cranial viesi.e., from foot to head, same as the CCD camera viewd. Theorientation of each image is such that the left side of eimage slice is lateralstoward outer side of breastd and theright side is medialstoward middle of the breastsd for the leftbreast, and vice versa for the right breast. For simplicitpresentation, slices at selected intervals are presented.the reconstructed data at FEM nodes is interpolated tolar grid of 0.2 cm spacing, each slice has 0.2 cm pixel sizboth x andz directions.

H. Image correlation analysis between MRI and DOT

As shown in Fig. 6sad, there are three standard orientat

FIG. 5. Orientation of three-dimensional reconstructed DOT imaCaudal-cranial slicessfoot to headd were arranged left to right, from sourto detector plane. The left side of each image is lateral and right smedial.

of views in tomography. Since the compression schemes o

Medical Physics, Vol. 32, No. 4, April 2005

r-

--

6

ce-

MRI ssagittald and DOTsaxiald are different, a true one-tone image comparison without distortion is not possible

In order to compare the tumor positions obtained by Mwith those obtained by DOT, one must derive a transfotion relating coordinates in the MRI axial image to coonates in the DOT axial image. We developed a simple sctransformation for this purpose, which by its nature caaccount for all of the deformations arising from breast cpression in orthogonal directions. The schematic of the tformation is illustrated in Fig. 6sbd. First, the breast dimesions of corresponding “central” image slicessi.e., slicescontaining the nippled were compared. We derived linescale factors from the ratio of breast length and breast win these corresponding central slices. In particular we descale factors a=XDOT

max /XMRImax, b=YDOT

max /YMRImax, and g

=ZDOTmax /ZMRI

max, where Ximax, Yi

max, Zimax is the longest linea

dimension of the breast in theX, Y, Z directions of theithsMRI or DOTd image.

We located the tumor center and tumor boundaries inMRI image by finding the region with high intensity duelarge gadolinium uptake. We then rescaled the tumor ccoordinater MRI

tumor to r DOTtumor by multiplying by scaling factor

si.e., a, b, andgd. The tumor boundaries are defined inrescaled fashion as well. Importantly, the tumor in the Dimage lies approximately within the volume defined byrescaling based on prechemotherapy MRI. We note thising approach assumes linear deformation and does nocount for elastic differences between tumor and surrountissue. Future models will be developed to account for elvariations and deformation due to the different compresschemes.

I. DOT tumor volume estimation

To define a tumor volume in the DOT images, a thresbased on the standard deviations of the reconstructed p

FIG. 6. sad Orientation of MRI view.sbd A sagittal view of MRI stopd andDOT sbottomd showing differences in compression. A linear scaling trformation scheme was utilized to find the tumor center with respect tnipple in DOT imagesr DOT

tumord from MRI r MRItumor.

feters was introduced. First, DOT parameters were decom-

Page 6: Diffuse optical tomography of breast cancer during neoadjuvant

for

tof a

-re-n

inn a

on

andosi-phy

argestmesize

ore

cm inter

al-cranial

1133 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1133

posed intoma andms8 at 690, 750, 786, and 830 nm. Theneach imagesTHC, StO2, CHb, CHbO2

, andma, ms8 at the fourwavelengthsd an averagespd and a standard deviationsspdwas calculated. The tumor region was defined byp.p+23sp, since values greater than 23sp have a 95% chancebe different from the average with the assumption oGaussian distribution. The average tumor volumesand standard deviationd was calculated by averaging all tumorgions defined by all images except StO2 swhere the variatiodid not exceed 23spd. Relative THC srTHC=THCtumor/THCnormald was calculated by averaging THCthe average tumor region and outside the tumor regio

FIG. 7. Three-dimensional reconstructed total hemoglobin concentratiin caudal-cranial view, from left to right. DOT images corresponding t

FIG. 8. Three-dimensional reconstructed images ofms8 at 786 nm. Image sl

view, from left to right. DOT images corresponding to after 4th, 5th, and 7

Medical Physics, Vol. 32, No. 4, April 2005

s

defined above.rTHC error bars were estimated basedstandard deviation of THCtumor and THCnormal.

III. RESULTS

According to the x-ray mammography, ultrasound,MRI, the primary cancer was located at the 12 o’clock ption in the left breast as shown in Fig. 3. The mammograand ultrasound reported adjacent multiple masses, the lbeing 2.132.232.1 cm in size at prechemotherapy tipoints. Prechemotherapy DCE-MRI measured the tumorto be 5.332.232.7 cm. MRI has been shown to be m

ages. Image slices from source to detection plane are presented at 1valsr 4th, 5th, and 7th chemotherapy were arranged from top to bottom.

rom source to detection plane are presented at 1 cm intervals in caud

on imo afte

ices f

th chemotherapy were arranged from top to bottom.
Page 7: Diffuse optical tomography of breast cancer during neoadjuvant

mor-in

tiong

ntra

nalupphe-

ention-re

on-

inft tourceifferged

omorare

con-wasumrbar.pyurceral

theck

edcen-allerrapy

eousrapy

HCvelyhe

asratioctionhesepic-the

arentfore,

intervals infter

ur-

1134 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1134

accurate in depicting the size and overall extent of tuthan mammography or ultrasound.6,62,63 Tumor size estimation by DCE-MRI at different time points is summarizedTable I. The approximate tumor location in prone posiassessed by palpation was,8.5 cm from the nipple alonthe breast contour. In our camera viewscaudal-cranial viewd,this tumor location corresponds to a distant upper ceposition from the nipple as shown in Fig. 3sbd.

The transillumination images in Fig. 4 show high sigattenuation at the surface blood vessel and around thecentral region particularly for data taken after the fourth cmotherapy cycle. This surface blood vessel was first idfied through observation while positioning and was cfirmed with DCE-MRI images. The transillumination pictuis quite sensitive to such surface features.

Three-dimensional DOT images of total hemoglobin ccentration sTHCd, reduced scattering coefficientsms8d at786 nm, and blood oxygen saturationsStO2d are presentedFigs. 7–9, respectively. In each figure, images from leright correspond to 3D DOT image slices taken from soto detector planes. The DOT images corresponding to dent time points within the chemotherapy cycle, are arranfrom top to bottom. Color bar scales are fixed fr5 to 15 cm−1 for ms8 at 786 nm, and from 50% to 100% fStO2. However, colorbar scales for THC are not fixed and

FIG. 9. Three-dimensional reconstructed blood oxygen saturation imcaudal-cranial view, from left to right. DOT images corresponding to a

TABLE I. Tumor size measured with DCE-MRI at different time points ding neoadjuvant chemotherapysAC: adriamycin+cytoxan,T: taxotered. Tu-mor volume was estimated by assuming ellipsoidal tumor shape.

TimeTumor size

scmdTumor volume

scm3d

Prechemotherapy 5.332.232.7 16.5After completion ofAC cycles 3.831.431.8 5.0After completion ofT cycles 3.031.631.3 3.3

Medical Physics, Vol. 32, No. 4, April 2005

l

er

-

-

adjusted at each time point to maximize the THC colortrast between tumor and normal region. The THC scaleadjusted so its maximum value was twice its minimvalue; this preserves the percentile changes in the colo

In the THC imagessFig. 7d after the fourth chemotheracycle, a high THC region is found in slices near the soplanes1–3 cm deep from surfaced and near the upper centpart of the breast. The lesion position corresponds totumor location estimated initially by palpation at 12 o’cloand 8.5 cm away from nipplesas shown in Fig. 3d. After thefifth chemotherapy cycle, the tumor region is still identifiby THC contrast near the source plane and in the uppertral region. However, the contrasted region appears smthan the corresponding region after the fourth chemothecycle. Also, the average THC decreased significantlysi.e.,from 21.4±1.4mM to 9.1±0.5mMd. The THC distributionafter the seventh chemotherapy cycle is more homogenthroughout the slices compared to previous chemothecycles. Within the original tumor margins, the high Tregion shifts toward outside of the tumor, leaving a relatilow THC region occupying most of the tumor extent. Taverage THC increased slightly between Figs. 7sbd and 7scdfrom 9.1±0.5mM to 12.5±0.6mM.

Figure 8 exhibits a similar trend. It shows higherms8 val-ues in the tumor region, and this region with highms8 shrinksover the course of treatment. Thems8 range does not varydramatically as the THC, and gives a higher contrastbetween tumor and normal tissue. Slices near the deteplane are affected by artifacts related to large vessels. Tartifacts bear close resemblance to the transilluminationture in Fig. 4, which is sensitive to large blood vessels onsurface near the detection plane. This effect is also appin the THC images, but to a much lesser degree. Thereslices near the detection planesi.e., within 1 cmd were ex-cluded in calculations of average values of THC, StO2, and

. Image slices from source to detection plane are presented at 1 cm4th, 5th, and 7th chemotherapy were arranged from top to bottom.

ages

ms8.

Page 8: Diffuse optical tomography of breast cancer during neoadjuvant

andgg-ned%

are

timetheium

earlyg-fncinmo-nentate

s atd toure

s oforsy

.ion

l vi-

12at-im-

ribedand

DOTnter

h

theith

ionsthe

RIolo-esh-

n ofasedn-OT,MRIe thetime

op tontraf

rr

OTlndAxiale

1135 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1135

StO2 images in Fig. 9 were relatively homogeneousdo not show contrast in the tumor regionsi.e., not exceedin23spd. Note, however, the overall StO2 value decreased sinificantly after the fifth chemotherapy cycle and remaiconstant thereaftersi.e., from 81.2% ±1.4% to 59.9% ±0.6to 61.1% ±1.0%d.

MRI images projected in sagittal and axial viewsshown in Fig. 10 for the prechemotherapy point, afterACtherapysweek 12d and after taxotere therapysweek 23d. Theintensity ranges were fixed among images at differentpoints. The intensity of the DCE-MRI image is higher intumor due to increased tumor vascularity and gadolincontrast uptake. Before chemotherapy, the tumor is clseen around 12 o’clock to 1 o’clocksupper quadrant in saittal view, near center in axial viewd. After completion ochemotherapy cycles, the size and intensity of the enharegion decreased significantly. Upon completion of chetherapy, MRI demonstrated an amorphous 5–6 cm nohancing soft tissue region with only a scattered puncform of enhancement. The majority of the visible masthis time point was poorly enhancing and was deemerepresent fibrosis. The tumor position and volume measby DCE-MRI is summarized in Table I.

Histology revealed extensive fibrosis with focal areainflammation, but with few vessels. However, viable tumcells, morphologically identical to the original core biopspecimensobtained prior to chemotherapyd were identifiedThey were diffusely scattered throughout the fibrotic regas individual cells and small groups. No macroscopica

FIG. 10. Dynamic-contrast enhanced MRI images of left breast. From tbottom, representative sagittal and axial slices along highest tumor coare shown for each time point: prechemotherapy, after completion oACcycles, and after completion of taxotere cycles.

able tumor mass was identified.

Medical Physics, Vol. 32, No. 4, April 2005

g

-

d

In both MRI and DOT, the tumor was found ino’clock, many centimeters away from the nipple. In antempt to assess the correlation between MRI and DOTages, the simple scaling transformation scheme descearlier was performed with the assumption that nippletumor center are good common reference points. Theaxial image slice corresponding to rescaled tumor cer DOT

tumor swith respect to nipple positiond is shown along witthe corresponding MRI axial image slice in Fig. 11sad. Thetumor center found in MRIsXd and DOT ssquared agreeswell with employment of this transformation. To showextent of the tumor in detail, the axial DOT images wsmallery intervalss0.4 cmd are shown in Fig. 11sbd. HigherTHC contrast in DOT are distributed in three dimensapproximately within the tumor margins defined fromscaled MRI data.

The tumor volume was defined independently for Mand DOT. MRI tumor volume was determined by a radigist, whereas DOT tumor volume was determined by throlding from the distribution of DOT parameters. Figure 12sadshows the decrease of tumor volume with progressiochemotherapy. MRI results show the tumor size decresignificantly after completion ofAC cycles. There is a geeral trend of decreasing tumor volume in MRI and Drespectively. The degree of change is different betweenand DOT, but direct comparison is not possible becausmeasurement time-point mismatch. The coincidingpoint is just after the fourth chemotherapy cyclescompletion

st

FIG. 11. Correlation of tumor position in three-dimensions.sad Tumor centeposition from MRI imagesa X in the axial sliced is transformed by a lineascaling scheme to the DOT coordinatessa square in the corresponding Daxial sliced. The scaled tumor centersa squared lies within a high totahemoglobin concentration region.sbd The scaled tumor boundaries arouthe tumor center is superimposed in three-dimensional DOT images.images shown with smallery intervals s0.4 cmd show the extent of thtumor in detail.

of ACd, where tumor volume measured by DOT is greater

Page 9: Diffuse optical tomography of breast cancer during neoadjuvant

OTna

srap

o-Thearemewith

es o-liz-scuring

ormgio

ofclesuclls.

s ef-

etersabil-car-

ndfectse intrast,from

ap-f pa-riedandfec-post-d inf thed by

posi-spa-ofa-

inceumor

-g. In

thed foringthentra-be

ngthsiveomeim-chro--loodboutula-20%

ob-el tis-

g con-mor

ndim-ave-lud-

ove

1136 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1136

than tumor volume measured by MRI. We note that Dtumor volume can be overestimated due to the ill-posedture of DOT, which introduces blurring.12 In DOT, there waa significant tumor volume decrease in each chemothecycle accompanied byrTHC decrease in Fig. 12sbd. Signifi-cant decreases ofrTHC were observed after the fifth chemtherapy cycle, which is the first taxotere chemotherapy.rTHC after fifth and seventh chemotherapy cyclesequivalent, but accompanied with significant tumor voludecrease, implying tumor neovasculature regressionchemotherapy.

IV. DISCUSSION

We have demonstrated three-dimensional DOT imagtotal hemoglobin concentrationsTHCd, tissue blood oxygenation sStO2d, and scattering coefficient are useful for locaing, quantifying, and tracking breast cancer based on valarity during neoadjuvant chemotherapy. THC and scatteshowed localized contrast between the cancer and the nregion. The THC increase in tumor is expected due to angenesis accompanying tumor growth.64 The enhancementscattering might be expected based on changes in nusize and on the increase in concentration of organellesas mitochondria due to the high metabolism in cancer ce64

FIG. 12. sad Decrease of tumor volume quantified by DOT and MRI,sbdChange ofrTHC in the tumor volume. Significant decrease inrTHC isnoted before and after the 5th chemotherapy.

Generally, StO2 might have been expected to be lower in the

Medical Physics, Vol. 32, No. 4, April 2005

-

y

f

-

al-

arh

tumor region due to high oxygen demand, however thifect was not apparent.

The localized changes of these physiological paramover time clearly demonstrate the dynamic imaging capity of the DOT method. The DOT measurements wereried out at time points just after completion ofACsadriamycin+cytoxand cycles and then just after the first athird taxotere cycles. The antivascular and apoptotic efof taxane65 are consistent with the significant decreastumor vasculature as measured in the tumor THC conthe tumor volume decrease, and the intensity decreaseDCE-MRI. Interestingly, the variations in average THCpear to be qualitatively correlated with measurements otient hematocrit over the same time period, which vafrom 38% to 33% to 36%. The trends observed by DOTMRI are consistent with pathologic findings about the eftiveness of chemotherapy. The carcinoma cells in thechemotherapy stage were finely and diffusely dispersefibrous connective tissue which represented the bulk oresidual mass. This remaining viable tumor was detecteMRI as focal enhancements and by DOT as small buttive THC contrast. Even though there was no significanttial contrast in StO2 for this subject, a significant decreaseaverage StO2 was observed over time; this type of informtion might also be useful for chemotherapy monitoring, stumor response can depend on the oxygenation of the tand surrounding tissues.66,67

Jakubowskiet al.41 reported THC and H2O as major contrast factors, but did not report the behavior of scatterinour case, water concentration was estimated byfrequency-domain measurements and was then fixecontinuous-wave image reconstruction. In addition to fixthe water and lipid concentration, we have tried fitting forwater concentration, but the reconstructed water concetion did not change much from its initial value. This maybecause our measurements used only four waveles,900 nmd, all of which were outside of the water-sensitrange. As for the scattering contrast, it is possible that sportion of this signal may arise from absorption-scatterage cross-talk, since the wavelengths used to discernmophores and scattering were not optimized.52 The scattering artifact near the detection plane resembled the bvessel in transillumination, and thus raises concerns ascattering-absorption cross-talk. However, our 3D simtions with noise in the same geometry find less thancross-talk between THC and scattering, whereas theserved scattering contrastsaveraged over tumor volumdamounts for more than a 60% increase over the normasues. We therefore suspect that the observed scatterintrast may indeed originate, at least partially, from tuphysiology.

To improve quantification of tumor optical contrast avalidate the DOT method in clinical settings, severalprovements are underway. Laser sources at optimal wlengths for separation of chromophore contributions incing water and scattering should further impr

quantification accuracy. Currently, the imperfect tumor re-
Page 10: Diffuse optical tomography of breast cancer during neoadjuvant

ausageandde

s isandandstud

toconof

he-athoi sizndeseallyure-ech-

fordju-ad-ultsg anionaerine tuer thtersdingay a

tru-hanstanrchnd

mai

lo-

her,o-D,with

rac-sys

aryn ch

s, G.aphiction,”

ckles,t can-

. W.emo-trast-

ar-uring

andtived III:

A.crine

in

mp,ffusealua-

for

lsen,im-

an,gra-ast,”

func-ncer

rentreen

. B.dif-

do,man

tt,. A.ional

, P.c im-f 19

, H.hke,Initial

ne-the

C.. Rin-f 50

Phys.

nd A.

1137 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1137

gion correlation between MRI and DOT data arises becthe linear approximations we have made for rescaling imdo not account for the shift of tumor due to compressiondeformation of the breast. Better modeling of the elasticformation of breast in different compression geometrieunder investigation. For better correlation between DOTMRI, it would be ideal to synchronize measurement timeemploy the same compression schemes. In this presentit was not possible to synchronize DOT and MRI duelogistics of two separate research protocols and thestraints of patient availability. The improved coordinationDOT and MRI is now being performed for neoadjuvant cmotherapy patients. The measurement of additional plogic parameters such as microvessel density or nuclewill be invaluable in correlating with tumor vasculature ascattering factors and thus confirming DOT findings; thparameters are not routinely measured in the clinic. Finwith respect to monitoring, more frequent DOT measments should improve the therapeutic value of this tnique.

V. CONCLUSION

We have demonstrated 3D diffuse optical tomographymonitoring physiological tumor responses during neoavant chemotherapy in a single patient with a locallyvanced breast cancer. We have also compared our resdynamic contrast enhanced magnetic resonance imaginpathologic analysis of the same patient. Three-dimensreconstructed total hemoglobin concentration and scattimages successfully localized tumors and quantified thmor volume decrease and the THC contrast decrease ovcourse of chemotherapy. These physiological paramemeasurable by DOT, may help to improve our understanof chemotherapy mechanisms, and hold potential to plrole in assessment of treatment response.

ACKNOWLEDGMENTS

The authors thank J. P. Culver for his work on the insmentation and for helpful discussions. The authors also tL. Sherman for her contribution as clinical research assiand Y. K. Choe for his help with illustrations. This reseawas supported by NIH R01-CA75124-04 aCALGB150007/150012.

adAuthor to whom correspondence should be addressed; [email protected]

1R. W. Carlson and A. M. Favret, “Multidisciplinary management ofcally advanced breast cancer,” Breast J.5s5d, 303–307s1999d.

2B. Fisher, J. Bryant, N. Wolmark, E. Mamounas, A. Brown, E. R. FisD. L. Wickerham, M. Begovic, A. DeCillis, A. Robidoux, R. G. Marglese, A. B. J. Cruz, H. J. L, A. W. Lees, N. V. Dimitrov, and B. H.“Effect of preoperative chemotherapy on the outcome of womenoperable breast cancer,” J. Clin. Oncol.16s8d, 2672–2685s1998d.

3G. Cocconi, B. Di Blasio, G. Alberti, G. Bisagni, E. Botti, and G. Pechia, “Problems in evaluating response of primary breast cancer totemic therapy,” Breast Cancer Res. Treat.4, 309–313s1984d.

4M. C. Segel, D. D. Paulus, and G. N. Hortobagyi, “Advanced primbreast cancer: Assessment at mammography of response to inductio

motherapy,” Radiology169, 49–54s1988d.

Medical Physics, Vol. 32, No. 4, April 2005

e

-

y,

-

-e

,

todlg-e,

kt

l:

-

e-

5S. J. Vinnicombe, A. D. MacVica, R. L. Guy, J. P. Sloan, T. J. PowleKnee, and J. E. Husband, “Primary breast cancer: Mammogrchanges after neoadjuvant chemotherapy, with pathologic correlaRadiology 198, 333–340s1996d.

6L. Esserman, N. Hylton, L. Yassa, J. Barclay, S. Frankel, and E. Si“Utility of magnetic resonance imaging in the management of breascer: Evidence for improved preoperative staging,” J. Clin. Oncol.17s1d,110–119s1999d.

7P. J. Drew, M. J. Kerin, T. Mahapatra, C. Malone, J. R. T. Monson, LTurnbull, and J. N. Fox, “Evaluation of response to neoadjuvant chradiotherapy for locally advanced breast cancer with dynamic conenhanced MRI of the breast,” Eur. J. Surg. Oncol.27, 617–620s2001d.

8M. V. Knopp, G. Brix, and S. L. Zeger, “MR mammography with phmacokinetic mapping for monitoring of breast cancer treatment dneoadjuvant therapy,” Magn. Reson Imaging Clin. N. Am.2s4d, 633–658s1994d.

9N. Tsuboi, Y. Ogawa, T. Inomata, D. Yoshida, S. Yoshida, T. Moriki,M. Kumon, “Changes in the findings of dynamic MRI by preoperaCAF chemotherapy for patients with breast cancer of stage II anPathologic correlation,”Oncol. Rep.6s4d, 727–732s1999d.

10A. Makris, T. J. Powles, S. Kakolyris, M. Dowsett, S. E. Ashley, andL. Harris, “Reduction in angiogenesis after neoadjuvant chemoendotherapy in patients with operable breast carcinoma,” Cancer85s9d, 1996–2000 s1998d.

11A. G. Yodh and D. A. Boas, “Functional imaging with diffusing light,”Biomedical Photonics HandbooksCRC Press, Boca Raton, 2003d, pp.21–1 to 21–45.

12J. P. Culver, R. Choe, M. J. Holboke, L. Zubkov, T. Durduran, A. SleV. Ntziachristos, B. Chance, and A. G. Yodh, “Three-dimensional dioptical tomography in the parallel plane transmission geometry: Evtion of a hybrid frequency domain/continuous wave clinical systembreast imaging,” Med. Phys.30, 235–247s2003d.

13H. Dehghani, B. W. Pogue, S. D. Jiang, B. Brooksby, and K. D. Pau“Three-dimensional optical tomography: Resolution in small-objectaging,” Appl. Opt. 42, 3117–3128s2003d.

14B. W. Pogue, S. P. Poplack, T. O. McBride, W. A. Wells, K. S. OstermU. L. Osterberg, and K. D. Paulsen, “Quantitative hemoglobin tomophy with diffuse near-infrared spectroscopy: Pilot results in the breRadiology 218, 261–266s2001d.

15V. Ntziachristos and B. Chance, “Probing physiology and moleculartion using optical imaging: Applications to breast cancer,” Breast CaRes. Treat.3, 41–46s2001d.

16V. Ntziachristos, A. G. Yodh, M. Schnall, and B. Chance, “ConcurMRI and diffuse optical tomography of breast after indocyanine genhancement,” Proc. Natl. Acad. Sci. U.S.A.97, 2767–2772s2000d.

17X. J. Gu, Q. Z. Zhang, M. Bartlett, L. Schutz, L. L. Fajardo, and HJiang, “Differentiation of cysts from solid tumors in the breast withfuse optical tomography,” Acad. Radiol.11, 53–60s2004d.

18H. Jiang, Y. Xu, N. Ifitimia, J. Eggert, K. Klove, L. Baron, and L. Fajar“Three-dimensional optical tomographic imaging of breast in a husubject,” IEEE Trans. Med. Imaging20, 1334–1340s2001d.

19A. Li, E. L. Miller, M. E. Kilmer, T. J. Brukilacchio, T. Chaves, J. StoQ. Zhang, T. Wu, M. Chorlton, R. H. Moore, D. B. Kopans, and DBoas, “Tomographic optical breast imaging guided by three-dimensmammography,” Appl. Opt.42, 5181–5190s2003d.

20Q. I. Zhu, M. M. Huang, N. G. Chen, K. Zarfos, B. Jagjivan, M. KaneHedge, and S. H. Kurtzman, “Ultrasound-guided optical tomographiaging of malignant and benign breast lesions: Initial clinical results ocases,” Neoplasia5, 379–388s2003d.

21M. A. Franceschini, K. T. Moesta, S. Fantini, G. Gaida, E. GrattonJess, W. W. Mantulin, M. Seeber, P. M. Schlag, and M. Kasc“Frequency-domain techniques enhance optical mammography:clinical results,” Proc. Natl. Acad. Sci. U.S.A.94, 6468–6473s1997d.

22A. Pifferi, P. Taroni, A. Torricelli, F. Messina, R. Cubeddu, and G. Dasini, “Four-wavelength time-resolved optical mammography in680–980-nm range,” Opt. Lett.28, 1138–1140s2003d.

23D. Grosenick, H. Wabnitz, K. T. Moesta, J. Mucke, M. Moller,Stroszczynski, J. Stossel, B. Wassermann, P. M. Schlag, and Hneberg, “Concentration and oxygen saturation of haemoglobin obreast tumours determined by time-domain optical mammography,”Med. Biol. 49s7d, 1165–1181s2004d.

24J. P. Culver, T. Durduran, T. Furuya, C. Cheung, J. H. Greenberg, a

G. Yodh, “Diffuse optical tomography of cerebral blood flow, oxygen-
Page 11: Diffuse optical tomography of breast cancer during neoadjuvant

Flow

ang,ow,moto

ros,sivepiglete in-

J.og-

S.mo-actio

ha-withthick

im-

. G.gen-

ave-Initia

L.sed

, D.is of

is byLett.

. M.rst in

i, R.am-

n of

M.eastam-

ner,pro-ectio

m, Lreas

g, J.y incas

, andmag-

etry,

ray

rans-

g andents.

ton-

igio,broad

B. J.com-.

tions.

. R.n inLett.

. R.on-

eme

. D.og-

ided

dh,ided,

iam-lthy

” Br.

ssue

d G.sts of

and

ell,ast

bbes,cy of

tent,”

rs of. N.Y.

P.reastTreat.

ctor:globins.

1138 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1138

ation, and metabolism in rat during focal ischemia,” J. Cereb. BloodMetab. 23, 911–924s2003d.

25T. Durduran, G. Yu, M. G. Burnett, J. A. Detre, J. H. Greenberg, J. WC. Zhou, and A. G. Yodh, “Diffuse optical measurement of blood flblood oxygenation, and metabolism in a human brain during sensoricortex activation,” Opt. Lett.29s15d, 1766–1768s2004d.

26D. M. Hueber, M. A. Franceschini, H. Y. Ma, Q. Zhang, J. R. BallesteS. Fantini, D. Wallace, V. Ntziachristos, and B. Chance, “Non-invaand quantitative near-infrared haemoglobin spectrometry in thebrain during hypoxic stress, using a frequency-domain multidistancstrument,” Phys. Med. Biol.46, 41–62s2001d.

27D. A. Boas, D. H. Brooks, E. L. Miller, C. A. DiMarzio, M. Kilmer, R.Gaudette, and Q. Zhang, “Imaging the body with diffuse optical tomraphy,” IEEE Signal Process. Mag.18, 57–75s2001d.

28V. Quaresima, M. Ferrari, M. A. Franceschini, M. L. Hoimes, andFantini, “Spatial distribution of vastus lateralis blood flow and oxyheglobin saturation measured at the end of isometric quadriceps contrby multichannel near-infrared spectroscopy,” J. Biomed. Opt.9, 413–420s2004d.

29U. Wolf, M. Wolf, J. H. Choi, L. A. Paunescu, L. P. Safonova, A. Miclos, and E. Gratton, “Mapping of hemodynamics on the human calfnear infrared spectroscopy and the influence of the adipose tissueness,” Adv. Exp. Med. Biol.510, 225–230s2003d.

30B. C. Wilson, M. S. Patterson, and L. Lilge, “Implicit and explicit dosetry in photodynamic therapy: A new paradigm,” Lasers Med. Sci.12s3d,182–199s1997d.

31H. Wang, M. E. Putt, M. J. Emanuele, D. B. Shin, E. Glatstein, AYodh, and T. M. Busch, “Treatment-induced changes in tumor oxyation predict photodynamic therapy outcome,” Cancer Res.64, 7553–7561 s2004d.

32H. Dehghani, B. W. Pogue, S. P. Poplack, and K. D. Paulsen, “Multiwlength three-dimensional near-infrared tomography of the breast:simulation, phantom, and clinical results,” Appl. Opt.42, 135–145s2003d.

33H. B. Jiang, N. V. Iftimia, Y. Xu, J. A. Eggert, L. L. Fajardo, and K.Klove, “Near-infrared optical imaging of the breast with model-bareconstruction,” Acad. Radiol.9, 186–194s2002d.

34R. L. Barbour, H. L. Graber, C. H. Schmitz, F. Tarantini, G. KhouryJ. Naar, T. F. Panetta, T. Lewis, and Y. Pei, “Time-frequency analysfunctional optical mammographic images,” Proc. SPIE4955, 84–92s2003d.

35Q. Zhu, N. Chen, and S. H. Kurtzman, “Imaging tumor angiogenesuse of combined near-infrared diffusive light and ultrasound,” Opt.28, 337–339s2003d.

36D. Grosenick, H. Wabnitz, H. H. Rinneberg, K. T. Moesta, and PSchlag, “Development of a time-domain optical mammograph and fivivo applications,” Appl. Opt.38, 2927–2943s1999d.

37D. Grosenick, K. T. Moesta, H. Wabnitz, J. Mucke, C. StroszczynskMacdonald, P. M. Schlag, and H. Rinneberg, “Time-domain optical mmography: Initial clinical results on detection and characterizatiobreast tumors,” Appl. Opt.42, 3170–3186s2003d.

38A. Torricelli, L. Spinelli, A. Pifferi, P. Taroni, R. Cubeddu, and G.Danesini, “Use of a nonlinear perturbation approach for in vivo brlesion characterization by multi-wavelength time-resolved optical mmography,” Opt. Express11, 853–867s2003d.

39V. E. Pera, E. L. Heffer, H. Siebold, O. Schutz, S. Heywang-KobrunL. Gotz, A. Heinig, and S. Fantini, “Spatial second-derivative imagecessing: An application to optical mammography to enhance the detof breast tumors,” J. Biomed. Opt.8, 517–524s2003d.

40B. J. Tromberg, N. Shah, R. Lanning, A. Cerussi, J. Espinoza, T. PhaSvaasand, and J. Butler, “Non-invasive in vivo characterization of btumors using photon migration spectroscopy,” Neoplasia2, 26–40s2000d.

41D. B. Jakubowski, A. E. Cerussi, F. Bevilacqua, N. Shah, D. HsianButler, and B. J. Tromberg, “Monitoring neoadjuvant chemotherapbreast cancer using quantitative diffuse optical spectroscopy: Astudy,” J. Biomed. Opt.9, 230–238s2004d.

42D. A. Boas, T. J. Gaudette, G. Strangman, X. Cheng, J. J. A. MarotaJ. B. Mandeville, “The accuracy of near infrared spectroscopy and iing during focal changes in cerebral hemodynamics,” Neuroimage13,76–90s2001d.

43

Y. S. Yang, H. L. Liu, X. D. Li, and B. Chance, “Low-cost frequency-

Medical Physics, Vol. 32, No. 4, April 2005

r

n

-

l

n

.t

e

domain photon migration instrument for tissue spectroscopy, oximand imaging,” Opt. Eng.36s5d, 1562–1569s1997d.

44E. K. Insko, T. J. Connick, M. D. Schnall, and S. G. Orel, “Multicoil arfor high resolution imaging of the breast,” Magn. Reson. Med.37s5d,778–784s1997d.

45A. Ishimaru,Wave Propagation and Scattering in Random MediasAca-demic Press, San Diego, 1978d.

46K. Furutsu, “On the diffusion equation derived from the space-time tport equation,” J. Opt. Soc. Am.70, 360–366s1980d.

47R. A. J. Groenhuis, H. A. Ferwerda, and J. J. Ten Bosch, “Scatterinabsorption of turbid materials determined from reflection measuremI. Theory,” Appl. Opt. 22, 2456–2462s1983d.

48T. Durduran, A. G. Yodh, B. Chance, and D. A. Boas, “Does the phodiffusion coefficient depend on absorption?” J. Opt. Soc. Am. A14s12d,3358–3365s1997d.

49J. R. Mourant, T. Fuselier, J. Boyer, T. M. Johnson, and I. J. B“Predictions and measurements of scattering and absorption overwavelength ranges in tissue phantoms,” Appl. Opt.39, 949–957s1997d.

50F. Bevilacqua, A. J. Berger, A. E. Cerussi, D. Jakubowski, andTromberg, “Broadband absorption spectroscopy in turbid media bybined frequency-domain and steady-state methods,” Appl. Opt39,6498–6507s2000d.

51S. R. Arridge and M. Schweiger, “Photon-measurement density funcPart 2: Finite-element-method calculations,” Appl. Opt.34s34d, 8026–8036 s1995d.

52A. Corlu, T. Durduran, R. Choe, M. Schweiger, E. M. C. Hillman, SArridge, and A. G. Yodh, “Uniqueness and wavelength optimizatiocontinuous-wave multispectral diffuse optical tomography,” Opt.28, 2339–2341s2003d.

53A. Corlu, R. Choe, T. Durduran, M. Schweiger, E. M. C. Hillman, SArridge, and A. G. Yodh, “Diffuse optical tomography with spectral cstraints and wavelength optimization,” Appl. Opt.sto be publishedd.

54S. R. Arridge and M. Schweiger, “A gradient-based optimization schfor optical tomography,” Opt. Express2s6d, 213–226s1998d.

55B. W. Pogue, T. O. McBride, J. Prewitt, U. L. Osterberg, and KPaulsen, “Spatially variant regularization improves diffuse optical tomraphy,” Appl. Opt. 38, 2950–2961s1999d.

56L. Kaufman and A. Neumaier, “PET regularization by envelope guconjugate gradients,” IEEE Trans. Med. Imaging15, 385–389s1996d.

57A. Corlu, R. Choe, T. Durduran, K. Lee, S. D. Konecky, and A. G. Yo“Regularization of diffuse optical tomography images by envelope guconjugate gradients,”OSA Technical Digest, ThD4 sOSA, WashingtonDC, 2004d.

58T. Durduran, R. Choe, J. P. Culver, L. Zubkov, M. J. Holboke, J. Gmarco, B. Chance, and A. G. Yodh, “Bulk optical properties of heafemale breast tissue,” Phys. Med. Biol.47, 2847–2861s2002d.

59H. Q. Woodard and D. R. White, “The composition of body tissues,J. Radiol. 59, 1209–1219s1986d.

60D. R. White, H. Q. Woodard, and S. M. Hammond, “Average soft-tiand bone models for use in radiation dosimetry,” Br. J. Radiol.60, 907–913 s1987d.

61N. A. Lee, H. Rusinek, J. C. Weinreb, R. Chandra, R. C. Singer, anM. Newstead, “Fatty and fibroglandular tissue volumes in the breawomen 20–83 years old: Comparison of x-ray mammographycomputer-assisted MR imaging,” Am. J. Radiol.168, 501–506s1997d.

62H. Mumtaz, M. A. Hall-Craggs, T. Davidson, K. Walmsley, W. ThurM. W. Kissin, and I. Taylor, “Staging of symptomatic primary brecancer with MR imaging,” Am. J. Roentgenol.169s2d, 417–424s1997d.

63C. Boetes, R. D. Mus, R. Holland, J. O. Barentsz, S. P. Strijk, T. WoJ. H. Hendriks, and S. H. Ruys, “Breast tumors: Comparative accuraMR imaging relative to mammography and use for demonstrating exRadiology 197, 743–747s1995d.

64S. Thomsen and D. Tatman, “Physiological and pathological factohuman breast disease that can influence optical diagnosis,” AnnAcad. Sci. 838, 171–193s1998d.

65G. Gasparini, E. Biganzoli, E. Bonoldi, A. Morabito, M. Fanelli, andBoracchi, “Angiogenesis sustains tumor dormancy in patients with bcancer treated with adjuvant chemotherapy,” Breast Cancer Res.65, 71–75s2001d.

66P. Vaupel, A. Mayer, S. Briest, and M. Hockel, “Oxygenation gain faA novel parameter characterizing the association between hemolevel and the oxygenation status of breast cancers,” Cancer Re63,

7634–7637s2003d.
Page 12: Diffuse optical tomography of breast cancer during neoadjuvant

eta-ic rel

nce,NIR,

1139 Choe et al. : DOT of breast cancer during neoadjuvant chemotherapy 1139

67P. Vaupel and M. Hockel, “Blood supply, oxygenation status and mbolic micromilieu of breast cancers: Characterization and therapeutevance,” Int. J. Oncol.17s5d, 869–879s2000d.

68

U. Sunar, J. Zhang, J. Du, T. Durduran, C. Zhou, G. Yu, A. Kilger, H.

Medical Physics, Vol. 32, No. 4, April 2005

-Quon, R. Lustig, L. Loevner, S. Nioka, A. G. Yodh, and B. Cha“Clinical responses of head & neck tumors to radiation therapy byspectroscopy,”OSA Technical Digest, FB7 sOSA, Washington, DC

2004d.