Citation
Differentiation of Malignant from Benign Breast Lesions Based on Functional Diffuse Optical Tomography

Material Information

Title:
Differentiation of Malignant from Benign Breast Lesions Based on Functional Diffuse Optical Tomography
Copyright Date:
2008

Subjects

Subjects / Keywords:
Breasts ( jstor )
Cancer ( jstor )
Hemoglobins ( jstor )
Image reconstruction ( jstor )
Images ( jstor )
Lesions ( jstor )
Mammography ( jstor )
Nodules ( jstor )
Tumors ( jstor )
Wavelengths ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
8/31/2006

Downloads

This item has the following downloads:


Full Text












DIFFERENTIATION OF MALIGNANT FROM BENIGN BREAST LESIONS BASED
ON FUNCTIONAL DIFFUSE OPTICAL TOMOGRAPHY















By

LIN CHEN


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Lin Chen

































This document is dedicated to the graduate students of the University of Florida.















ACKNOWLEDGMENTS

I would like to thank Dr. Huabei Jiang, my thesis advisor, for his guidance. I would

like to express appreciation to all the members in the Biomedical Optics Laboratory of

Florida. My special thanksgiving goes to Changqing Li and Xiaoping Liang, for many

discussions and advices in this project. I appreciate Qizhi Zhang and Hongzhi Zhao in

their help for the procedure of clinic experiments.

The Oconee Memorial Hospital and the Greenville Hospital have been a great help

to our clinic experiments in introducing patients and guidance on the lesion position.

I thank my parents for their care, love and support on my study.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES ....................................................... ............ .............. .. vii

L IST O F FIG U R E S .............. ............................ ............. ........... ... ........ viii

ABSTRACT ........ .............. ............. ...... .......... .......... ix

CHAPTER

1 IN TRODU CTION ................................................. ...... .................

2 T H E O R Y .................................................................................................... . 4

In tro d u ctio n ................... ...................4.............................
R reconstruction A lgorithm .................................................. .............................. 5
Characterization of Spectra in Breast Tissue............... ..............................................11

3 M ATERIALS AND M ETHOD S ........................................ ......................... 15

G general A approach ....................................................... ......... .. .............15
Instrum ent and Calibration ............................................................ ............... 17
S y stem S etu p .................................................... ................ 17
System C alibration M ethod........................................... ................ ..... .......... 19
Phantom Experiments for System Calibration ............................................. 20
Results of Calibration Experiments .......................................... ...............20
Clinical Experim ents ......................................... ... .... ........ ......... 24
Fitting C consideration ........................ ................ ... .... ........ ......... 25
A absorption Fitting .................................. .. .. ...... ............ 25
Scattering Fitting .................................. .. .. .. ........ ...............27

4 RESULTS AND DISCU SSION S......................................... .......................... 28

Tumor Result-Images for Selected Examples ........................................ ............28
D iscu ssion s ........................................................ ................. 4 1
H em oglobin Concentration ........................................ ........................... 41
Represented cases.............................................. ........41
4 1 C ases.................................................. 43



v









W ater C ontent................................................... 51
Scattering Spectra ........ .................... ........... .............. 53
C on clu sion .................................................................................................5 5

L IST O F R E FE R E N C E S ........................................ ........... ................ ...........................57

B IO G R A PH IC A L SK E TCH ...................................................................... ..................60
















LIST OF TABLES

Table page

2-1 The Optical Properties of Human Female Breast............ ........................12

3-1 The Extinction Coefficients of All the Wavelengths in the Experiments ...............16

3-2 The Number of Lesions Selected for Experiment and Analysis .................................25
















LIST OF FIGURES


Figure page

2-1 Flow chart of the Newton-type iteration for estimating the distribution of the
optical properties ............. ..................... .................. ......... 10

3-1 Schematic of 2 experimental systems .............. ..... ............................. .......... 18

3-2 Geometry of the phantom configuration .............. ........... ............................... 21

3-3 Comparison between the reconstructed absorption images of original data and of
calibrated data for cases of 2:1 absorption contrast ............................................. 22

3-4 Reconstructed absorption images for heterogeneous phantoms with or without
calibration from different diameters in the case of 1.4:1 absorption contrast..........23

4-1 Corresponding position from report to optical 2D images............................. 29

4-2 The X-ray mammography, reconstructed optical properties and functional
information images of a selected nodule case............... ........ .... ............... 30

4-3 The X-ray mammography, reconstructed optical properties and functional
information images of a selected calcification case................. .... .............. 33

4-4 The X-ray mammography, reconstructed optical properties and functional
inform ation im ages of a selected cyst case ................................... ..................36

4-5 The X-ray mammography, reconstructed optical properties and functional
inform ation im ages of a selected cancer case .................................. ............... 39

4-6 Graph illustrates the quantitative value for the raw de-oxy hemoglobin
concentration and oxy-hemoglobin concentration ..............................................44

4-7 Mean hemoglobin concentrations of reported diseased lesion ..............................45

4-8 Graph illustrates the quantitative value for the maximum hemoglobin
concentration and the corresponding oxygen saturations.............................51

4-9 Graph illustrates the mean value of scattering power with the total hemoglobin
concentration. a and the patient age b in reported diseased lesion......................54















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

DIFFERENTIATION OF MALIGNANT FROM BENIGN BREAST LESIONS BASED
ON FUNCTIONAL DIFFUSE OPTICAL TOMOGRAPHY

By

Lin Chen

August 2006

Chair: Huabei Jiang
Major Department: Biomedical Engineering

In this thesis, optical images based on a clinical study with patients including 11

cancer and 30 benign cases were processed and analyzed. Using multi-spectral diffuse

optical tomography systems coupled with finite element reconstruction algorithms, we

first obtained optical absorption and scattering maps of the breast and then derived tissue

functional images from the recovered absorption and scattering images at multi-

wavelengths. In order to obtain accurate in vivo images, a calibration database was

developed which was based on a series of homogenous phantom measurements with a

range of phantom dimensions. Hemoglobin (both oxy and de-oxy) concentrations and

water content images were obtained from the multi-spectral absorption images, and a Mie

scattering theory approximation was applied to extract scattering amplitude and power.

Functional parameter images of the 41 cases were investigated, and correlation plots of

different function parameters were illustrated and compared among 4 disease categories

including cancer, cyst, nodule and calculation. We found that the majority of the









carcinomas exhibited increased total hemoglobin concentration compared to the healthy

and other benign tissues, and the correlation between total hemoglobin concentration and

oxygen saturation of these diseased tissues showed a clear separation between malignant

and benign lesions, while the separation among the benign lesions is not apparent for the

cases examined.














CHAPTER 1
INTRODUCTION

Optical methods for the detection of breast cancer, especially for the early detection

of cancer, can be traced back to as early as 1929 [1], when it was introduced by Max

Culter. Decades of studies improved the diffuse optical tomography (DOT) as a non-

invasive imaging technique that could provide quantitative absorption and also a

scattering distribution. In the 1980s, Carlsen [2] introduced spectral breast imaging by

restricting the light source of a medical transillumination imager. Because of the

introduction of these technologies, optical imaging of human tissue using near-infrared

(NIR) light provides the possibility of obtaining new types of physiological information

from the tissue in vivo, while the traditional method, using conventional x-ray

mammography techniques could only provide structural information.

The near-infrared light passing through breast tissues is sensitive to several

physiological components such as hemoglobin, water, lipid, melanin, carotene, proteins,

DNA, and so forth. Thus NIR absorption in breast tissue is influenced by hemoglobin

concentration, oxygen saturation, water content, and to a lesser extent by lipid. Therefore,

NIR techniques could be fashioned into an inexpensive and portable alternative solution

for distinguishing malignant (even in an early stage) from benign diseased or health

tissues. NIR can accomplish this by obtaining quantitative hemoglobin concentration,

oxygen saturation, water fraction and other functional information from absorption

distribution of tissue. The goal of the studies described in this thesis is a pilot approach to

evaluate the possibility from substantive patient cases.









Our experimental device was developed as a silicon photodiodes-based DOT system,

and employed a finite element algorithm for the frequent-domain optical data

reconstruction based on a well-known diffusion equation. In Chapter 2, we review the

reconstruction algorithm for absorption and reduced scattering coefficients in detail. Also,

it includes a general review on the spectra characterization of breast tissue.

The experimental system and the preparation for clinical examination are covered in

Chapter 3. In order to assist in obtaining the first quantitative reconstructed data of both

absorption and reduced scattering coefficients, the calibration method was employed. We

built a database based on a series of homogenous phantom with a range of dimensions

(from 60mm to 11cm with an increment of 10mm) to fit with various sizes of clinical

human breasts. Meanwhile, clinical experiments were performed on more than 100

volunteers including those with malignant cancer, benign diseased and healthy breast

conditions. Typical cases with both mammography and ultrasound reports (and biopsy

reports if they exist) were selected, reconstructed and evaluated. We also introduced our

fitting methods on both absorption and scatting spectra based on the simplified models in

Chapter 3.

Results of the clinical experiments are presented in Chapter 4. With a comparison to

the mammography and ultrasound reports, we were able to recognize the tumors in the

corresponding position, and evaluate the functional information. We noticed specific

differences for total hemoglobin, corresponding to physiological and pathological

knowledge, with different kinds of tumors. A series of analyses were carried out for the

purpose of investigating these visible images and quantitative values.









These analyses attempt to provide a basis for the aid of diagnosing malignant cancer

and other breast diseases.














CHAPTER 2
THEORY

Introduction

Yearly mammograms are recommended starting at age 40 and continuing for as long

as a woman is in good health; and clinical breast exam should be taken as part of a

periodic health exam, preferably at least every three years for women in their 20s and 30s,

and every year for women, for the presence of breast cancer which is one in eight women

in the United States. Thus, diffuse optical tomography (DOT), tries to investigate an

alternative method for the early detection of preclinical breast cancer.

Currently, conventional x-ray mammography and palpation are the most common

method for breast cancer detection. However, obvious limitations to conventional x-ray

mammography have been recognized. For example, conventional x-ray mammography is

not suitable for young women in early pre-menopausal stage, by reason of their increased

cellularity and subsequent radiodense tissue structure. That is to say, due to hormone

fluctuations, the pre-menopausal women with preclinical breast cancer are at increased

risk of more rapid tumor growth. In addition, the positive predictive value of

conventional x-ray mammography is quite low in both medical and economic terms, and

as a result, numerous biopsies are required to be performed each year. Moreover, women

with the familial gene for breast cancer (e.g., family history, genetic tendency, past breast

cancer) might experience risk when subjected to the x-irradiation.

As a result, DOT, a non-ionizing, non-invasive near-infrared optical imaging holds

great promise to become an alternative for breast cancer screening, especially for cancer









in early stage. Using a laser light source, this optical method attempts to produce an

image of the inside of the breast, with unique capability for screening high radiodense

breasts usually for premenopausal women. Recent studies have suggested that biomedical

optical imaging of breast tissue has significant advantages for breast cancer detection and

diagnosis, which helps a lot for retaining the corresponding treatment to be keeping pace

with the increased incidence of the breast cancer. Meanwhile, this method has no harm to

human body even for patients with familial gene for breast cancer; and the

instrumentation for optical imaging is much lower in cost than that for x-ray

mammography.

Further, the DOT could obtain quantitative absorption and scattering distributions

from breast tissue, which can not be measured by conventional x-ray mammography or

other radiologic techniques. The spectral dependence of quantitative tissue absorption /u

and reduced scattering /u,' distributions could provide tissue functional information in

breast with the introduction of near-infrared (NIR) migration spectroscopy. Multi-spectral

measurements helps for determination of the concentrations of de-oxy and oxy-

hemoglobin, water, and other components in breast; and the scattering properties of the

tissue could also yield important physiological information, such as the scattering

amplitude and scattering power. These typical values within the breast are believed to

help doctors for better diagnoses on breast diseases.

Reconstruction Algorithm

Our reconstruction algorithm for absorption and reduced scattering coefficients,

previously described in detail [3, 4], is an iterative finite element algorithm based on the

well-known diffusion equation









V. DVO(r, o) (/a -c) 0(r, o) = -S(r, m) (1)

where D is the diffusion coefficient, D(r, o) is the radiance, /u is the absorption

coefficient, c is the wave speed in the medium, S(r, o) is the source term as an eo"t

time variation is assumed. And the diffusion coefficient D can be expressed as


D = (2)
3[u, + (1 g)]

where u, is the scattering coefficient and g is the average cosine of the scattering angle.

And the reduced scattering coefficient is defined as /' '= (1- g)/, .With known /au and

/u distribution, the diffusion equation becomes a standard boundary value problem for

spatially varying radiance subject to appropriate boundary conditions (BC's). There are

three classical boundary conditions for this diffusion equation:

i) specification of the field, D (Dirichlet or Type I);

ii) specification of its flux, -DVDO (Neumann or Type II);

iii) specification of a relationship between field and flux (mixed or Type III)

In our study, we employed Type III BC's in the reconstruction algorithm, that is

DV(O h = a
related to the internal reflection, which can be derived from the Fresnel reflection

coefficient.

For the finite element forward solution, ( and F = -DV(O are expanded as the

sum of coefficients multiplied by a set of locally spatially varying Lagrangian basis

functions

N
(=1 (3a)
J-1









M
F = -Fb (3b)
J-1

where O is the known basis and 0,, F] are the respective radiance and flux at node j.


Similarly, /u, and D are expanded as a collection of unknown parameters multiplied by

a known spatially varying expansion function

K
D = Dk Vk (4a)
k=l

L
U' = 1Y(P (4b)
l1

As a result, the diffusion equation becomes

N K L \

=1 k= I 1=1 C
=1 [ (5)
M
=-(S,)+ F j Ods
j=1

which could be express in the matrix form as

[A]{} ={b} (6)

where the elements of matrix [A] are


a I =- ZDk V I(IZdU -j ) (7a)
\ k=- l=1 ),,1C i


as ( ) indicating integration over the problem domain.


The column vector {~} is composed of the photon density D, at node i.


And {b} is filled with elements that

M
b= -(S)+ aZ j f Ods (7b)
J-1









where f expresses integration over the boundary surface with FJ replaced by ao4 as

Type III BC defined, and M is the number of boundary nodes.

With finite element discretization, the photon density (computed optical data) is

obtained as the solution of the diffusion equation. Then a regularized Newton's method is

exploited here to update the initially guessed optical property distribution iteratively in

order to minimize an object function composed of a weighted sum of the squared

difference between computed and measured optical data at the medium surface. We

assume that the computed and/or measured values of D or F are analytic functions of

D and /u, and D and /u are independent since /, > u. Then 0 and F could be

Taylor expanded as an assumed (D, u,) distribution, which is a perturbation away from

some other distribution, (D,/ ), such that a discrete set of radiance and flux values can

be expressed as


O D =+ + AD + Au, +. (8a)
aD a8/,

SF 8F
F(DA)= F(D,l)+- +AD + A, +... (8b)
aD a0I,

where AD = D-D and A/u = ji -/u,. If the assumed optical property distribution is

close to the true profile, the left-hand side of (8) can be considered as true data (either

imposed or observed), and the relationship truncated to produce

JA = Y" 'c (9a)

where Y0 and 'Y are observed and calculated [based on the estimated (D,/u,)

distribution] data, either 0D or F, depending on boundary conditions for i = 1, 2,...,M









locations and Dk for k = 1,2,...K and u, for 1 = 1,2,...,L; and J is Jacobian matrix

consisting of derivatives of Y with respect to D or /u at each boundary observation

node.

aY, aY, aY, aY, a'Y, ay,1
aDI aD2 OD a/ i a/k2 a/kL
aw, aw, aw, aw, aw, aw,
OT2 OT2 OT2 OT2 OT 2 OT 2
J= aD~ aD2 aDK a ,i au/2 au/L (9b)


8M, 8T, d 8 Mp, 8TM 8aMp
aD, OD2 aD2 aOK aO2 aduL

And A, is the vector that gives the perturbation of /u and D

AD,
AD2


AI I
A{,- A K T 2 WC 2 (9c)




A/pL

In order to realize an invertible system of equations for Az, the Eq. (9a) could be

multiplied by J' on both sides to obtain

JTJAx = JT (Y"- C) (10)

which can be used for updating the optical property distribution.

As the matrix J'J is known to be ill conditioned, techniques should be performed to

regularize or stabilize the decomposition of JJ Thus, a quantity is adding to the

diagonal of J'J in practice, and the problem transformed to









(J'j+AI)AX


jT (T, -c)


(11)


where I is the identity matrix and A may be a scalar or a diagonal matrix.


Figure 2-1 Flow chart of the Newton-type iteration for estimating the distribution of the
optical properties









By adding a contribution to the diagonal terms in Eq. (11), the matrix JWJ is made

more diagonally dominant, which improves its invertibility. Hence, there is no need for

any "exact" solution from Eq. (10), which is already an approximation.

The flow chart in Fig. 2-1 describes the iterative update of /,u and D to approach

the true profile starting from a uniform initial guess.

Characterization of Spectra in Breast Tissue

Physiologically, the breast is a turbid, light scattering medium combined with

different shapes of absorbers, scatterers, fluorophores and anisotropic interfaces [5]. For

biomedical optical imaging, the techniques aiding for cancer diagnosis depend primarily

on detection for the aberrations of reflected, transmitted and emitted light that due to the

physiological characterization or cellular growth of cancer and the host response to the

cancer. Thus, it is important for us to get to the bottom of physiological and pathological

factors of human breast disease that can influence optical diagnosis.

There are two primary optical properties, the absorption coefficient /, and the

reduced scattering coefficient pu,', which determine the propagation of the diffusive light

through the breast. The diffuse optical tomography allows for measuring quantitative

absorption and scattering distribution of tissues at any wavelength of interest, hence it is

possible to use the spectra of which to obtain tissue functional information, such as

quantify typical values of hemoglobin concentration, oxygen saturation, water content,

scattering power, scattering amplitude and so forth within the breast tissue.

With a single integrating sphere technique, a few measurements of optical properties

have been done in vitro on both normal and diseased breast tissues shown in Table 2-1

[6]. The optical coefficients for the tissues with fibrocystic disease, fibroadenomas and










ductal carcinoma have no significant difference with the normal tissues. However, due to

the material of their experiments, which are tissue specimens of human breast, blood

drainage caused by the surgical and pathological dissections of the breasts have greatly

diminished the contributions of hemoglobin when measuring this group of optical

properties.

Table 2-1. The Optical Properties of Human Female Breast

Tissue Type
Optical ,A
Properties (nm) Fibro-adenoma
Glandular (3) Adipose (7) Fibrocystic (8) a Carcinoma (9)
(6)
Absorption 540 0.358 0.156 0.227 0.057 0.164 +0.006 0.438 0.314 0.307 0.099
Coefficient 700 0.047 0.011 0.070 0.008 0.022 0.009 0.052 0.047 0.045 0.012
/P (mm 1) 900 0.062 0.005 0.075 0.008 0.027 0.011 0.072 0.053 0.050 0.015
Reduced 540 2.44 0.58 1.03 0.19 2.17 0.33 1.110.30 1.90 0.51
Scattering 700 1.42+0.30 0.86+0.13 1.34+0.19 0.72+0.17 1.18+0.31
Coefficient
I' (mm1 900 0.99+0.20 0.79+0.11 0.95+0.17 0.53+0.14 0.890.26
Modified from Peters, V. G. et al. 1990. Optical properties of normal and diseased breast tissues in the visible and
near-infrared. Phys. Med. Biol. 35: 1317-1334
The numbers in parentheses give the number of tissue specimens examined for each tissue type. [6]

Recent studies have demonstrated that the near-infrared photon migration is sensitive

to several important tissue biochemical compositions; for instance, especially in the

400-600 nm range, hemoglobin is strong, and could even be seen as some hemoglobin

contamination in some of the spectra. Thus, dual-wavelength was introduced into DOT

researches for quantifying the concentrations of de-oxy hemoglobin (reduced hemoglobin

Hb -R) and oxy-hemoglobin (Hb 0,) in tissue. However, research suggests that the

near-infrared light absorption in breast is produced by more than just hemoglobin. In the

intact living human breast, the most significant light absorbers include hemoglobin,

melanin, water, carotene, proteins and DNA. As reported, the absorption increases toward

shorter wavelengths owing to the protein absorption, and toward longer wavelengths due









to water absorption. That is to say, water and lipid, although they are just weak near-

infrared light absorbers, for their high abundance in the breast, could shadow a significant

influence on absorption relative to hemoglobin in the wavelength of 900-1000nm range,

especially water.

The scattering properties of the tissue also yield important physiologic and

pathological information. However, the distribution of near-infrared light scattering in

tissue is not well understood yet. However, multi-spectral near-infrared light

measurements of the reduced scattering coefficient /u,' have still shown there are

relationship between scattering and wavelength. The scattering decreases with the

increase of wavelength. At shorter wavelengths below 600 nm, the scattering behavior is

likely dominated by scattering from the periodicity and size of refractive index

fluctuations of the collagen fibrils in the size range of 70 nm to hundreds of nm; whereas

at longer wavelength behavior beyond 600 nm, the scattering behavior is increasingly

dominated by scattering from the larger cylinders of collagen fibers (2-3 .m diameter)

composed of collagen fibrils. Similar to other biological tissues, all cellular and extra-

cellular components within the breast tissue contribute to light scattering; thus, besides

the collagen fibers at the micro and macro scale, other tissue components also contribute

to the overall scattering, but the collagen fibers probably play the main role in scattering

behaviors. Thus, information about the types of the scattering centers within certain

lesion of breast tissue could be provided by spectral measurements to the scattering

coefficient.

Recent advances in our experimental studies have provided reconstruction datum and

images with separate /u and ', and phantom experiments have presented solid






14


evidence that the reconstruction value has no clear differences with the practical designed

value. In this work, the principal near-infrared light absorbers within the breast tissue are

assumed to be de-oxy hemoglobin (Hb -R), oxy-hemoglobin (Hb 0 ) and water

(HO). For the scattering coefficient /,', we made use of some apriori information,

that is, the introduction of the simple power-law which is well known in the near-infrared

light theory. As a result, the scattering amplitude and the scattering power were reported

in each experiment.














CHAPTER 3
MATERIALS AND METHODS

General Approach

The chromophore concentration on the absorbance of near-infrared light depends on

'ua = [c]l, where E is the molar extinction coefficient (Mlcmn ), [c] is the

concentration of chromophore c (ML 1), and / is the photon path length (cm). the path

length / is increased by scattering and is not known apriori. As reported, the absorption

coefficients translate into tissue chromophore concentrations based on the equation

u, =2.303E [c] (12)

where the factor of 2.303 originates from the base conversion between the logarithm for

absorbance and the natural logarithm for /u [7, 8, 9].

We assume that the chromophores contributing to /u, in the human breast tissues are

principally de-oxy hemoglobin (Hb R), oxy-hemoglobin (Hb 0 ) and water (HO)

(Cope 1991; Sevick et al. 1991) Thus, the concentrations of components in the tissue we

need to determine for study includes the concentration of de-oxy hemoglobin [Hb] (in

units of /jM ), oxy-hemoglobin [HbO, ] (in units of /jM ) and water [HO] (in units of

percentage) in the tissue. The respective extinction coefficient E" for a given
[cYom ]

chromophore at wavelength A, could be obtained from literature values. In the

experiments, as the absorption coefficients t/ are measured from 3 to 10 wavelengths

optionally [10, 11], we could use at least three equations to determine the three unknown









quantities of the functional concentrations of [Hb], [HbO,], and [HO]. The extinction

coefficients of all the wavelengths are listed in the following table [12]:

Table 3-1. The Extinction Coefficients of All the Wavelengths in the Experiments
A (nm) Hb2 ( 1 H ( 1 Optical Absorption of Water
A (nm) HbO (M 'cmn ') Hb(Ml'cmn1) ) (
Compendium (cm1 )
673 287.8 2668.24 0.00478
690 276 2051.96 0.005535
733 403.2 1102.2 0.023144
775 683.2 1188.28 0.027565
808 856 723.52 0.021773
840 1022 692.36 0.039494
915 1219.8 778.22 0.076
922 1225.6 777.04 0.09268
965 1175.6 484.34 0.470099

785 735.4 977.04 0.025933
808 856 723.52 0.021773
830 974 693.04 0.0320205

The scattering spectrum of tissue yields information on the nature of the scattering

particles. In general, u,'(A) is the sum of contributions from the various tissue scatterers.

Unfortunately, the detailed information about these individual scatterers was not well

understood yet. Since it has been observed that the reduced scattering coefficient has a

general trend to decrease as the wavelength increased, with some apriori information,

the simple power-law dependence were employed to fit the near-infrared light scattering

in tissue [7, 8]:

u,'= AASP (13)

where A is the arbitrary model parameters for amplitude (a constant), A is he

wavelength (in nanometers), and SP, the magnitude of the exponent, is the scatter power.

As known, the value of scattering power SP increases significantly as the decreasing of

the scattering center size, combined with the optical wavelength. In the case of Rayleigh









scattering (d << A, where d is the scattering center size, and A is the responsible

wavelength), SP = 4 is well established. As scattering objects become larger size, the

scatter power SP decreases to approximately 1 for larger Mie-like scatters (d ); and

the value of SP decreases to zero when d > A [8, 13].

Instrument and Calibration

System Setup

There are two imaging systems used for the clinic experiments are automated multi-

channel frequency-domain systems, both of which employ multiple diode lasers that

provide visible and near-infrared light. The DOT system setups are schematically shown

in the Following Figure. One could provide near-infrared laser at 3 different wavelengths

(785, 808, and 830nm), and the other at 10 different wavelengths (638, 673, 690, 733,

775, 808, 840, 915, 922 and 965nm). The ring of the instruments holds different layers of

fibre bundles, and each of those layers has 16 source and 16 detection fibre bundles

alternately distributed by turns. The radio-frequency intensity-modulated near-infrared

beams are transmitted to the optical switch, and sequentially be passed to the source

probes that gently touched the surface of the experiment materials or clinic human breasts.

The diffused light collected by the detection fibre bundles is then sensed by the detection

Units, which convert the light intensity into voltage signals, which are collected by the

computer through the data acquisition board. We use the measured data for absorption

and scattering images through our reconstruction algorithm.






18



Source fibre
bundles
S10 O Detection
10 Lasers
10 LasersOptical Switch fibre bundles Detection

Units



Laser
Current


-Data

CCDs DC motor Acquisition
SBoards

Computer


a.


Reference signal
Laser diode Moving stage PMT 1
cotrolls Fiber-optic bundles
ctr ietl Movin stage Neutral-c nsity filter

Laser PMT 2
head



Sample grf
rf generator rf generator f
amplifier
Phase control


Nanostep
controller Computer



b

Figure 3-1. Schematic of 2 experimental systems. a. Schematic of the experimental
system with 10 wavelengths b. Schematic of the experimental system with 3
wavelengths









System Calibration Method

It has been demonstrated that the system calibration could provide helpful aid to the

quantitative reconstructions of both absorption and scattering coefficients in turbid media

[14]. In clinical experiments, the diameters of breasts various from person to person, a

data base for calibration has better to be established. For 2D imaging experiments, the

calibration procedure could be described as following approaches:

(1) Make a group of homogeneous phantoms with different diameters of interest for
data base.

(2) Perform experiments individually with the homogeneous phantoms and collect the
measured data {D } respectively for each phantom.

(3) For each set of measured dataD find the respective initial values of absorption
coefficient /, reduced scattering coefficient /,' and the boundary conditions
coefficient a.

(4) Generate a 2D finite element mesh with the same diameter for each phantom. Using
a unit source intensity for the 16 illuminated positions, simulate the 2D photon
propagation with the initial values of the optical properties and the boundary
conditions coefficient a obtained in the former approach, then a new set of data
Dy could be generated.

(5) Obtain a factor matrix f, using the following equation

f, =D/D* i, j=1,2,---,16 (14)

The factor matrix fy could be added into the calibration data base. For experimental

data of whether heterogeneous phantom or clinical breast, multiply f/ from the

homogeneous phantom with the nearest diameter by the experiment data set (E }, a final


calibrated data set E) } could be obtained, and used for further image reconstruction


E=f i, j=1,2,,16 (15)









Phantom Experiments for System Calibration

Phantom is used as an object to make in imitation of biological tissues in terms of

absorption and scattering coefficients. In our study, the phantom materials employ

composition of Intralipid as the scatterer and India Ink as the absorber, as the Intralipid is

an aqueous suspension consist of glycerin, lecithin, soybean and water. A boiled agar

powder solution in a concentration of 2% is chosen as the hardener to solidify the

aqueous mixture of Intralipid and India Ink, taking advantage of its non-absorption and

low-turbidity.

Considering the various size of human breast, homogeneous phantoms in different

diameters (60mm, 70mm, 90mm, 100mm, and 110mm) were prepared for the imaging

experiments under the calibration method. The optical properties of phantoms in

calibration experiments are the same: /u = 0.005 mm1, and u '= 1.0 mm1.

Results of Calibration Experiments

To examine the results of calibration data base, heterogeneous phantoms that have

similar diameters to the homogeneous phantoms were employed during the examination

for calibration experiments. The optical properties for the background of the

heterogeneous phantoms are still the same with the homogeneous phantoms:

u, = 0.005 mm1, and /,'= 1.0 mm 1; and a single target was embedded in each

homogenous background phantom, with position departs from the center. Thus, one

14mm diameter hole was cylindrical drilled in each homogenous background phantom

for the inclusions of the target. As the related researches have shown that there is obvious

difficulty to get quantitative optical images under conditions of very low absorption

contrast, and this instance do not happen under conditions with just low scattering









contrast or both low absorption and low scattering contrast. Considering this, the optical

coefficients of the off-center targets for the heterogeneous phantoms in the calibration

examine experiments were set as low absorption contrast only, that is /, = 0.010 mm 1,

and /,'= 1.0 mm 1. The following figure depicts the geometrical configurations for the

test cases of phantom diameter under study.





R, =d /2
|R = 7mm










Figure 3-2. Geometry of the phantom configuration

Figure 3-3 present the reconstructed absorption images, both without calibration and

with calibration, from phantom experiments in diameters 60mm and 70mm, respectively,

take the imaging data from the same iteration of reconstructed data with the same filter

times, and from the same wavelength (take A = 922nm, in which wavelength it is almost

the most difficult than most other wavelengths for quantitative absorption images).

From the Fig. 3-3, the results accords very good with our former conclusion. In other

words, when the heterogeneous phantoms has the same diameter as the homogenous

phantom that used for obtaining calibration data, the calibration method could improve

the quantitative optical images in size, shape and value, even under the conditions of very

low absorption contrast.



















a b




ID





c d

Figure 3-3 Comparison between the reconstructed absorption images of original data (left
column) and of calibrated data (right column) for cases of 2:1 absorption
contrast. a and c are the reconstructed absorption image from the original data
in diameter 60mm (first row) and 70mm (second row) respectively; b and d
are the corresponding absorption image from the data with calibration
compare respectively to image a and c.

In clinical experiments, the sizes of human breasts were not the same as the

diameters we set for calibration phantom experiment. Thus, another group of experiments

were performed. As the space between different homogenous phantoms is 10mm, we

could use the calibration data from phantoms whose diameters border upon the diameter

of the heterogeneous phantoms. Take the data from heterogeneous phantom (Absorption

only, /u = 0.007 mm and / '= 1.0 mm-)) with diameter of d = 100 mm for example,

we use calibration data from homogenous phantoms with diameters of 90mm, 100mm,










and 110mm respectively. Figure 3.4 present the corresponding reconstructed absorption

images for A = 922nm .












olo a 3
a b


5k,










c d

Figure 3-4 Reconstructed absorption images for heterogeneous phantoms with or without
calibration from different diameters in the case of 1.4:1 absorption contrast

In the Figure 3-4 (where the Figure 3-4 a is the reconstructed absorption image for

original data without any calibration, and Figure 3-4 b, 3-4 c, and 3-4 d are the

reconstructed absorption images for data with calibration of 100mm, 90mm, and 110mm

respectively), obviously the results are much better for calibrated data, even with

calibration data from that of homogenous phantoms with different diameters. Comparing

to the reconstructed image of 100mm-calibrated data, there are slight shifts for the target

position for 90mm-calibrated and 1 10mm-calibrated data, and the sizes of target changes

also. The maximum values of reconstructed data from 90mm-calibrated data and 110mm-









calibrated data have errors around 5% compared to the maximum values of reconstructed

data from 100mm-calibrated data, and the minimum values do not show visible errors

among the three images. Since 10mm is the largest distance in the calibration data base,

we can conclude that the errors of the values are acceptable in the experiments, and the

calibration data base we obtained is suitable for most clinical case.

Clinical Experiments

Over the past years, more than 100 volunteers, healthy or having pathological lesions

in their breasts, ranging from 30 years old to 80, have been enrolled in this study of

optical clinical human female breast experiments. All these cases could be divided into

two sets: patients in the first set have strong evidence of abnormality, and the patients

grouped in the other set have mammograms with unclear significance.

After each patient was informed about this experiment, the patients would undergo

the procedure lasting about half an hour. The ring holding the optical fibre bundles was

gently attached to examine the breast (only one breast) under the lesion position based on

guidance or suggestion from professional doctors, without any discomfort or even

significant pressure on the breast. When started, the breast would be illuminated by laser

beams from a series of source probes, and at the same time, the detectors from multiple

positions around the breast would collect the diffused light transmitted from the breast

tissue. After these datum collections, our reconstruction algorithm could be applied for

further study.

In this paper we present representative cases from the selected subset of abnormal

cases from 2 groups of patients (from 3 wavelengths to 10 wavelengths selectively,

examined by two different system respectively), generate breast images with comparison

to the mammograms, obtain absorption and reduced scattering coefficients with









calibration methods, and finally gain and analyze the optical properties and further

functional information revealed by chromophore concentration from absorption

coefficient /u and scattering properties from scattering coefficient /,'

The diseased human female breast could be divided into two categories as benign

and cancer. The benign breast tumor includes Fibroadenomas, Fibrocustic Disease (Cyst)

and miscellaneous lesions such as lipomas, blunt trauma, mastitis tissue and even

ruptured or leaky silicone implants. These benign lesions are distinct pathologic entities,

but for detection and diagnosis, they can be and usually are intermixed within the same

breast. Thus, we selected clinical cases listed in the following table.

Table 3-2 The Number of Lesions Selected for Experiment and Analysis
Cancer Lesions Benign Lesions
Cyst Calcification Nodules
OMH 4 4 3 14
Greenville 7 2 3 4

The data obtained in the OMH (Oconee Memorial Hospital) were collected for 10, 7

or 5 wavelengths, considering the size of the breast; and the data obtained in the hospital

of Greenville were collected for 2 or 3 wavelengths.

Fitting Consideration

Absorption Fitting

With the measured absorption coefficients /u and relative absorption spectrum, a

weighted least-squares problem is put forward for recovering the concentrations of

absorbers in breast tissues. The dependence equations









-(U ) (E, E()m(.) c,

(17)


_a(_n) l() ,m(An) _Cm

could be written in a more general form

{/ I = [E]{c} (18)

As {( } is a vector containing the measured /u values for N wavelengths, and {c}

is the vector containing the concentrations of M different chromophores interested in the

study. And [E] is the relative extinction coefficients N xM matrix, where the traditional

literature molar extinction coefficients should be convert into extinction coefficients by

multiply a 2.303). Hence, a general solution for this matrix problem could be expressed

as

c = ([E]T [E])l [E]T ({ (19)

To minimize calculation errors, normalizing scheme were employed to balance the

variation of elements in the extinction coefficients matrix [E] by normalize columns of

extinction coefficients for each chromophore with their respective maximum. Assuming

M, i = 1,2, m are the maximum value for extinction coefficient { (A )}, j= 1,2, *, n,

the normalized extinction coefficient matrix convert into

1 1
-- (, ).... (,
M1 M
E= : "'. : (20)
1 1
M, M.









Then I =E E and finally c, = 1 (21)


For these cases we report four hemoglobin parameters: [Hb], [HbO2 ], total

hemoglobin concentration HbT, and the hemoglobin saturation SO2, where

[HbT]= [Hb]+[HbO2] (22a)


S [bo2 (22b)
2 Hb]+ [HbO2

Scattering Fitting

The wavelength-dependent tissue reduced-scattering coefficient is assumed to take

on this simplified Mie-scattering form /u,(2) = AA s as described in the beginning of

this chapter. A and SP are related to the size, the index of refraction, and the

concentration of scatterers in the tissue as well as the index of refraction of the

surrounding medium. This Mie-scattering form is judged as a robust nonlinear form, and

is transformed into a linear form

In/u'(A) = In A SP x In (23)

For every two wavelengths A, and A, we could obtain a relative In A and SP ,

hence, the final In A and SP were gained as the average value of a group of relative

In A and SP,

I1 n- n n(n -1)
A= exp~f InA. ,as n' = 1 (24a)
n _1 _1+1 2


SP = S(24b)
n Z=1 J=i+1

where n is the number of measured wavelengths.















CHAPTER 4
RESULTS AND DISCUSSIONS

Tumor Result-Images for Selected Examples

All these volunteers that were selected for our studies including benign and cancer

patients, have 41 abnormal lesions, which could be divided into 4 primary kinds of breast

tumors: 18 benign nodules, 6 calcifications, 6 cysts, and 11 cancers. All lesions were

examined by professional physician and showed clear signs on mammography or have

distinct evidences on other examination reports. Figure 4-1 shows a rough relationship

between the clinic breast front views (which was used for the breast examination reports)

and our 2D optical images. In the following section, we will discuss some selected cases

(in Fig 4.2 4.5, we present one selected patient for each disease category), in order to

gain a general view before further quantitative analysis.

Nodule

This patient is a 43-year old white female with two smooth lobulated solid nodules

presented in the medial aspect of the left breast at the 9:30 o'clock. The more anterior

nodule measures 5.8mm x 4.8mm x 8.3mm, and the mid left breast nodule measures

6.9mm x 3.6mm x 9.1mm. These two nodules lasted for at least six months before our

optical imaging, and were recommended to follow up examinations in another six months

to access stability. The optical image experiments were done by the 10-wavelengh-

system, and 64 sources and 64 detectors were distributed uniformly for four planes along

the surface of patient's left breast at the lesion position (the respective radii of the four










layers that attach the breast are: r, = 46.42 mm, r2 = 44.06 mm, r3 = 42.18 mm, and

r4 = 40.30 mm; 16 sources and 16 detectors at each plane). As layer 1 (r, = 46.42 mm)

showed the most similar results compare to the mammography and ultrasounds report, so

we took the plane 1 for further studies. And the x-ray mammography (Figure 4-2a) and

our optical imaging results (Figure 4-2b 4-2e) are presented as figures 4-2.


12:00 12:00
Right I Left


9:00 3:00 9:00
\ ITl\ ^


6:00


3:00


6:00


3:00


6:00


9:00


12:00


Value


Figure 4-1 Corresponding position from report to optical 2D images. a The Front View
on Breast Examination Report. b The 2D Image View for Experiment Results



















L MLO


638nm 673nm 690nm




808nm 840nm 915nm


733nm 775nm


@1


922nm


[HbO,] [Hb] Water (%) HbT StO2 (%)

c
Figure 4-2 The X-ray mammography, reconstructed optical properties and functional
information images of a selected nodule case. a. The X-ray Mammography
for this patient. b. The Respective Absorption Coefficients Images from
Wavelength 638nm to 922nm. c The Resolved Chromophore Concentration
Images from Absorption Coefficients. d Respective Reduced Scattering
Coefficients Images from Wavelength 638nm to 922nm. e Scattering
Amplitude and Scattering Power Images


L LUL












638nm 673nm



0nm 0nm
808nm 840nm


690nm 733nm


@II@


915nm


922nm


Scatt Ampl. Scatt Power
e

Figure 4-2 Continued.

Hence, the two disease lesions near the 9:30 position were distinct at the

corresponding position of optical images for both absorption and reduced scattering

coefficients from the shortest wavelength 638nm to 840nm. Unfortunately, for the

longest three wavelengths, we could not obtain reconstructed images with lesions legible

for further study, due to the weaker signals. Therefore, we use those qualified

reconstructed datum for resolving chromophore concentrations, scattering amplitude, and

scattering power. Those two tumors were distinct around 9:00 o'clock position in images

for de-oxy hemoglobin concentration, oxy-hemoglobin concentration, and water content:

all the functional information of the two tumors shows an increase compared to the


775nm









surrounding tissue. Total hemoglobin concentration HbT, and the hemoglobin oxygen

saturation SO, were also gained. Only the HbT shows increases at the tumors'

positions, but at the corresponding position, no tumor emerged in shape. For resolved

scattering amplitude and power images, tumors are visible, showing increases for

scattering amplitude and decreases for scattering power at their respective positions.

Calcification

This patient was a 65-year-old white female with the presence of coarse

calcifications which were clustered together in the upper outer right breast at 10 o'clock

position. No underlying soft tissue component is apparent on the spot compression views.

In addition to the ultrasound, there is a fairly well circumscribed slightly lobular

hypoechoic focus associated with the calcifications, which measures approximately 5mm

x 8mm in size and could represent fat necrosis. The optical image experiments were also

done by the 10-wavelengh-system, with the radii for three planes along the surface of

patient's right breast at the lesion position are: r, = 36.43 mm, r2 = 34.07 mm, and

r = 32.19 mm (the probes for the 4th plane was not touched fully on the breast because of

its own shape). Layer 1 (r, = 36.43 mm ) showed the best results compared to the

mammography and ultrasounds report. And the x-ray mammography (Figure 4-3a) and

our optical imaging results (Figure 4-3b 4-3e) are presented as Figures 4-3.




















R ML


638nm 673nm 690nm 733nm 775nm



808nm 840nm 915nm 922nm
b



I Ope1 @1
[HbO,] [Hb] Water (%) HhT S02 (o%)
c
Figure 4-3 The X-ray mammography, reconstructed optical properties and functional
information images of a selected calcification case. a The X-ray
Mammography for this patient. b The Respective Absorption Coefficients
Images from Wavelength 638nm to 922nm. c The Resolved Chromophore
Concentration Images from Absorption Coefficients. d Respective Reduced
Scattering Coefficients Images from Wavelength 638nm to 922nm. e
Scattering Amplitude and Scattering Power Images


R CC

























k I




Scatt Ampl. Scatt Power

e

Figure 4-3. Continued.

Still, the absorption and reduced scattering images were vivid, and the lesion position

is in accordance with the examination reports, except the images for the longest 3

wavelengths. Clearly, the calcification lesion at 10 o'clock position is apparent on almost

all the resolved functional coefficients images except for the concentration of SO2.

Remarkably, the nodule which could be fat necrosis was not apparent on mammograms

and only shadowed on the ultrasound examination, but did appear clearly on our optical

images for most wave lengths in terms of both absorption and reduced scattering

coefficients. Also, the resolved images for chromophore concentration (except SO2),

scattering amplitude and scattering power, shadowed this nodule near the corresponding


638nm 673nm 690nm 733nm 775nm




808nm 840nm 915nm 922nm









location. Both the calcification and nodule lesion present in silhouette, with increases in

chromophore concentrations and scattering amplitude, and decreases in scattering power.

Cyst

This patient was a 43-year-old female during our optical breast experiments. The

reports clinical recorded that there was a persistent 2.3cm diameter smooth round nodular

density, which was demonstrated to be a benign cyst. The physician indicated the cyst to

be located at the 12 o'clock position. A moderate to large amount of dense residual

fibroglandular tissue within the left breast was also present on the mammography. A

small asymmetric oval area of density could be seen in the inferior half of the left breast

on the MLO view; however, ultrasound examination suggested no dominant masses in

the region of the palpable abnormality. To explain this, her ultrasound report also pointed

out that the density of the breast was such that it might decrease sensitivity of

mammography for detection of malignancy. The 10-wavelengh-system again were

employed for the corresponding optical imaging experiments, performing four-plane

measurements for the lesion position, and the respective radii are: r, = 49.04 mm,

r2 = 46.68 mm, r, = 44.80 mm, and r4 = 42.92 mm, and the fourth layer were selected

out for further researches.

We could see visible tumor at the 12:00 o'clock position in reconstructed absorption

and reduced scattering coefficient images for the middle region of wavelengths. For

resolved functional coefficients, there are corresponding increases for this tumor at the

respective position in images for de-oxy hemoglobin concentration [Hb], oxy-

hemoglobin concentration [HbO,], water content, total hemoglobin concentration HbT

and scattering power; however, no significant decreases or increases were present at the









diseased lesion in the image of scattering amplitude, and neither in the oxygen saturation

SO2 image. Interestingly, those tissue density that recorded in her mammography reports,

do increase the value contrast to the cyst value in the resolved chromophore

concentration images, comparing to the images of absorption and reduced scattering

coefficients, which made them more similar to the normal background in the

chromophore concentration images.


Figure 4-4 The X-ray mammography, reconstructed optical properties and functional
information images of a selected cyst case. a The X-ray Mammography for
this patient b The Respective Absorption Coefficients Images from
Wavelength 673nm to 922nm c The Resolved Chromophore Concentration
Images from Absorption Coefficients c Respective Reduced Scattering
Coefficients Images from Wavelength 673nm to 922nm e Scattering
Amplitude and Scattering Power Images


L CC


L MLO



















[Hb120 [Hb] Water (%) HbT S2, Q/)
c



@11 01 @1
673nm 690nm 733nm 775nm



808nm 840nm 915nm 922nm
d

Figure 4-4 Continued


















Scatt Ampl. Scatt Power

e


Figure 4-4 Continued

Cancer

This patient was a 62-year-old white female with two cancer masses diagnosed with

the needle biopsy report. After our optical image experiment, the patient had surgery on

her breast and removing the lesions. In her screening mammography reports, these two

cancer lesions were reported as two increasing microcalcifications, one in the

inferomedial quadrant near the dominant mass and the other in the central aspect of the

mid breast. A few other small clusters of microcalcifications are seen laterally in the right

breast, and more likely to be related to a benign process such as sclerosing adenosis or

cystic hyperplasia. In addition, there was a small to moderate amount of dense residual

fibro-glandular tissue in the breast.

This experiment was performed on the the 10-wavelengh-system, with only two

layers of source-detector probes touch patient's breast surface at the lesion position and

the respective radii are: r, = 65.60 mm, and r2 = 63.24 mm, and the experiment datum

collected in the second plane showed better results.

Here, the two cancer tumors were present at the correct positions according to the

surgery reports for reconstructed absorption images, but were not as clear in the









reconstructed reduced scattering images. After analysis, the two cancer tumors were more

clear-cut in silhouette as in de-oxy hemoglobin concentration [Hb], oxy-hemoglobin


concentration [HbO ], water content, total hemoglobin concentration HbT, scattering

amplitude and scattering power. In this case, corresponding increases were present at

images for [Hb], [HbO2], HbT, and scattering power; and decreases were present at

images for water content and scattering amplitude. In addition, decreases in oxygen

saturation S,02 could also be observed at the cancer tumors lesions, although the margins

were not so fairly well circumscribed. Again, the uncertain tissues that have low contrasts

compared to the cancer tumors' values in the absorption and reduced scattering

coefficient images, do increase their value contrast in the resolved chromophore

concentration images.







673nm 690nm 733nm 775nm





808nm 840nm 915nm 922nm

a


Figure 4-5 The X-ray mammography, reconstructed optical properties and functional
information images of a selected cancer case. a The Respective Absorption
Coefficients Images from Wavelength 638nm to 922nm b The Resolved
Chromophore Concentration Images from Absorption Coefficients c
Respective Reduced Scattering Coefficients Images from Wavelength 638nm
to 922nm d Scattering Amplitude and Scattering Power Images











[HbO ] [Hb] Water (%) HbT StO2 (%)
b




I I 5 I sI
673nm 690nm 733nm 775nm



808nm 840nm 915nm 922nm


Figure 4-5 Continued


Scatt Ampl. Scatt Power









Discussions

In this section, we will discuss the functional properties we obtained. Different

function parameters were illustrated in correlation plots and compared, and the major part

of our discussion is on hemoglobin concentration. Followed a general view on the

selected cases we presented and the correlation plots of [Hb] [HbO2], total hemoglobin

concentration HbT were investigated with age and oxygen saturation StO2 in order to

evaluate the feasibility of separating different breast diseases. Water content and

scattering power were also briefly analyzed with problems presented.

Hemoglobin Concentration

Represented cases

Before quantitative analysis, we obtained some general results that emerged from

these selected cases we present above for different breast disease categories. All subjects

with confirmed pathologic abnormalities produced respective localized increases in both

de-oxy hemoglobin concentration [Hb] and oxy-hemoglobin concentration [HbO2]:

(a) Nodule case (the two nodules combined at 9:30 o'clock): the resolved images

recorded them as: Nodule 1 has a mean de-oxy hemoglobin concentration [Hb] for about

12/iM (peaked at approximate 16/iM ), and the mean oxy-hemoglobin concentration

[HbO2 ] for about 45/iM (peaked at approximate 59/iM ); Nodule 2 has a mean de-oxy

hemoglobin concentration [Hb] for about 10/iM (peaked at approximate 14/iM ), and

the mean oxy-hemoglobin concentration [HbO2] for about 30/iM (peaked at

approximate 40/iM ). Thus, the relative mean total hemoglobin HbT is about 55/MU[

(peaked at approximate 70/jM ) and 43/IM (peaked at approximate 50/iM)









respectively. No decreases for the hemoglobin oxygen saturation SO, were observed at

the corresponding location for these two nodules.

The additional nodule lesion in the calcification case had a mean value about 10/IM

for [Hb] (peaked at approximate 13/iM) and 47/Mi for [HbO, ](peaked at approximate

60/iM); hence its relative average HbT is about 56/IM ( peaked at approximate

71/iM), with no clear decreases in SO, were observed at the corresponding location.

(b) Calcification Case (calcification located at 10 o'clock position): This lesion was

imaged with a mean value 9/iM in [Hb] (peaked at approximate 1 l/M ) and 45/uM

for [HbO, ] (peaked at approximate 58/iM ), which suggested the mean total hemoglobin

concentration HbT was 54/Mi (peaked at approximate 68/iM ). Again, no decreases in

the hemoglobin oxygen saturation SO, were observed.

(c) Fibrocustic Disease (Cyst) Case: (a simple cyst located at approximate 12:00

o'clock position): The resolved images indicate a mean de-oxy hemoglobin concentration

[Hb] of 13/M[ (peaked at approximate l8/IM) and a mean oxy-hemoglobin

concentration [HbO,] of 19/iM (peaked at approximate 40/iM ). Therefore, the mean

total hemoglobin concentration HbT was close to 3 l/iM (peaked at approximate

49/iM ), and no decreases for corresponding location of SO, similar to the other cases.

(d) Cancer Cases (Two cancer tumors): The mean values of [Hb] for tumor 1 was

20/jM (peaked at approximate 27/iM ), and the relative [HbO, ] was 5 l1/M (peaked at

approximate 62/iM ); similarly the mean values of [Hb] for tumor 2 was 15/iM (peaked

at approximate 17/iM ), and the relative [HbO ] was 45/iM (peaked at approximate









5 1uM ). Their corresponding total hemoglobin concentrations HbT averaged at

approximate 70/iM and 60/iM, and peaked at about 87/iM and 68/iM respectively.

In contrast to the benign cases, legible decreases could be recognized in the

corresponding locations of the hemoglobin oxygen saturation SO, image, although it

was not clear-cut.

At this stage of our clinical investigation, it is obvious to note that the abnormal

tissues have higher hemoglobin concentration comparing to their surrounding normal

background; and these increases should correlate with the blood vessel density. The two

cancer tumors presented the highest hemoglobin concentration, whereas the reflected cyst

case presented the lowest value. It appeared that the nodules and the calcifications in the

cases we presented did not show distinct differences in the total hemoglobin

concentration, nor did their individual de-oxy hemoglobin concentrations and oxy-

hemoglobin concentrations. Thus, questions about the possibility on diagnose aid from

the hemoglobin resolution remained to be answer, and a furthermore approach then

started.

41 Cases

Total hemoglobin concentration

In Figure 4-6, we presented all cases for de-oxy hemoglobin concentration [Hb] and

oxy-hemoglobin concentration [HbO,] from reported diseased lesions (as (a) illustrates

the mean concentration of the diseased lesions, and (b) illustrated their corresponding

peak values), slight gap emerged as differentiation the malignant cancer tumors from

benign tumors accord with the selected cases we presented above. As the malignant

cancer tumors showed high values in both [Hb] and [HbO, ], total hemoglobin









44



concentration (average values) were calculated and graphed in Figure 4-7 with patient


ages. Three patients did not repeat their ages and were not included in the graph. The


average total hemoglobin concentration of the corresponding diseased lesions were


approximate 35y/M (calcification), 80/uM (nodule), and 5/pM (lymph node nodule)



respectively.



Mean Values


U
100


80


60


40
*
20

0o
0 5 10 15 20 25 30 35 40 45 50
[Hb] (uM)



Peak Value

160

140

120

100

80



40



0
ot-------------------------


n 10


20 i0 40 50 60 70 80 90


[Hb] (uM)
b

Figure 4-6 Graph illustrates the quantitative value for the raw de-oxy hemoglobin

concentration [Hb] and oxy-hemoglobin concentration [HbO2] from reported

diseased lesions, a mean values from the lesions, and b peak value. For the

special cases that illustrated lower concentration comparing to their

surrounding tissues, we picked out the corresponding minimum values.


*Nodule
Cancer
Cyst
Calcification


*Nodule Hb02
mCancer Hb02
CystHb02
Call Hb02












140

120

100 -

80 Nodule
80
U Cancer
S60 m Cyst
60 Calcification
$ t
40

20


0
0 i-------------------------
0 20 40 60 80 100
Age (year)

Figure 4-7 Mean hemoglobin concentrations of reported diseased lesion

Physiologically, it has been demonstrated that most miscellaneous lesions (nodule

cases, including blunt trauma, mastitis and so forth), usually stimulate increased blood

flow into the very tender, firm affected area [5]. Another familiar tumor always reported

as miscellaneous lesions, lipomas, is a benign fat tumors made up of mature adipose cells.

Lipomas might have similar absorption and scattering characteristics as the surrounding

fat, but should tend to have more blood vessels than them; however, its vascularity may

not be sufficient enough to differentiate the tumors from the normal tissues. By and large,

most miscellaneous lesions should reflect higher hemoglobin concentrations due to a

more sufficient localized blood flow; even in the lipomas cases, though they might not

have distinct higher hemoglobin concentrations than normal tissue (but should be no less

than that of normal tissues), their corresponding hemoglobin concentrations should help

differentiate them as they are always correlated with plenty of surrounding fat which

have extremely low hemoglobin concentrations. Not only the typical cases we presented









out here, but all the 17 miscellaneous lesions we investigated, the nodule lesions could be

estimated with higher hemoglobin concentrations comparing to its surrounding tissues

except two: one was reported as indistinct abnormalities of fibroglandular, and the other

one was at last demonstrated to be clusters of lymph node (a small bean-shaped structures

that are found throughout the body, acting for filter substances in lymph and help fight

infection and disease); both of which should have lower vascularity physiologically. The

mean values for total hemoglobin concentrations of normal breasts usually ranged from

greater than 20/iM to less than 50/iM as K. D. Paulsen et al investigated in clinical

cases [15, 16, 17], the hemoglobin concentrations for our nodule lesions were mostly

more than or approximate at 50/AM in average values (11 cases), 4 cases approximate in

the normal ranges, and the indistinct fibroglandular (approximate 16/iM in average) and

distinct lymph node (approximate 5k/M in average) referred just now has lower total

hemoglobin concentrations than normal tissue as showed in its corresponding image. The

results accorded well with what the Physiological and Pathology theories suggested.

The fibrocystic disease or fibrocystic change of breast tissue covers a broad spectrum

of localized or diffuse stomal and glandular alterations that can produce diffuse masses,

distinct tumors and/or non-palpable lesions [5]. The cases we selected for investigation

have been diagnosed with large or small liquid-filled cysts formed by the dilatation of

ducts. Customarily, cysts are a very common breast disease and rarely associated with

cancer. Unlike cancerous tumors which are solid, cysts are masses that generally

containing dark brown fluid in the breast. The fluid obtained from cysts usually contains

degenerated cells, secretions and blood that may or may not have distinctive absorption

characteristics. Theoretically, the fluid-filled cysts should reflect lower total hemoglobin









concentration than the solid tumors for the reason of containing plenty of non-bloody

fluid. Comparing to what we obtained that illustrated in graph of Figure 4-6, by

appearances, most cysts did exhibit obvious lower HbT than most other diseased lesion

categories.

The third category we focused on, calcifications, are tiny flecks of calcium, like

grains of salt, in the soft tissue of the breast that can sometimes indicate the presence of

an early breast cancer [5]. Calcifications usually are not detected by ultrasound, but on

the other hand, they could appear on a mammogram. Using our technique, we could

image calcifications lesions, and differentiate them in relative chromophore concentration

images successfully. But when making comparisons to other lesion categories in

hemoglobin concentration, we could not differentiate benign calcifications clearly from

miscellaneous lesions and cancer cases. Studies reported that big calcifications

(diagnosed as "macrocalcifications") are not always associated with cancer. Groups of

small calcifications huddled together, (diagnosed as "clusters of microcalcifications") are

usually combined to extra breast cell activity-most of which are non-cancerous extra

cell growth. But sometimes clusters of microcalcifications, the tiny calcium deposits

occurring in areas of early cancer, could be regarded as the first x-ray evidence of cancers.

Hence, the calcifications lesion should have higher hemoglobin concentration correlated

with extra cell growth needing for a more sufficient localized blood flow, and some cases

that have clusters of microcalcifications might demonstrate some comparability to early

cancer cases since the malignant tumors are usually associated with increased

microcalcifications.









The malignant tumors in breast are cancer that may be life-threatening [5].

Biologically, it has been demonstrated that various growth factors and cytokines inheres

in cancer cells that could stimulate host responses of inflammation, blood vessel

proliferation (cancer related angiogenesis) and fibrosis (desmoplasia). Hence, it is easy to

apprehend that the malignant cancer tumors could reflect extremely high hemoglobin

concentrations comparing to most benign tumors.

Most of the diseased lesions involved in this investigation indicated increased total

hemoglobin concentration compared to their surrounding tissue except the two nodule

cases we referred to before; and all the diseased tumors could be distinguished from its

surrounding tissues from quantitative hemoglobin concentration images in accord with

the reported position. It appears that the optical mammography could allow most of the

physical masses that were present on mammography are detectable and distinguishable

not only for the images of optical properties (/u, and /,'), but also for the images of

hemoglobin concentration in vivo. Quantitatively, the results plotted in graph of Figure 4-

6 indicated that the mean values of total hemoglobin concentration can secrete functional

information of different diseased lesions. The solid tumors (including solid nodules,

calcifications and malignant cancer tumors) generally indicate significantly higher

hemoglobin concentration than the tumors that contains non-bloody fluid (include cyst,

fibro-glandular and lymph node cases); and the malignant cancer tumors generally

shadowed significant higher hemoglobin concentration than most benign tumors except

some cases that containing clusters of micro-calcifications. However, not all the lesions

have significantly higher/lower hemoglobin concentration than the whole breast tissue

plane for imaging, as even the normal breast itself is not a homogeneous organ, illustrated









in Figure 4-6 illustrated, we could detect that most of the non-bloody-liquid-contain

tumors shadowed their corresponding average hemoglobin concentration in the normal

tissue range. On the other hand, for the solid tumor cases, the possibility for quantitative

hemoglobin concentration aiding diagnosis introduced its existence; since most solid

tumors distinguished themselves not only from their surrounding tissue in the

corresponding images directly, but also shadowed higher values compare to the normal

tissues. It is also interesting to note that the malignant cancer tumors appeared to produce

high hemoglobin concentration, as the hemoglobin concentration for most benign tumors

are less than 60/iM on average, while the malignant cancer tumors were usually greater

than or at least approximate 60/iM .

Hemoglobin concentration and oxygen saturation

The total hemoglobin concentration appears to be a promising avenue for a

diagnostic aid by quantitative optical chromophore concentrations image, providing a

deeper view from hemoglobin concentrations. The total hemoglobin concentration is one

factor correlated by de-oxy hemoglobin concentration and oxy-hemoglobin concentration,

and the other important factor combined with them is the oxygen saturation.

Theoretically, the oxygen saturation which is related to the pO2 of oxygen in tissue,

could help in evaluating the hypoxia in tumors [17]. Evidence have been accumulated

that hypoxia is usually involved in a vicious circle of a fundamental biologic mechanism

of the malignant tumors. Hence, the hypoxia was regarded as one of the critical factor for

separating malignancies from benign tumors. As a result, in several cases of the

malignant tumor exhibit smaller oxygen saturation compared to healthy tissue, but there

are also cases for which the opposite holds true. As known, results of cases performed by









Fantini et al (1998), Tromberg et al (2000) and Chernomordik et al (2002) observed

decreases in oxygen saturation of the tumor [18, 19, 20], whereas McBride et al (2002)

and K. D. Paulsen et al (2005) each reported a case with no essential difference in

oxygen saturation between tumor and healthy tissue [21, 22]. Meanwhile, for clinical

cases, Dirk Grosenick et al (2004) recorded that blood oxygen saturation enhanced in

some cases and reduced in some other cases under the processing of dual wavelength

time-domain optical mammography [23, 24].

From our oxygen saturation images, we could distinguish 4 malignant cancer tumors

from its surrounding tissues with obvious oxygen saturation decreases; and 1 malignant

cancer tumors have slightly decreases comparing to its surrounding health tissues in

oxygen saturation, but the difference is not so significant to define the tumor shape; the

other 6 malignant cancer tumors did not show differences in their oxygen saturation

images. For the benign tumors, we observed 2 unclear decreases in the tumor position (1

nodule and calcificationn, 3 clear decreases (1 calcification and 2 nodule including the

fibroglandular nodule case), and 1 increase oxygen saturation cyst case, whereas the other

cases presented no evident differences between the diseased lesion and healthy tissue.

Thus, our results seem to suggest that oxygen saturation decreases were more often

visible in vivo for malignant cancer tumors (5 out of 11 lesions) and calcification lesions

which usually be regarded as signs for early cancer (2 out 6 cases). In Contrast, benign

nodules usually do not differentiate obvious differences in oxygen saturation from their

surrounding healthy breast tissue. For more quantitative analysis, as most tumors could

not be distinguish out from normal tissues, whereas their corresponding total hemoglobin

concentration images did show distinct difference from their surrounding tissue; hence,







51



we picked out one point from each tumor, i.e. the point showed peak value for total


hemoglobin concentration (for those tumors showed lower HbT than their surrounding


tissue, we picked out the point with minimum value in them) and then calculated their


corresponding oxygen saturations. These quantitative data were plotted in a HbT StO2


plane showed as Figure 4-7. Clearly, compared to Figure 4-6 just have one functional


coefficient of HbT, the data in Figure 4-7 may be an improvement to distinguish the


malignant cancer tumors from benign tumors. However, this approach fails to yield more


information to distinguish fluid-filled cysts than using total hemoglobin concentrations


alone.


1
0.9
0.8

0. 7
0.6
0. 4
r

0. 4
> 0.3

0. 2
0. 1
0


*Nodule
* Cancer
Cyst
Calcification


0 20 40 60 80 100
HbT (uM)


120 140 160


Figure 4-8 Graph illustrates the quantitative value for the maximum hemoglobin
concentration and the corresponding oxygen saturations in reported diseased
lesion

Water Content

As above discussed, our model results of hemoglobin concentration could accord


with the nature of tumors, not only visible but also quantitative for both malignant and


4L ; 1


';7'









benign cases. However, it may not be so promising when making quantitative analysis on

the water content obtained by the absorption coefficients /,.

Similar to what was presented by the selected examples, most tumor cases could be

visible and distinguish out at the acceptable location in accordance with what the

mammography or ultrasound reports indicated. But the values of water content we gain

were not reasonable enough to make further quantitative analysis: negative values (as the

cancer cases we presented) or values exceed 100% (as the first nodule case we presented)

existed in some cases.

The water influence on the absorption coefficients in breast tissue was not as

significant as hemoglobin especially for light in short wavelength range. Only in the long

wavelength, particularly in the 900-1000nm range, did absorption by water present a

stronger influence. The wavelength we choose for these experiments were between

638nm to 965nm, which were considered to be in the range of central wavelength region,

and the tissue absorption is the lowest which could provide the allowance for light

penetration through a few centimeters of tissue. From the reconstructed absorption

coefficient image, we could notice that we always obtain clear images in most short

wavelength that influenced more by hemoglobin. The experiment data for longest three

wavelengths we considered most helpful in separate the water characterization usually

were too weak to gain qualify enough reconstructed optical coefficients. That might be

the primary reason for the problems on the values of water contents. Besides, our model

for chromophore concentration calculation was a simplified model which does not

account enough for the complexity of the breast tissue. Based on this model, the influence

from other absorbers such as melanin, carotene, proteins, DNA, vegetable oil and lipids









were neglected, which might reflect uncertain factors on the results. It has also been

demonstrated that interactions of various chromophores with other tissue and cellular

constituents may cause spectral shifts in measurements taken in vivo when compared to

those taken in vitro or as purified chromophores in solution. In addition, the tumor

location and visible size may influence the accuracy of tumor optical properties. All these

factors could be added into the uncertainty of results of the chromophores concentration.

Scattering Spectra

The scattering spectra were thought to provide important physiological content. To

best of our knowledge, we expect the collagen-rich-tissue-rich-breast would display a

higher scatter power than the breasts with more adipose. Visibly, the image results we

obtained for scattering power using 10-wavelength experiment produced evidence to

encourage the view that most of the diseased lesions could be distinguished clearly from

the scattering power images (except 4 nodule and 1 cyst cases), with lesion positions well

accord with the pathological diagnoses, whereas the results from 3-wavelength were not

as good. Figure 4-9 presents the mean value of scattering power with their mean total

hemoglobin concentration and age. There were no evident gaps between different tumors

in the HbT SP (scattering power) plane that were observed. Tromberg et al (2002)

have reported that scattering power showed decrease with age in normal breast [7], in our

tumor cases, the decreases combined with age were not significant. Moreover, images of

scattering amplitude failed to illustrate anything.























M









20 40 60 80 100 120 14(
HbT (uM)



















____________________*_________l___


I I


0 10 20 30 40 50
Age (Year)


*Nodule
* Cancer
Cyst
Calcification


*Nodule
* Cancer
Cyst
Calcification


60 70 80 90


Figure 4-9 Graph illustrates the mean value of scattering power with the total hemoglobin
concentration. a and the patient age b in reported diseased lesion.


As the algorithm accounting for scattering amplitude and scattering power employed


a non-linear to linear transformation and a linear to non-linear transformation; besides,


2. 5




2




i 1.5






0.


0.5









there are more than 2 equations for solving the two scattering parameters, thus, the

combined equations might be ill-conditioned. Any small inaccuracy in the measurements

of optical coefficients might be enlarged while calculation. Hence, it leaves future work

for algorithm improvements on this.

Conclusion

During this investigation, patients suspected of having different breast diseases, were

imaged using a multi-wavelength optical scanning in the frequency domains. Optical

properties were reconstructed in certain wavelengths ranging from 638nm to 965nm, and

calculating the physiological functional information such as hemoglobin concentration,

water content, scattering amplitude and power, with the introduction of their

corresponding spectra. Detectability of tumors for these physiological functional

coefficients was recorded and evaluated through comparison with breast x-ray

mammography and ultrasound reports, or biopsy reports if suitable.

For 41 diseased lesions we selected, all of them could be detected and distinguished

by oxy-hemoglobin and de-oxy hemoglobin concentration. Additional factors such as

total hemoglobin concentration and oxygen saturation were also obtained and analyzed.

The quantitative values of total hemoglobin concentration were in accord with the

physiological and pathological nature of different tumors. For most tumors we detected,

total hemoglobin concentration was larger compared to its surrounding tissues (except for

one lymph node case and one fibro-glandular case); and especially for malignant cancer

tumors, evident increases could be observed more distinct than most of the benign tumors.

Furthermore, when combined with the oxygen saturation, it seems promising that this

optical image technique using continuous-wave will allow us to distinguish malignant

tumors from benign. From this study, we can explore approaches using mathematical






56


classification from all these optical and physiological coefficients that we obtained, or a

possibility of carcinomas diagnosis assistance could be found.

Both water content and scattering power were obtained, and tumors were visible in

positions in accordance with mammography and ultrasound examination reported in the

corresponding images, but deeper quantitative analysis is a remaining problem. Hence,

algorithms accounting for each of these will require further improvements before they

can be used.















LIST OF REFERENCES


1. M. Cutler, "Transillumination as an aid in the diagnosis of breast lesions," Surg.
Gynecol. Obstet. 48 (1929) 721-728.

2. E. Carlsen, "Transillumination Light Scanning," Diagnostic Imaging 4 (1982) 28-
34.

3. K. D. Paulsen and H. Jiang, "Spatially varying optical property reconstruction
using a finite element diffusion equation approximation," Med. Phys. 22 (1995)
691-702.

4. Y. Xu, X. Gu, T. Khan, and H. Jiang, "Absorption and scattering images of
heterogeneous scattering media can be simultaneously reconstructed by use of dc
data," Appl. Opt. 41 (2002) 5427-5437.

5. S. Thomsen and D. Tatman, "Physiological and Pathological Factors of Human
Breast Disease That Can Influence Optical Diagnosis," Annals of the New York
Academy of Sciences 838 (1998) 171-193.

6. V. G. Peters, D. R. Wyman, M. S. Patterson, and G. L. Frank, "Optical properties
of normal and diseased breast tissues in the visible and near-infrared," Phys. Med.
Biol. 35 (1990) 1317-1334.

7. A. E. Cerussi, D. Jakubowski, N. Shah, F. Bevilacqua, R. Lanning, A. J. Berger, D.
Hsiang, J. Butler, R. F. Holcombe, and B. J. Tromberg, "Spectroscopy enhances the
information content of optical mammography," J. Biomed. Opt. 7 (2002) 60-71.

8. B. W. Pogue, S.g Jiang, H. Dehghani, C. Kogel, S. Soho, S. Srinivasan, X. Song,
Tor D. Tosteson, S. P. Poplack, and K. D. Paulsen, "Characterization of
hemoglobin, water, and NIR scattering in breast tissue: analysis of intersubject
variability and menstral cycle changes," J. Biomed. Opt, 9 (2004) 541-522.

9. B. J. Tromberg, O. Coquoz, J. B. Fishkin, T. Pham, E. R. Anderson, J. Butler, M.
Cahn, J. D. Gross, V. Venugopalan, and D. Pham, "Non-invasive measurements of
breast tissue optical properties using frequency-domain photon migration," Phil.
Trans. R. Soc. Lond. B. 352 (1997) 661-8.

10. A. Corlu, T. Durduran, R. Choe, M. Schweiger, E. M. Hillman, S. R. Arridge, and
A. G. Yodh, "Uniqueness and wavelength optimization in continuous-wave
multispectral diffuse optical tomography," Opt. Lett. 28 (2003) 2339-2341.









11. A. Corlu, R. Choe, T. Durduran, K. Lee, M. Schweiger, S. R. Arridge, E. M.
Hillman, and A. G. Yodh, "Diffuse optical tomography with spectral constraints
and wavelength optimization," Appl. Opt. 44 (2005) 2082-2093.

12. S. J. Matcher, M. Cope, and D. T. Delpy, "In vivo measurements of the wavelength
dependence of tissue-scattering coefficients between 760 and 900 nm measured
with time-resolved spectroscopy," Appl. Opt. 36 (1997) 386-396.

13. A. E. Cerussi, A. J. Berger, F. Bevilacqua, N. Shah, D. Hakubowski, J. Butler, R. F.
Holcombe, B. J. Tromberg, "Sources of absorption and scattering contrast for Near-
Infrared optical mammography," Acad Radiol 8 (2001) 211-218.

14. C. Li and H. Jiang, "A calibration method in diffuse optical tomography," J. Opt.
A: Pure Appl. Opt. 6 (2004) 844-852.

15. B. W. Pogue, S. P. Poplack, T. O. McBride, W. A. Wells, O. K. S., U. L. Oserberg,
and K. D. Paulsen, "Quantitative hemoglobin tomography with diffuse near-
infrared spectroscopy: pilot results in the breast," Radiology 218 (2001) 261-6.

16. S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, C. Kogel, S. Soh, J. J. Gibson,
T. D. Tosteson, S. P. Poplack, and K. D. Paulsen, "Interpreting hemoglobin and
water concentration, oxygen saturation and scattering measured in vivo by near-
infrared breast tomography," PNAS 100 (2003) 12349-12354.

17. M. Hockel, P. Baupel, "Tumor hypoxia: definitions and current clinical, biologic,
and molecular aspects," J. Natl. Cancer Inst. 93 (2001) 266-276.

18. K. T.Moesta, S. Fantini, H. Jess, S. Totkas, M. A. Franceschini, M. Kaschke, and P.
M.Schlag, "Contrast features of breast cancer in frequency-domain laser scanning
mammography," J. Biomed. Opt. 3 (1998) 129-36.

19. B. J. Tromberg, N. Shah, R. Lanning, A. Cerussi, J. Espinoza, T. Pham, L.
Svaasand, and J. Butler, "Non-invasive in vivo characterization of breast tumors
using photon migration spectroscopy," Neoplasia 2 (2000) 26-40.

20. V. Chemomordik, D. W. Hattery, D. Grosenick, H. Wabnitz, H. Rinneberg, K. T.
Moesta, P. M. Schlag, and A. Gandjbakhche, "Quantification of optical properties
of a breast tumor using random walk theory," J. Biomed. Opt. 7 (2002) 80-7.

21. T. O.McBride, B. W. Pogue, S. Poplack, S. Soho, W. A. Wells, S. Jiang, U. L.
Oserberg, and K. D. Paulsen, "Multispectral near-infrared tomography: a case
study in compensating for water and lipid content in hemoglobin imaging of the
breast," J. Biomed. Opt. 7 (2002) 72-9.

22. S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, and K. D. Paulsen, "Spectrally
constrained NIR tomography for breast imaging: simulations and clinical results,"
SPIE 5693 (2005) 293-300.






59


23. D. Grosenick, H. Wabnitz, K. T. Moesta, J. Mucke, M. Moller, C. Stroszczynski, J.
Stossel, B. Wassermann, P. M. Schlag, and H. Rinneberg, "Concentration and
oxygen saturation of haemoglobin of 50 breast tumours determined by time-domain
optical mammography," Phys. Med. Biol. 49 (2004) 1165-81.

24. D. Grosenick, K. T. Moesta, H. Wabnitz, J. Mucke, C. Stroszczynski, R.
Macdonald, P. M. Schlag, and H. Rinneberg, "Time-domain optical
mammography: initial clinical results on detection and characterization of breast
tumors," Appl. Opt. 42 (2003) 3170-3186.















BIOGRAPHICAL SKETCH

My name is Lin Chen. I was born on March 14, 1981, as a girl in Jiangjing, a little

town of Chongqing, in southwest China. After accomplishing high school with a brilliant

record, I fought my way out of millions of candidates into Peking University, one of the

most prestigious and selective universities in China. As an undergraduate in Peking

University (1999.7.-2003.6.), I intended to major in electronics engineering. During this

period, I enrolled in some courses and projects on optics and signal processing, and I

found myself enamored with the area, especially the optical imaging. After graduation, I

pursued graduate school at my advisor's suggestion, and then enrolled in the master's

program of Clemson University. At the Physics and Astronomy Department in Clemson

University, my advisor, Prof. Huabei Jiang, brought me into the research for biomedical

optical imaging of human breast. Under his guidance, I began to do clinical experiments

and build mathematical models for tissue functional information reconstruction and

analysis. In 2005, I transferred to the University of Florida with the whole Biomedical

Optics Laboratory, and continued my master's Program and researches in the Biomedical

Engineering Department here. I enrolled in courses on anatomy and physiology, which

helps a lot in my research. Now, I have achieved some approaches on the differentiation

of malignant from benign breast lesions based on functional diffuse optical tomography.




Full Text

PAGE 1

DIFFERENTIATION OF MALIGNANT FROM BENIGN BREAST LESIONS BASED ON FUNCTIONAL DIFFUSE OPTICAL TOMOGRAPHY By LIN CHEN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Lin Chen

PAGE 3

This document is dedicated to the graduate students of the University of Florida.

PAGE 4

iv ACKNOWLEDGMENTS I would like to thank Dr. Huab ei Jiang, my thesis advisor, for his guidance. I would like to express appreciation to all the member s in the Biomedical Optics Laboratory of Florida. My special thanksgiving goes to Changqing Li and Xiaoping Liang, for many discussions and advices in this project. I appreciate Qizhi Zhang and Hongzhi Zhao in their help for the procedur e of clinic experiments. The Oconee Memorial Hospital and the Greenville Hospital have been a great help to our clinic experiments in introducing patients and guidance on the lesion position. I thank my parents for their care, love and support on my study.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT....................................................................................................................... ix CHAPTER 1 INTRODUCTION........................................................................................................1 2 THEORY......................................................................................................................4 Introduction................................................................................................................... 4 Reconstruction Algorithm............................................................................................5 Characterization of Spect ra in Breast Tissue..............................................................11 3 MATERIALS AND METHODS...............................................................................15 General Approach.......................................................................................................15 Instrument and Calibration.........................................................................................17 System Setup.......................................................................................................17 System Calibration Method.................................................................................19 Phantom Experiments for System Calibration....................................................20 Results of Calibration Experiments.....................................................................20 Clinical Experiments..................................................................................................24 Fitting Consideration..................................................................................................25 Absorption Fitting...............................................................................................25 Scattering Fitting.................................................................................................27 4 RESULTS AND DISCUSSIONS...............................................................................28 Tumor Result-Images for Selected Examples............................................................28 Discussions.................................................................................................................41 Hemoglobin Concentration.................................................................................41 Represented cases.........................................................................................41 41 Cases........................................................................................................43

PAGE 6

vi Water Content......................................................................................................51 Scattering Spectra................................................................................................53 Conclusion..................................................................................................................55 LIST OF REFERENCES...................................................................................................57 BIOGRAPHICAL SKETCH.............................................................................................60

PAGE 7

vii LIST OF TABLES Table page 2-1 The Optical Properties of Human Female Breast.....................................................12 3-1 The Extinction Coefficients of All the Wavelengths in the Experiments................16 3-2 The Number of Lesions Select ed for Experiment and Analysis..............................25

PAGE 8

viii LIST OF FIGURES Figure page 2-1 Flow chart of the Newton-type iterati on for estimating the distribution of the optical properties......................................................................................................10 3-1 Schematic of 2 experimental systems......................................................................18 3-2 Geometry of the phantom configuration..................................................................21 3-3 Comparison between the reconstructed abso rption images of or iginal data and of calibrated data for cases of 2:1 absorption contrast.................................................22 3-4 Reconstructed absorption images for heterogeneous phantoms with or without calibration from different diameters in the case of 1.4:1 absorption contrast..........23 4-1 Corresponding position from report to optical 2D images.......................................29 4-2 The X-ray mammography, reconstructe d optical properties and functional information images of a selected nodule case..........................................................30 4-3 The X-ray mammography, reconstructe d optical properties and functional information images of a se lected calcification case.................................................33 4-4 The X-ray mammography, reconstructe d optical properties and functional information images of a selected cyst case..............................................................36 4-5 The X-ray mammography, reconstructe d optical properties and functional information images of a selected cancer case..........................................................39 4-6 Graph illustrates the quantitativ e value for the raw de-oxy hemoglobin concentration and oxy-hemoglobin concentration ..................................................44 4-7 Mean hemoglobin concentrations of reported diseased lesion.................................45 4-8 Graph illustrates the quantitativ e value for the maximum hemoglobin concentration and the corre sponding oxygen saturations.........................................51 4-9 Graph illustrates the mean value of scattering power with the total hemoglobin concentration. a and the patient ag e b in reported diseased lesion.........................54

PAGE 9

ix Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DIFFERENTIATION OF MALIGNANT FROM BENIGN BREAST LESIONS BASED ON FUNCTIONAL DIFFUSE OPTICAL TOMOGRAPHY By Lin Chen August 2006 Chair: Huabei Jiang Major Department: Biomedical Engineering In this thesis, optical images based on a clinical study with patients including 11 cancer and 30 benign cases were processed and analyzed. Using multi-spectral diffuse optical tomography systems coupled with fin ite element reconstruction algorithms, we first obtained optical absorption and scattering maps of the breast and then derived tissue functional images from the recovered ab sorption and scattering images at multiwavelengths. In order to obtain accurate in vivo images, a calibration database was developed which was based on a series of homogenous phantom measurements with a range of phantom dimensions. Hemoglobin (both oxy and de-oxy) concentrations and water content images were obtained from th e multi-spectral absorption images, and a Mie scattering theory approximation was applied to extract scattering amplitude and power. Functional parameter images of the 41 cases were investigated, and correlation plots of different function parameters were illustrat ed and compared among 4 disease categories including cancer, cyst, nodule and calcula tion. We found that the majority of the

PAGE 10

x carcinomas exhibited increased total hemoglobi n concentration compared to the healthy and other benign tissues, and the correlation between total hemoglobi n concentration and oxygen saturation of these diseased tissues s howed a clear separation between malignant and benign lesions, while the separation among th e benign lesions is not apparent for the cases examined.

PAGE 11

1 CHAPTER 1 INTRODUCTION Optical methods for the detection of breast cancer, especially for the early detection of cancer, can be traced back to as early as 1929 [1], when it was introduced by Max Culter. Decades of studies improved the diffuse optical tomography (DOT) as a noninvasive imaging technique that could pr ovide quantitative ab sorption and also a scattering distribution. In the 1980s, Carlsen [2] introduced spectral breast imaging by restricting the light source of a medical transillumination imager. Because of the introduction of these technologies, optical im aging of human tissue using near-infrared (NIR) light provides the possibi lity of obtaining new types of physiological information from the tissue in vivo while the traditional met hod, using conventional x-ray mammography techniques could only provide structural information. The near-infrared light passing through breast tissues is sensitive to several physiological components such as hemoglobin, water, lipid, melanin, carotene, proteins, DNA, and so forth. Thus NIR absorption in br east tissue is influenced by hemoglobin concentration, oxygen saturation, wa ter content, and to a lesser extent by lipid. Therefore, NIR techniques could be fashioned into an inexpensive and portable alternative solution for distinguishing malignant (even in an earl y stage) from benign diseased or health tissues. NIR can accomplish this by obtaining quantitative hemoglobin concentration, oxygen saturation, water fraction and other functional information from absorption distribution of tissue. The goal of the studies described in this thesis is a pilot approach to evaluate the possibility from substantive patient cases.

PAGE 12

2 Our experimental device was developed as a silicon photodiodes-based DOT system, and employed a finite element algorithm for the frequent-domain optical data reconstruction based on a well-known diffusi on equation. In Chapter 2, we review the reconstruction algorithm for absorption and redu ced scattering coefficients in detail. Also, it includes a general review on the spectra characteriz ation of breast tissue. The experimental system and the preparation for clinical examination are covered in Chapter 3. In order to assist in obtaining the first quantitative reconstructed data of both absorption and reduced scatte ring coefficients, the calibration method was employed. We built a database based on a se ries of homogenous phantom w ith a range of dimensions (from 60mm to 11cm with an increment of 10 mm) to fit with various sizes of clinical human breasts. Meanwhile, clinical expe riments were performed on more than 100 volunteers including th ose with malignant cancer, beni gn diseased and healthy breast conditions. Typical cases with both mamm ography and ultrasound reports (and biopsy reports if they exist) were selected, recons tructed and evaluated. We also introduced our fitting methods on both absorption and scatting spectra based on the simplified models in Chapter 3. Results of the clinical experiments are pr esented in Chapter 4. With a comparison to the mammography and ultrasound reports, we were able to recognize the tumors in the corresponding position, and ev aluate the functional inform ation. We noticed specific differences for total hemoglobin, correspond ing to physiological and pathological knowledge, with different kinds of tumors. A se ries of analyses were carried out for the purpose of investigating these visibl e images and quantitative values.

PAGE 13

3 These analyses attempt to provide a basis for the aid of diagnos ing malignant cancer and other breast diseases.

PAGE 14

4 CHAPTER 2 THEORY Introduction Yearly mammograms are recommended starti ng at age 40 and continuing for as long as a woman is in good health; and clinical br east exam should be taken as part of a periodic health exam, preferably at least ev ery three years for women in their 20s and 30s, and every year for women, for the presence of breast cancer which is one in eight women in the United States. Thus, diffuse optical tomography (DOT), tries to investigate an alternative method for the early dete ction of preclinical breast cancer. Currently, conventional x-ray mammogra phy and palpation are the most common method for breast cancer detection. However, obvious limitations to conventional x-ray mammography have been recognized. For exam ple, conventional x-ray mammography is not suitable for young women in early pre-menopa usal stage, by reason of their increased cellularity and subsequent ra diodense tissue structure. That is to say, due to hormone fluctuations, the pre-menopaus al women with preclinical br east cancer are at increased risk of more rapid tumor growth. In addition, the positive predictive value of conventional x-ray mammography is quite low in both medical and economic terms, and as a result, numerous biopsies are required to be performed each year. Moreover, women with the familial gene for breast cancer (e.g., family history, genetic tendency, past breast cancer) might experience risk when subjected to the x-irradiation. As a result, DOT, a non-ionizi ng, non-invasive near-infra red optical imaging holds great promise to become an alternative for breast cancer screening, especially for cancer

PAGE 15

5 in early stage. Using a laser light source, this optical method attempts to produce an image of the inside of the breast, with uni que capability for scr eening high radiodense breasts usually for premenopausal women. Recent studies have suggested that biomedical optical imaging of breast tissue has significan t advantages for breast cancer detection and diagnosis, which helps a lot fo r retaining the corre sponding treatment to be keeping pace with the increased inci dence of the breast cancer. Meanwh ile, this method has no harm to human body even for patients with familial gene for breast cancer; and the instrumentation for optical imaging is much lower in cost than that for x-ray mammography. Further, the DOT could obtain quantitative absorption and scattering distributions from breast tissue, which can not be meas ured by conventional x-ray mammography or other radiologic techniques. The spectral de pendence of quantitativ e tissue absorption a and reduced scattering s distributions could provide ti ssue functional information in breast with the introduction of near-infrared (NIR) migration spectro scopy. Multi-spectral measurements helps for determination of the concentrations of de-oxy and oxyhemoglobin, water, and other components in breast; and the scattering properties of the tissue could also yield important physiol ogical information, such as the scattering amplitude and scattering power. These typical values within the breast are believed to help doctors for better di agnoses on breast diseases. Reconstruction Algorithm Our reconstruction algorithm for absorpti on and reduced scattering coefficients, previously described in detail [3, 4], is an iterative finite element algorithm based on the well-known diffusion equation

PAGE 16

6 (,)(,)(,)i a cDrrSr (1) where D is the diffusion coefficient, (,) r is the radiance, a is the absorption coefficient, c is the wave speed in the medium, (,) Sr is the source term as an ite time variation is assumed. And the diffusion coefficient D can be expressed as 1 3[(1)]asD g (2) where s is the scattering coefficient and g is the average cosine of the scattering angle. And the reduced scattering coefficient is defined as '(1) s sg .With known a and s distribution, the diffusion equation beco mes a standard boundary value problem for spatially varying radiance subject to appr opriate boundary conditions (BC’s). There are three classical boundary conditions for this diffusion equation: i) specification of the field, (Dirichlet or Type I); ii) specification of its flux, ˆ Dn (Neumann or Type II); iii) specification of a relationship betwee n field and flux (mixed or Type III) In our study, we employed Type III BC’s in the reconstruction algorithm, that is ˆ Dn where ˆ n is the unit vector normal to the boundary surface, and is related to the internal reflection, which can be derived from the Fresnel reflection coefficient. For the finite element forward solution, and ˆ FDn are expanded as the sum of coefficients multiplied by a set of locally spatially varying Lagrangian basis functions 1 N jj j (3a)

PAGE 17

7 1 M jj jFF (3b) where j is the known basis and j jF are the respective radiance and flux at node j Similarly, a and D are expanded as a collection of unknown parameters multiplied by a known spatially varyin g expansion function 1 K kk kDD (4a) 1L all l (4b) As a result, the diffusion equation becomes 111 1 NKL jkkjillij jkl M ijji ji D c SFds (5) which could be express in the matrix form as Ab (6) where the elements of matrix A are 11 KL ijkkjillij kli aD c (7a) as indicating integration over the problem domain. The column vector is composed of the photon density i at node i. And b is filled with elements that 1 M iijji jbSds (7b)

PAGE 18

8 where expresses integration ove r the boundary surface with jF replaced by j as Type III BC defined, and M is the number of boundary nodes. With finite element discretization, the photon density (computed optical data) is obtained as the solution of the diffusion equa tion. Then a regularized NewtonÂ’s method is exploited here to update the in itially guessed optical property distribution iteratively in order to minimize an object function compos ed of a weighted sum of the squared difference between computed and measured optical data at the medium surface. We assume that the computed and/or measured values of or F are analytic functions of D and a and D and a are independent since s a Then and F could be Taylor expanded as an assumed ,aD distribution, which is a perturbation away from some other distribution, ,aD such that a discrete set of radiance and flux values can be expressed as ,,aaa aDDD D (8a) ,,aaa aFF FDFDD D (8b) where DDD and aaa If the assumed optical property distribution is close to the true profile, the left-hand side of (8) can be c onsidered as true data (either imposed or observed), and the re lationship truncated to produce ocJ (9a) where o and c are observed and calculated [based on the estimated ,aD distribution] data, either or F, depending on boundary conditions for 1,2,,iM

PAGE 19

9 locations and kD for 1,2,kK and l for 1,2,,lL ; and J is Jacobian matrix consisting of derivatives of with respect to D or a at each boundary observation node. 111111 1212 222222 1212 KL KLDDD DDD J 1212 MMMMMM KLDDD (9b) And is the vector that gives the perturbation of a and D 1 2 1 2 K LD D D 1 2 o o o o M 1 2 c c c c M (9c) In order to realize an invert ible system of equations for the Eq. (9a) could be multiplied by TJ on both sides to obtain TTocJJJ (10) which can be used for updating the optical property distribution. As the matrix TJJ is known to be ill conditioned, techniques should be performed to regularize or stabilize the decomposition of TJJ. Thus, a quantity is adding to the diagonal of TJJ in practice, and the problem transformed to

PAGE 20

10 TTocJJIJ (11) where I is the identity matrix and may be a scalar or a diagonal matrix. Figure 2-1 Flow chart of the Newton-type iter ation for estimating the distribution of the optical properties START Initial Value ,aD Forward Computation FEM solution of Diffusion Equation (,)(,)(,)i a cDrrSr Measured data o i Converged? 2 24 110M co ii i STOP Inverse Computation Build Jacobian Matrix: Solve matrix Equation: TTocI Update Optical Property Values ,aD 'iiiDDD 'iii

PAGE 21

11 By adding a contribution to the diagonal terms in Eq. (11), the matrix TJJ is made more diagonally dominant, which improves its invertibility. Hence, there is no need for any “exact” solution from Eq. (10), which is already an approximation. The flow chart in Fig. 2-1 desc ribes the iterative update of a and D to approach the true profile starting from a uniform initial guess. Characterization of Spectra in Breast Tissue Physiologically, the breast is a turbid, light scattering medium combined with different shapes of absorbers, scatterers, fluorophores and anisotropic interfaces [5]. For biomedical optical imaging, the techniques aiding for cancer diagnosis depend primarily on detection for the aberrations of reflected, transmitted and emitted light that due to the physiological characterization or cellular grow th of cancer and the host response to the cancer. Thus, it is important for us to get to the bottom of physiologi cal and pathological factors of human breast disease that can influence optical diagnosis. There are two primary optical prope rties, the absorption coefficient a and the reduced scattering coefficient s which determine the propagation of the diffusive light through the breast. The diffuse optical tomography allows for measuring quantitative absorption and scattering distribution of tissues at any wavelength of interest, hence it is possible to use the spectra of which to obt ain tissue functional information, such as quantify typical values of hemoglobin concentration, oxygen saturation, water content, scattering power, scattering amplitude and so forth within the breast tissue. With a single integrating sphere technique, a few measurements of optical properties have been done in vitro on both normal and diseased brea st tissues shown in Table 2-1 [6]. The optical coefficients for the tissues with fibrocystic diseas, fibroadenomas and

PAGE 22

12 ductal carcinoma have no signifi cant difference with the normal tissues. However, due to the material of their experiments, which are tissue specimens of human breast, blood drainage caused by the surgical and pathological dissections of the breasts have greatly diminished the contributions of hemoglobin when measuring this group of optical properties. Table 2-1. The Optical Propert ies of Human Female Breast Tissue Type Optical Properties (nm) Glandular (3) Adipose (7) Fibrocystic (8) Fibro-adenoma (6) Carcinoma (9) 540 0.3580.156 0.2270.057 0.1640.006 0.4380.314 0.3070.099 700 0.0470.011 0.0700.008 0.0220.009 0.0520.0470.0450.012 Absorption Coefficient a (1mm) 900 0.0620.005 0.0750.008 0.0270.011 0.0720.0530.0500.015 540 2.440.58 1.030.19 2.170.33 1.110.30 1.900.51 700 1.420.30 0.860.13 1.340.19 0.720.17 1.180.31 Reduced Scattering Coefficient s (1mm) 900 0.990.20 0.790.11 0.950.17 0.530.14 0.890.26 Modified from Peters, V. G. et al. 1990. Optical propertie s of normal and diseased brea st tissues in the visible and near-infrared. Phys. Med. Biol. 35: 1317-1334 The numbers in parentheses give the number of tissue specimens examined for each tissue type. [6] Recent studies have demonstrated that the near-infrared photon migr ation is sensitive to several important tissue biochemical compositions; for instance, especially in the 400~600 nm range, hemoglobin is strong, and co uld even be seen as some hemoglobin contamination in some of the spectra. Thus dual-wavelength was introduced into DOT researches for quantifying the concentrations of de-oxy hemoglobin (reduced hemoglobin HbR ) and oxy-hemoglobin (2HbO ) in tissue. However, research suggests that the near-infrared light absorption in breast is produced by more th an just hemoglobin. In the intact living human breast, the most signi ficant light absorbers include hemoglobin, melanin, water, carotene, proteins and DNA. As reported, the absorption increases toward shorter wavelengths owing to the protein ab sorption, and toward longer wavelengths due

PAGE 23

13 to water absorption. That is to say, wate r and lipid, although they are just weak nearinfrared light absorbers, for their high abunda nce in the breast, could shadow a significant influence on absorption relative to hem oglobin in the wavelength of 900~1000nm range, especially water. The scattering properties of the tissue also yield im portant physiologic and pathological information. However, the dist ribution of near-infrare d light scattering in tissue is not well understood yet. Howeve r, multi-spectral ne ar-infrared light measurements of the reduced scattering coefficient s have still shown there are relationship between scatteri ng and wavelength. The scattering decreases with the increase of wavelength. At shorter waveleng ths below 600 nm, the scattering behavior is likely dominated by scattering from the pe riodicity and size of refractive index fluctuations of the collagen fibrils in the size range of 70 nm to hundreds of nm; whereas at longer wavelength behavior beyond 600 nm, th e scattering behavior is increasingly dominated by scattering from the larger cyli nders of collagen fibers (2-3 m diameter) composed of collagen fibrils. Similar to othe r biological tissues, all cellular and extracellular components within the breast tissue contri bute to light scattering; thus, besides the collagen fibers at the mi cro and macro scale, other tissu e components also contribute to the overall scattering, but th e collagen fibers probably play the main role in scattering behaviors. Thus, information about the type s of the scattering centers within certain lesion of breast tissue could be provided by spectral measurements to the scattering coefficient. Recent advances in our experimental studi es have provided reconstruction datum and images with separate a and s and phantom experiments have presented solid

PAGE 24

14 evidence that the reconstruction value has no cl ear differences with the practical designed value. In this work, the principal near-infra red light absorbers with in the breast tissue are assumed to be de-oxy hemoglobin (HbR ), oxy-hemoglobin (2HbO) and water (2HO). For the scattering coefficient s we made use of some a priori information, that is, the introduction of th e simple power-law which is well known in the near-infrared light theory. As a result, the scattering am plitude and the scattering power were reported in each experiment.

PAGE 25

15 CHAPTER 3 MATERIALS AND METHODS General Approach The chromophore concentration on the absorb ance of near-infrared light depends on acl, where is the molar extinction coefficient (11 M cm ), c is the concentration of chromophore c (1 M L ), and l is the photon path length (cm). the path length l is increased by scatte ring and is not known a priori. As reported, the absorption coefficients translate into tissue chrom ophore concentrations based on the equation 2.303ac (12) where the factor of 2.303 originates from th e base conversion between the logarithm for absorbance and the natural logarithm for a [7, 8, 9]. We assume that the chro mophores contributing to a in the human breast tissues are principally de-oxy hemoglobin (HbR ), oxy-hemoglobin (2HbO ) and water (2HO) (Cope 1991; Sevick et al. 1991) Thus, the concentrations of components in the tissue we need to determine for study includes th e concentration of de-oxy hemoglobin Hb (in units of M ), oxy-hemoglobin 2HbO (in units of M ) and water 2HO (in units of percentage) in the tissue. The re spective extinction coefficient chrom for a given chromophore at wavelength could be obtained from literature values. In the experiments, as the absorption coefficients a are measured from 3 to 10 wavelengths optionally [10, 11], we could use at least three equations to determine th e three unknown

PAGE 26

16 quantities of the functio nal concentrations of Hb, 2HbO, and 2HO. The extinction coefficients of all the wavelengths are listed in the following table [12]: Table 3-1. The Extinction Coefficients of All the Wavelengths in the Experiments (nm) 2HbO(11 M cm)Hb(11 M cm ) Optical Absorption of Water Compendium (1cm) 673 287.8 2668.24 0.00478 690 276 2051.96 0.005535 733 403.2 1102.2 0.023144 775 683.2 1188.28 0.027565 808 856 723.52 0.021773 840 1022 692.36 0.039494 915 1219.8 778.22 0.076 922 1225.6 777.04 0.09268 965 1175.6 484.34 0.470099 785 735.4 977.04 0.025933 808 856 723.52 0.021773 830 974 693.04 0.0320205 The scattering spectrum of tissue yields in formation on the nature of the scattering particles. In general, 's is the sum of contributions fr om the various tissue scatterers. Unfortunately, the detailed information about these individual scatterers was not well understood yet. Since it has been observed that the reduced scattering coefficient has a general trend to decrease as the wavelength increased, with some a priori information, the simple power-law dependence were employe d to fit the near-inf rared light scattering in tissue [7, 8]: 'SP sA (13) where A is the arbitrary model paramete rs for amplitude (a constant), is he wavelength (in nanometers), and SP, the magnitude of the expo nent, is the scatter power. As known, the value of scattering power SP increases significantly as the decreasing of the scattering center size, combined with the optical wavelength. In the case of Rayleigh

PAGE 27

17 scattering (d where d is the scattering center size, and is the responsible wavelength), 4SP is well established. As scattering objects become larger size, the scatter power SP decreases to approximately 1 for larger Mie-like scatters (d ); and the value of SP decreases to zero when d [8, 13]. Instrument and Calibration System Setup There are two imaging systems used for the clinic experiments are automated multichannel frequency-domain systems, both of which employ multiple diode lasers that provide visible and near-infrared light. The DOT system setups are schematically shown in the Following Figure. One could provide ne ar-infrared laser at 3 different wavelengths (785, 808, and 830nm), and the other at 10 di fferent wavelengths (638, 673, 690, 733, 775, 808, 840, 915, 922 and 965nm). The ring of the instruments holds different layers of fibre bundles, and each of those layers ha s 16 source and 16 detection fibre bundles alternately distributed by turns. The radi o-frequency intensity-modulated near-infrared beams are transmitted to the optical switch, and sequentially be passed to the source probes that gently touched the surface of the experiment mate rials or clinic human breasts. The diffused light collected by the detection fi bre bundles is then sens ed by the detection Units, which convert the light intensity into voltage signals, which are collected by the computer through the data acquisition board. We use the measured data for absorption and scattering images through our reconstruction algorithm.

PAGE 28

18 Figure 3-1. Schematic of 2 experimental syst ems. a. Schematic of the experimental system with 10 wavelengths b. Schematic of the experimental system with 3 wavelengths Detection fibre bundles Source fibre bundles DC motor CCDs Computer 10 Optical Switch 10 Lasers Laser Current Detection Units Data Acquisition Boards a. b

PAGE 29

19 System Calibration Method It has been demonstrated that the system calibration could provide helpful aid to the quantitative reconstructions of both absorption a nd scattering coefficients in turbid media [14]. In clinical experiments, the diameters of breasts various from person to person, a data base for calibration has be tter to be established. For 2D imaging experiments, the calibration procedure could be de scribed as following approaches: (1) Make a group of homogeneous phantoms with different diameters of interest for data base. (2) Perform experiments individually with the homogeneous phantoms and collect the measured data ijD respectively for each phantom. (3) For each set of measured dataijD, find the respective initial values of absorption coefficient a reduced scattering coefficient s and the boundary conditions coefficient (4) Generate a 2D finite element mesh with the same diameter for each phantom. Using a unit source intensity for the 16 illumi nated positions, simulate the 2D photon propagation with the initial values of the optical properties and the boundary conditions coefficient obtained in the former approach, then a new set of data ijD could be generated. (5) Obtain a factor matrix ijfusing the following equation *, i, j=1,2,,16ijijijfDD (14) The factor matrix ijf could be added into the calibration data base. For experimental data of whether heterogeneous phantom or clinical breast, multiply ijf from the homogeneous phantom with the nearest diameter by the experiment data set ij E a final calibrated data set ij E could be obtained, and used for further image reconstruction *, i, j=1,2,,16ijijijEfE (15)

PAGE 30

20 Phantom Experiments for System Calibration Phantom is used as an object to make in imitation of biological tissues in terms of absorption and scattering coefficients. In our study, the phantom materials employ composition of Intralipid as the scatterer and I ndia Ink as the absorber, as the Intralipid is an aqueous suspension consist of glycerin, lecithin, soybean and water. A boiled agar powder solution in a concentration of 2% is chosen as the hardener to solidify the aqueous mixture of Intralipid and India Ink, taking advantage of its non-absorption and low-turbidity. Considering the various size of human br east, homogeneous phantoms in different diameters (60mm, 70mm, 90mm 100mm, and 110mm) were prepared for the imaging experiments under the calibration method. The optical properties of phantoms in calibration experiments are the same:10.005 amm and 1'1.0 smm. Results of Calibration Experiments To examine the results of calibration data base, heterogeneous phantoms that have similar diameters to the homogeneous phant oms were employed during the examination for calibration experiments. The optical properties for the background of the heterogeneous phantoms are still the same with the homogeneous phantoms: 10.005 amm, and 1'1.0 smm ; and a single target was embedded in each homogenous background phantom, with positi on departs from the center. Thus, one 14mm diameter hole was cylindrical drilled in each homogenous background phantom for the inclusions of the target. As the relate d researches have shown that there is obvious difficulty to get quantitative optical imag es under conditions of very low absorption contrast, and this instance do not happen under conditions with just low scattering

PAGE 31

21 contrast or both low absorption and low sca ttering contrast. Consider ing this, the optical coefficients of the off-center targets for the heterogeneous phant oms in the calibration examine experiments were set as low absorption contrast only, that is 10.010 amm and 1'1.0 smm. The following figure depicts the ge ometrical configurations for the test cases of phantom diameter under study. Figure 3-2 Geometry of the phantom configuration Figure 3-3 present the reconstructed absorp tion images, both without calibration and with calibration, from phantom experiments in diameters 60mm and 70mm, respectively, take the imaging data from the same iteration of reconstructed data with the same filter times, and from the same wavelength (take 922nm in which wavelength it is almost the most difficult than most other wavele ngths for quantitative ab sorption images). From the Fig. 3-3, the results accords very good with our former conclusion. In other words, when the heterogeneous phantoms ha s the same diameter as the homogenous phantom that used for obta ining calibration data, the cal ibration method could improve the quantitative optical images in size, shape and value, even under the conditions of very low absorption contrast. s 27 R mm 1/2 R d

PAGE 32

22 a b c d Figure 3-3 Comparison between th e reconstructed absorption imag es of original data (left column) and of calibrated data (right column) for cases of 2:1 absorption contrast. a and c are the r econstructed absorption imag e from the original data in diameter 60mm (first row) and 70mm (second row) respectively; b and d are the corresponding absorption image from the data with calibration compare respectively to image a and c. In clinical experiments, the sizes of human breasts were not the same as the diameters we set for calibration phantom expe riment. Thus, another group of experiments were performed. As the space between di fferent homogenous pha ntoms is 10mm, we could use the calibration data from phantoms whose diameters border upon the diameter of the heterogeneous phantoms. Take the da ta from heterogeneous phantom (Absorption only, 10.007 amm and 1'1.0 smm ) with diameter of 100 dmm for example, we use calibration data from homogenous phantoms with diamet ers of 90mm, 100mm,

PAGE 33

23 and 110mm respectively. Figure 3.4 present th e corresponding recons tructed absorption images for 922 nm a b c d Figure 3-4 Reconstructed absorp tion images for heterogeneou s phantoms with or without calibration from different diameters in the case of 1.4:1 absorption contrast In the Figure 3-4 (where the Figure 3-4 a is the reconstructed absorption image for original data without any calibration, and Figure 3-4 b, 3-4 c, and 3-4 d are the reconstructed absorption im ages for data with calibra tion of 100mm, 90mm, and 110mm respectively), obviously the re sults are much better for cal ibrated data, even with calibration data from that of homogenous pha ntoms with different diameters. Comparing to the reconstructed image of 100mm-calibrated data, there are slight shifts for the target position for 90mm-calibrated and 110mm-calibrated data, and the sizes of target changes also. The maximum values of reconstructed data from 90mm-calibrated data and 110mm-

PAGE 34

24 calibrated data have errors around 5% compared to the maximum values of reconstructed data from 100mm-calibrated data, and the mi nimum values do not show visible errors among the three images. Since 10mm is the larg est distance in the calibration data base, we can conclude that the errors of the valu es are acceptable in the experiments, and the calibration data base we obtained is suitable for most clinical case. Clinical Experiments Over the past years, more than 100 voluntee rs, healthy or having pathological lesions in their breasts, ranging from 30 years old to 80, have been enrolled in this study of optical clinical human female breast experime nts. All these cases co uld be divided into two sets: patients in the first set have st rong evidence of abnormality, and the patients grouped in the other set have mamm ograms with unclear significance. After each patient was informed about this experiment, the patients would undergo the procedure lasting about ha lf an hour. The ring holding the optical fibre bundles was gently attached to examine the breast (only one breast) under the lesion position based on guidance or suggestion from professional doc tors, without any di scomfort or even significant pressure on the breast. When starte d, the breast would be illuminated by laser beams from a series of source probes, and at the same time, the de tectors from multiple positions around the breast would collect the diffused light transmitted from the breast tissue. After these datum collections, our r econstruction algorithm could be applied for further study. In this paper we present representative cas es from the selected subset of abnormal cases from 2 groups of patients (from 3 wa velengths to 10 wavelengths selectively, examined by two different system respectivel y), generate breast images with comparison to the mammograms, obtain absorption and reduced scattering coefficients with

PAGE 35

25 calibration methods, and finally gain and an alyze the optical pr operties and further functional information rev ealed by chromophore concentration from absorption coefficient a and scattering properties from scattering coefficient s The diseased human female breast could be divided into two categories as benign and cancer. The benign breast tumor includes Fibroadenomas, Fibrocustic Disease (Cyst) and miscellaneous lesions such as lipomas, blunt trauma, mastitis tissue and even ruptured or leaky silicone implants. These beni gn lesions are distinct pathologic entities, but for detection and diagnosis, they can be and usually are intermixed within the same breast. Thus, we selected clinical ca ses listed in the following table. Table 3-2 The Number of Lesions Sele cted for Experiment and Analysis Benign Lesions Cancer Lesions Cyst Calcification Nodules OMH 4 4 3 14 Greenville 7 2 3 4 The data obtained in the OMH (Oconee Memo rial Hospital) were collected for 10, 7 or 5 wavelengths, considering the size of the breast; and the data obt ained in the hospital of Greenville were collected for 2 or 3 wavelengths. Fitting Consideration Absorption Fitting With the measured absorption coefficients a and relative absorption spectrum, a weighted least-squares problem is put forw ard for recovering the concentrations of absorbers in breast tissues The dependence equations

PAGE 36

26 1111 1 1()()() ()()()am annmnmc c (17) could be written in a more general form a E c (18) As a is a vector containing the measured a values for N wavelengths, and c is the vector containing the concentrations of M different chromophores interested in the study. And E is the relative extinction coefficients NM matrix, where the traditional literature molar extinction coefficients should be convert into extin ction coefficients by multiply a 2.303). Hence, a general solution for this matrix problem could be expressed as 1 TT acEEE (19) To minimize calculation errors, normalizi ng scheme were employed to balance the variation of elements in the extinction coefficients matrix E by normalize columns of extinction coefficients for each chromophore with their respective maximum. Assuming 1,2,i M im are the maximum value for extinction coefficient (), 1,2,,ijjn the normalized extinction coefficient matrix convert into 111 1 1 111 ()() 11 ()()m m nmn mMM E MM (20)

PAGE 37

27 Then 1,TT acEEE and finally 1ii icc M (21) For these cases we report four hemoglobin parameters: Hb, 2HbO, total hemoglobin concentration HbT, and the hemoglobin saturation 2 tSO where 2HbTHbHbO (22a) 2 2 2 tHbO SO HbHbO (22b) Scattering Fitting The wavelength-dependent tissue reduced-scatt ering coefficient is assumed to take on this simplified Mie-scattering form ()SP sA as described in the beginning of this chapter. A and SP are related to the size, the index of refraction, and the concentration of scatterers in the tissue as well as the index of refraction of the surrounding medium. This Mie-sc attering form is judged as a robust nonlinear form, and is transformed into a linear form ln()lnlnsASP (23) For every two wavelengths i and j we could obtain a relative lnijA and ijSP, hence, the final lnA and SP were gained as the averag e value of a group of relative lnijA and ijSP 1 111(1) expln, as '2nn ij ijinn AAn n (24a) 1 111 'nn ij ijiSPSP n (24b) where n is the number of measured wavelengths.

PAGE 38

28 CHAPTER 4 RESULTS AND DISCUSSIONS Tumor Result-Images for Selected Examples All these volunteers that were selected for our studies including benign and cancer patients, have 41 abnormal lesions, which could be divided into 4 prim ary kinds of breast tumors: 18 benign nodules, 6 calcifications, 6 cy sts, and 11 cancers. All lesions were examined by professional physician and show ed clear signs on mammography or have distinct evidences on other examination reports. Figure 41 shows a rough relationship between the clinic breast front views (which was used for the breast examination reports) and our 2D optical images. In the following se ction, we will discuss some selected cases (in Fig 4.2 – 4.5, we present one selected pa tient for each disease ca tegory), in order to gain a general view before fu rther quantitative analysis. Nodule This patient is a 43-year old white fema le with two smooth lobulated solid nodules presented in the medial aspect of the left br east at the 9:30 o’clock. The more anterior nodule measures 5.8mm 4.8mm 8.3mm, and the mid left breast nodule measures 6.9mm 3.6mm 9.1mm. These two nodules lasted fo r at least six months before our optical imaging, and were recommended to fo llow up examinations in another six months to access stability. The optical image experiments were done by the 10-wavelenghsystem, and 64 sources and 64 detectors were di stributed uniformly fo r four planes along the surface of patient’s left breast at the lesi on position (the respective radii of the four

PAGE 39

29 layers that attach the breast are: 146.42 rmm 244.06 rmm 342.18 rmm and 440.30 rmm ; 16 sources and 16 detectors at each plane). As layer 1 (146.42 rmm ) showed the most similar results compare to the mammography and ul trasounds report, so we took the plane 1 for further studies. And the x-ray mammography (Figure 4-2a) and our optical imaging results (Figure 4-2b – 4-2e) are presented as figures 4-2. Figure 4-1 Corresponding position from report to optical 2D images. a The Front View on Breast Examination Report. b The 2D Image View for Experiment Results Right Left 12:00 12:00 3:00 3:00 6:00 6:00 9:00 9:00 12:00 3:00 6:00 9:00 a Value b

PAGE 40

30 Figure 4-2 The X-ray mammography, reconstr ucted optical propert ies and functional information images of a selected nod ule case. a. The X-ray Mammography for this patient. b. The Respective Absorption Coefficients Images from Wavelength 638nm to 922nm. c The Re solved Chromophore Concentration Images from Absorption Coefficients d Respective Reduced Scattering Coefficients Images from Wavele ngth 638nm to 922nm. e Scattering Amplitude and Scattering Power Images c 2HbO Hb Water (%) HbT 2 tSO (%) b 638nm 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm L_CC L_MLO a

PAGE 41

31 Figure 4-2 Continued. Hence, the two disease lesions near th e 9:30 position were distinct at the corresponding position of optical images fo r both absorption and reduced scattering coefficients from the shortest wavele ngth 638nm to 840nm. Unfortunately, for the longest three wavelengths, we could not obtain rec onstructed images with lesions legible for further study, due to the weaker si gnals. Therefore, we use those qualified reconstructed datum for resolving chromophor e concentrations, scattering amplitude, and scattering power. Those two tumors were dist inct around 9:00 oÂ’clock position in images for de-oxy hemoglobin concentration, oxy-hemo globin concentration, and water content: all the functional information of the two tu mors shows an increase compared to the 638nm 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm d Scatt Ampl. Scatt Power e

PAGE 42

32 surrounding tissue. Total he moglobin concentration HbT, and the hemoglobin oxygen saturation 2 tSO were also gained. Only the HbT shows increases at the tumors’ positions, but at the corresponding position, no tumor emerged in shape. For resolved scattering amplitude and power images, tumo rs are visible, showing increases for scattering amplitude and decreases for scatte ring power at their respective positions. Calcification This patient was a 65-year-old white fe male with the presence of coarse calcifications which were cluste red together in the upper outer right breast at 10 o’clock position. No underlying soft tissu e component is apparent on the spot compression views. In addition to the ultrasound, there is a fairly well ci rcumscribed slightly lobular hypoechoic focus associated with the calcifi cations, which measures approximately 5mm 8mm in size and could represent fat necrosis. The optical image experiments were also done by the 10-wavelengh-system, with the ra dii for three planes along the surface of patient’s right breast at the lesion position are: 136.43 rmm 234.07 rmm and 332.19 rmm (the probes for the 4th plane was not touched fully on the breast because of its own shape). Layer 1 (136.43 rmm ) showed the best results compared to the mammography and ultrasounds report. And th e x-ray mammography (Figure 4-3a) and our optical imaging results (Figure 4-3b – 4-3e) are presented as Figures 4-3.

PAGE 43

33 c Figure 4-3 The X-ray mammography, reconstr ucted optical propert ies and functional information images of a selected calcification case. a The X-ray Mammography for this patient. b The Respective Absorption Coefficients Images from Wavelength 638nm to 922n m. c The Resolved Chromophore Concentration Images from Absorption Coefficients. d Respective Reduced Scattering Coefficients Images from Wavelength 638nm to 922nm. e Scattering Amplitude and Scattering Power Images 638nm 673nm 690nm 733nm 775nm 808n m 840n m 915n m 922n m b R_CC R_ML a

PAGE 44

34 Figure 4-3. Continued. Still, the absorption and reduced scattering images were vivid, and the lesion position is in accordance with the examination re ports, except the images for the longest 3 wavelengths. Clearly, the calcifi cation lesion at 10 oÂ’clock pos ition is apparent on almost all the resolved functional coefficients images except for the concentration of 2 tSO. Remarkably, the nodule which could be fat necrosis was not apparent on mammograms and only shadowed on the ultrasound examinati on, but did appear cl early on our optical images for most wave lengths in terms of both absorption an d reduced scattering coefficients. Also, the resolved images for chromophore conc entration (except 2 tSO), scattering amplitude and scat tering power, shadowed this nodule near the corresponding 638nm 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm d Scatt Ampl. Scatt Power e

PAGE 45

35 location. Both the calcifica tion and nodule lesion present in silhouette, with increases in chromophore concentrations and scattering amp litude, and decreases in scattering power. Cyst This patient was a 43-year-old female dur ing our optical breast experiments. The reports clinical recorded th at there was a persistent 2. 3cm diameter smooth round nodular density, which was demonstrated to be a beni gn cyst. The physician i ndicated the cyst to be located at the 12 oÂ’clock position. A m oderate to large amount of dense residual fibroglandular tissue within the left breast was also present on the mammography. A small asymmetric oval area of density could be s een in the inferior half of the left breast on the MLO view; however, ultrasound examin ation suggested no dominant masses in the region of the palpable abnormality. To expl ain this, her ultrasound report also pointed out that the density of the breast was su ch that it might decrease sensitivity of mammography for detection of malignancy. The 10-wavelengh-system again were employed for the corresponding optical imaging experiments, performing four-plane measurements for the lesion position, and the respective radii are: 149.04 rmm 246.68 rmm 344.80 rmm and 442.92 rmm and the fourth layer were selected out for further researches. We could see visible tumor at the 12:00 oÂ’clock position in rec onstructed absorption and reduced scattering coefficient images for the middle region of wavelengths. For resolved functional coefficients, there are corresponding increases for this tumor at the respective position in images fo r de-oxy hemoglobin concentration Hb, oxyhemoglobin concentration 2HbO, water content, total hemoglobin concentration HbT and scattering power; however, no significant de creases or increases were present at the

PAGE 46

36 diseased lesion in the image of scattering am plitude, and neither in the oxygen saturation 2 tSO image. Interestingly, those tissue density that recorded in her mammography reports, do increase the value contrast to the cyst value in the resolved chromophore concentration images, comparing to the im ages of absorption and reduced scattering coefficients, which made them more similar to the normal background in the chromophore concentration images. Figure 4-4 The X-ray mammography, reconstr ucted optical propert ies and functional information images of a selected cyst case. a The X-ray Mammography for this patient b The Respective Absorption Coefficients Images from Wavelength 673nm to 922nm c The Re solved Chromophore Concentration Images from Absorption Coefficients c Respective Reduced Scattering Coefficients Images from Wave length 673nm to 922nm e Scattering Amplitude and Scattering Power Images L_CC L_MLO a

PAGE 47

37 c Figure 4-4 Continued 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm b 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm d

PAGE 48

38 Figure 4-4 Continued Cancer This patient was a 62-year-old white fema le with two cancer masses diagnosed with the needle biopsy report. Afte r our optical image experiment, the patient had surgery on her breast and removing the lesions. In he r screening mammography reports, these two cancer lesions were reported as two in creasing microcalcifications, one in the inferomedial quadrant near the dominant mass a nd the other in the central aspect of the mid breast. A few other small clusters of microc alcifications are seen laterally in the right breast, and more likely to be related to a be nign process such as sclerosing adenosis or cystic hyperplasia. In addition, there was a small to moderate amount of dense residual fibro-glandular tissue in the breast. This experiment was performed on the the 10-wavelengh-system, with only two layers of source-detector pr obes touch patientÂ’s breast surf ace at the lesi on position and the respective radii are: 165.60 rmm and 263.24 rmm and the experiment datum collected in the second plan e showed better results. Here, the two cancer tumors were present at the correct positi ons according to the surgery reports for reconstructed absorpti on images, but were not as clear in the Scatt Ampl. Scatt Power e

PAGE 49

39 reconstructed reduced scatteri ng images. After analysis, the two cancer tumors were more clear-cut in silhouette as in de-oxy hemoglobin concentration Hb, oxy-hemoglobin concentration 2HbO, water content, total hemoglobin concentration HbT, scattering amplitude and scattering power. In this cas e, corresponding increases were present at images for Hb, 2HbO, HbT, and scattering power; and d ecreases were present at images for water content and scattering am plitude. In addition, decreases in oxygen saturation 2 tSO could also be observed at the cance r tumors lesions, although the margins were not so fairly well circum scribed. Again, the uncertain ti ssues that have low contrasts compared to the cancer tumorsÂ’ values in the absorption and reduced scattering coefficient images, do increase their valu e contrast in the resolved chromophore concentration images. Figure 4-5 The X-ray mammography, reconstr ucted optical propert ies and functional information images of a selected can cer case. a The Respective Absorption Coefficients Images from Wavele ngth 638nm to 922nm b The Resolved Chromophore Concentration Images from Absorption Coefficients c Respective Reduced Scattering Coefficien ts Images from Wavelength 638nm to 922nm d Scattering Amplitude and Scattering Power Images 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm a

PAGE 50

40 Figure 4-5 Continued b 2HbO Hb Water (%) HbT 2 tSO (%) c 673nm 690nm 733nm 775nm 808nm 840nm 915nm 922nm Scatt Ampl. Scatt Power d

PAGE 51

41 Discussions In this section, we will discuss the functional properties we obtained. Different function parameters were illustrated in correl ation plots and compared and the major part of our discussion is on hemoglobin concen tration. Followed a general view on the selected cases we presented and the correlation plots of 2[][]HbHbO total hemoglobin concentration HbT were investigated with age and oxygen saturation 2StO in order to evaluate the feasibility of separating diffe rent breast diseases. Water content and scattering power were also briefly an alyzed with problems presented. Hemoglobin Concentration Represented cases Before quantitative analysis, we obtained some general results that emerged from these selected cases we present above for di fferent breast disease categories. All subjects with confirmed pathologic abnormalities produr ed respective localized increases in both de-oxy hemoglobin concentration Hb and oxy-hemoglobin concentration 2HbO: (a) Nodule case (the two nodules combined at 9:30 oÂ’clock): the resolved images recorded them as: Nodule 1 has a mean de-oxy hemoglobin concentration Hb for about 12 M (peaked at approximate 16 M ), and the mean oxy-hemoglobin concentration 2HbO for about 45 M (peaked at approximate 59 M ); Nodule 2 has a mean de-oxy hemoglobin concentration Hb for about 10 M (peaked at approximate 14 M ), and the mean oxy-hemoglobin concentration 2HbO for about 30 M (peaked at approximate 40 M ). Thus, the relative mean total hemoglobin HbT is about 55 M (peaked at approximate 70 M ) and 43 M ( peaked at approximate 50 M )

PAGE 52

42 respectively. No decreases fo r the hemoglobin oxygen saturation 2 tSOwere observed at the corresponding location for these two nodules. The additional nodule lesion in the calcification case had a mean value about 10 M for Hb(peaked at approximate 13 M ) and 47 M for 2HbO(peaked at approximate 60 M ); hence its relative average HbT is about 56 M ( peaked at approximate 71 M ), with no clear decreases in 2 tSO were observed at the corresponding location. (b) Calcification Case (calci fication located at 10 oÂ’cloc k position): This lesion was imaged with a mean value 9 M in Hb (peaked at approximate 11 M ) and 45 M for 2HbO(peaked at approximate 58 M ), which suggested the mean total hemoglobin concentration HbT was 54 M (peaked at approximate 68 M ). Again, no decreases in the hemoglobin oxygen saturation 2 tSO were observed. (c) Fibrocustic Disease (Cyst) Case: (a simple cyst located at approximate 12:00 oÂ’clock position): The resolved images indi cate a mean de-oxy hemoglobin concentration Hb of 13 M (peaked at approximate18 M ) and a mean oxy-hemoglobin concentration 2HbO of 19 M (peaked at approximate 40 M ). Therefore, the mean total hemoglobin concentration HbT was close to 31 M (peaked at approximate 49 M ), and no decreases for corresponding location of 2 tSO similar to the other cases. (d) Cancer Cases (Two cancer tumors): The mean values of Hb for tumor 1 was 20 M (peaked at approximate 27 M ), and the relative 2HbO was 51 M (peaked at approximate 62 M ); similarly the mean values of Hb for tumor 2 was 15 M (peaked at approximate 17 M ), and the relative 2HbO was 45 M (peaked at approximate

PAGE 53

43 51 M ). Their corresponding total hemoglobin concentrations HbT averaged at approximate 70 M and 60 M and peaked at about 87 M and 68 M respectively. In contrast to the benign cases, legible decreases c ould be recognized in the corresponding locations of the hemoglobin oxygen saturation 2 tSO image, although it was not clear-cut. At this stage of our clinical investigation, it is obvious to note that the abnormal tissues have higher hemoglobin concentra tion comparing to their surrounding normal background; and these increases should correlate with the blood vessel density. The two cancer tumors presented the highest hemoglobin concentration, whereas the reflected cyst case presented the lowest value. It appeared that the nodules and the calcifications in the cases we presented did not show distin ct differences in the total hemoglobin concentration, nor did thei r individual de-oxy hemogl obin concentrations and oxyhemoglobin concentrations. Thus, questions a bout the possibility on diagnose aid from the hemoglobin resolution remained to be answer, and a furthermore approach then started. 41 Cases Total hemoglobin concentration In Figure 4-6, we presented all cases for de-oxy hemoglobin concentration Hband oxy-hemoglobin concentration 2HbO from reported diseased lesions (as (a) illustrates the mean concentration of the diseased le sions, and (b) illustra ted their corresponding peak values), slight gap emerged as differe ntiation the malignant cancer tumors from benign tumors accord with the selected cases we presented above. As the malignant cancer tumors showed high values in both Hb and 2HbO, total hemoglobin

PAGE 54

44 concentration (average values) were calcula ted and graphed in Fi gure 4-7 with patient ages. Three patients did not repeat their ages and were not include d in the graph. The average total hemoglobin concentration of the corresponding diseased lesions were approximate 35 M (calcification), 80 M (nodule), and 5 M (lymph node nodule) respectively. a b Figure 4-6 Graph illustrate s the quantitative value fo r the raw de-oxy hemoglobin concentration Hband oxy-hemoglobin concentration 2HbO from reported diseased lesions, a mean values from the lesions, and b p eak value. For the special cases that illustrated lowe r concentration comparing to their surrounding tissues, we picked out the corresponding minimum values.

PAGE 55

45 Figure 4-7 Mean hemoglobin concentra tions of reported diseased lesion Physiologically, it has been demonstrated that most miscellaneous lesions (nodule cases, including blunt trauma, mastitis and so forth), usually stimulate increased blood flow into the very tender, firm affected area [5]. Another familiar tumor always reported as miscellaneous lesions, lipomas, is a benign fat tumors made up of mature adipose cells. Lipomas might have simila r absorption and scattering characteristics as the surrounding fat, but should tend to have more blood vessels than th em; however, its vascularity may not be sufficient enough to differentiate the tu mors from the normal tissues. By and large, most miscellaneous lesions should reflect higher hemoglobi n concentrations due to a more sufficient localized blood flow; even in the lipomas cases, though they might not have distinct higher hemoglobin concentratio ns than normal tissue (but should be no less than that of normal tissues), their corres ponding hemoglobin concentrations should help differentiate them as they are always co rrelated with plenty of surrounding fat which have extremely low hemoglobin concentrations Not only the typical cases we presented

PAGE 56

46 out here, but all the 17 miscellaneous lesions we investigated, the nodule lesions could be estimated with higher hemoglobi n concentrations comparing to its surrounding tissues except two: one was reported as indistinct abnormalities of fibroglandular, and the other one was at last demonstrated to be clusters of lymph node (a small bean-shaped structures that are found throughout the body, acting for filt er substances in lymph and help fight infection and disease); both of which shoul d have lower vascular ity physiologically. The mean values for total hemoglobin concentra tions of normal breasts usually ranged from greater than 20 M to less than 50 M as K. D. Paulsen et al investigated in clinical cases [15, 16, 17], the hemoglobin concentr ations for our nodule lesions were mostly more than or approximate at 50 M in average values (11 cases), 4 cases approximate in the normal ranges, and the indistin ct fibroglandular (approximate 16 M in average) and distinct lymph node (approximate 5 M in average) referred just now has lower total hemoglobin concentrations than normal tissue as showed in its corresponding image. The results accorded well with what the Physio logical and Pathology theories suggested. The fibrocystic disease or fi brocystic change of breast ti ssue covers a broad spectrum of localized or diffuse stomal and glandular alterations that can produce diffuse masses, distinct tumors and/or non-palpable lesions [5]. The cases we selected for investigation have been diagnosed with large or small li quid-filled cysts formed by the dilatation of ducts. Customarily, cysts are a very common br east disease and rarely associated with cancer. Unlike cancerous tumors which are solid, cysts are masses that generally containing dark brown fluid in the breast. Th e fluid obtained from cysts usually contains degenerated cells, secretions and blood that may or may not have distinctive absorption characteristics. Theoretically, the fluid-fill ed cysts should reflect lower total hemoglobin

PAGE 57

47 concentration than the solid tumors for the reason of containing plenty of non-bloody fluid. Comparing to what we obtained th at illustrated in graph of Figure 4-6, by appearances, most cysts di d exhibit obvious lower HbT than most other diseased lesion categories. The third category we focused on, calcific ations, are tiny flecks of calcium, like grains of salt, in the soft tissue of the breas t that can sometimes indicate the presence of an early breast cancer [5]. Calcifications usually are not detected by ultrasound, but on the other hand, they could appear on a mammogram. Using our technique, we could image calcifications lesions, and differentiate them in relative chromophore concentration images successfully. But when making comp arisons to other lesion categories in hemoglobin concentration, we could not differe ntiate benign calcific ations clearly from miscellaneous lesions and cancer cases. St udies reported that big calcifications (diagnosed as “macrocalcifications”) are not always associated with cancer. Groups of small calcifications huddled together, (diagnosed as “clusters of micr ocalcifications”) are usually combined to extra breast cell activity—m ost of which are non-cancerous extra cell growth. But sometimes clusters of mi crocalcifications, the tiny calcium deposits occuring in areas of early cancer could be regarded as the first x-ray evidence of cancers. Hence, the calcifications lesion should have higher hemoglobin concentration correlated with extra cell growth needing for a more su fficient localized blood flow, and some cases that have clusters of microcalcifications mi ght demonstrate some comparability to early cancer cases since the malignant tumors are usually associated with increased microcalcifications.

PAGE 58

48 The malignant tumors in breast are can cer that may be life-threatening [5]. Biologically, it has been demons trated that various growth factors and cytokines inheres in cancer cells that could stimulate host responses of inflammation, blood vessel proliferation (cancer related angiogenesis) and fibrosis (desmoplasia). Hen ce, it is easy to apprehend that the malignant cancer tumors could reflect extremely high hemoglobin concentrations comparing to most benign tumors. Most of the diseased lesions involved in th is investigation indi cated increased total hemoglobin concentration compared to th eir surrounding tissue except the two nodule cases we referred to before; and all the diseas ed tumors could be distinguished from its surrounding tissues from quantitative hemogl obin concentration images in accord with the reported position. It app ears that the optical mammogra phy could allow most of the physical masses that were present on mammogr aphy are detectable and distinguishable not only for the images of optical properties (a and s ), but also for the images of hemoglobin concentration in vivo. Quantitative ly, the results plotted in graph of Figure 46 indicated that the mean values of total hemoglobin concentration can secrete functional information of different diseased lesions The solid tumors (including solid nodules, calcifications and malignant cancer tumors ) generally indicate significantly higher hemoglobin concentration than the tumors that contains non -bloody fluid (i nclude cyst, fibro-glandular and lymph node cases); a nd the malignant cancer tumors generally shadowed significant higher hemoglobin con centration than most benign tumors except some cases that containing clusters of micr o-calcifications. However, not all the lesions have significantly higher/lower hemoglobin concentration than the whole breast tissue plane for imaging, as even the normal breast itself is not a homogeneous organ, illustrated

PAGE 59

49 in Figure 4-6 illustrated, we could detect that most of the non-bloody-liquid-contain tumors shadowed their corre sponding average hemoglobin co ncentration in the normal tissue range. On the other hand, for the solid tumor cases, the possibi lity for quantitative hemoglobin concentration aiding diagnosis intr oduced its existence; since most solid tumors distinguished themselves not onl y from their surround ing tissue in the corresponding images directly, but also sha dowed higher values compare to the normal tissues. It is also interesting to note that the malignant cancer tumors appeared to produce high hemoglobin concentration, as the hemogl obin concentration for most benign tumors are less than 60 M on average, while the malignant cancer tumors were usually greater than or at least approximate 60 M Hemoglobin concentration and oxygen saturation The total hemoglobin concen tration appears to be a promising avenue for a diagnostic aid by quantitative optical chrom ophore concentrations image, providing a deeper view from hemoglobin concentrations The total hemoglobin concentration is one factor correlated by de-oxy hemoglobin con centration and oxy-hemoglobin concentration, and the other important factor combined with them is the oxygen saturation. Theoretically, the oxygen saturati on which is related to the 2 p O of oxygen in tissue, could help in evaluating the hypoxia in tumo rs [17]. Evidence have been accumulated that hypoxia is usually involved in a vicious circle of a fundamental biologic mechanism of the malignant tumors. Hence, the hypoxia was regarded as one of th e critical factor for separating malignancies from benign tumors. As a result, in several cases of the malignant tumor exhibit smaller oxygen saturati on compared to healt hy tissue, but there are also cases for which the opposite holds tr ue. As known, results of cases performed by

PAGE 60

50 Fantini et al (1998), Tromberg et al (2000) and Chernomordik et al (2002) observed decreases in oxygen saturation of the tumor [18, 19, 20], whereas McBride et al (2002) and K. D. Paulsen et al (2005) each reported a case with no essential difference in oxygen saturation between tumor and healthy tissue [21, 22]. Meanwhile, for clinical cases, Dirk Grosenick et al (2004) recorded that blood oxyge n saturation enhanced in some cases and reduced in some other cas es under the processing of dual wavelength time-domain optical mammography [23, 24]. From our oxygen saturation images, we coul d distinguish 4 malignant cancer tumors from its surrounding tissues w ith obvious oxygen saturation d ecreases; and 1 malignant cancer tumors have slightly decreases comp aring to its surrounding health tissues in oxygen saturation, but the difference is not so significant to define the tumor shape; the other 6 malignant cancer tumors did not s how differences in their oxygen saturation images. For the benign tumors, we observed 2 unclear decreases in the tumor position (1 nodule and 1calcification), 3 cl ear decreases (1 calcificat ion and 2 nodule including the fibroglandular nodule case), and 1 increase oxygen saturation cy st case, whereas the other cases presented no evident differences between the diseased lesion and healthy tissue. Thus, our results seem to suggest that oxygen saturation decreases were more often visible in vivo for malignant cancer tumors (5 out of 11 lesions) and calc ification lesions which usually be regarded as signs for early cancer (2 out 6 cases). In Contrast, benign nodules usually do not differentiate obvious di fferences in oxygen saturation from their surrounding healthy breast tissue. For more qua ntitative analysis, as most tumors could not be distinguish out from normal tissues whereas their corresponding total hemoglobin concentration images did show distinct di fference from their surrounding tissue; hence,

PAGE 61

51 we picked out one point from each tumor, i. e. the point showed peak value for total hemoglobin concentration (for those tumors showed lower HbTthan their surrounding tissue, we picked out the point with minimu m value in them) and then calculated their corresponding oxygen saturations. These qua ntitative data were plotted in a 2HbTStO plane showed as Figure 4-7. Clearly, compar ed to Figure 4-6 just have one functional coefficient of HbT, the data in Figure 4-7 may be an improvement to distinguish the malignant cancer tumors from benign tumors. Ho wever, this approach fails to yield more information to distinguish fluid-filled cyst s than using total hem oglobin concentrations alone. Figure 4-8 Graph illustrates the quantit ative value for the maximum hemoglobin concentration and the co rresponding oxygen saturations in reported diseased lesion Water Content As above discussed, our model results of hemoglobin concentration could accord with the nature of tumors, not only visible but also quantitative for both malignant and

PAGE 62

52 benign cases. However, it may not be so prom ising when making quantitative analysis on the water content obtained by the absorption coefficients a Similar to what was presented by the select ed examples, most tumor cases could be visible and distinguish out at the accepta ble location in accord ance with what the mammography or ultrasound reports indicated. But the values of water content we gain were not reasonable enough to make further qua ntitative analysis: nega tive values (as the cancer cases we presented) or values exceed 100% (as the first nodule case we presented) existed in some cases. The water influence on the absorption coe fficients in breast tissue was not as significant as hemoglobin especi ally for light in short wave length range. Only in the long wavelength, particularly in the 900-1000n m range, did absorption by water present a stronger influence. The wavelength we c hoose for these experiments were between 638nm to 965nm, which were considered to be in the range of centr al wavelength region, and the tissue absorption is the lowest wh ich could provide the allowance for light penetration through a few centimeters of ti ssue. From the reconstructed absorption coefficient image, we could notice that we always obtain clear images in most short wavelength that influenced more by hemoglobi n. The experiment data for longest three wavelengths we considered most helpful in separate the water ch aracterization usually were too weak to gain qualify enough recons tructed optical coeffici ents. That might be the primary reason for the problems on the valu es of water contents. Besides, our model for chromophore concentration calculati on was a simplified model which does not account enough for the complexity of the breast tissue. Based on this model, the influence from other absorbers such as melanin, carot ene, proteins, DNA, vegetable oil and lipids

PAGE 63

53 were neglected, which might reflect uncertain factors on th e results. It has also been demonstrated that interactions of various chromophores with other tissue and cellular constitutents may cause spectral shifts in measurements taken in vivo when compared to those taken in vitro or as purified chromophores in solution. In addition, the tumor location and visible size may infl uence the accuracy of tumor optical properties. All these factors could be added into the uncertainty of results of the chromophores concentration. Scattering Spectra The scattering spectra were thought to provi de important physiological content. To best of our knowledge, we expect the colla gen-rich-tissue-rich-b reast would display a higher scatter power than the breasts with more adipose. Visibly, the image results we obtained for scattering power using 10-wavelength experi ment produced evidence to encourage the view that most of the diseased lesions could be dis tinguished clearly from the scattering power images (except 4 nodule and 1 cyst cases ), with lesion positions well accord with the pathological diagnoses, whereas the results from 3-wavelength were not as good. Figure 4-9 presents the mean value of scattering power w ith their mean total hemoglobin concentration and age. There were no evident gaps between different tumors in the HbTSP (scattering power) plane that were observed. Tromberg et al (2002) have reported that scattering power showed decrease with ag e in normal breast [7], in our tumor cases, the decreases combined with age were not significant. Moreover, images of scattering amplitude failed to illustrate anything.

PAGE 64

54 a b Figure 4-9 Graph illustrates the mean value of scattering power with the total hemoglobin concentration. a and the patient ag e b in reported diseased lesion. As the algorithm accounting for scattering amplitude and scattering power employed a non-linear to linear transfor mation and a linear to non-linear transformation; besides,

PAGE 65

55 there are more than 2 equations for solving the two scattering parameters, thus, the combined equations might be ill-conditioned. Any small inaccuracy in the measurements of optical coefficients might be enlarged while calculation. Hence, it leaves future work for algorithm improvements on this. Conclusion During this investigation, patients suspected of having different br east diseases, were imaged using a multi-wavelength optical scanning in the frequency domains. Optical properties were reconstructed in certain wa velengths ranging from 638nm to 965nm, and calculating the physiological functional inform ation such as hemoglobin concentration, water content, scattering amplitude a nd power, with the introduction of their corresponding spectra. Detectab ility of tumors for thes e physiological functional coefficients was recorded and evalua ted through comparison with breast x-ray mammography and ultrasound reports, or biopsy reports if suitable. For 41 diseased lesions we selected, all of them could be detected and distinguished by oxy-hemoglobin and de-oxy hemoglobin concen tration. Additional factors such as total hemoglobin concentration and oxygen sa turation were also obt ained and analyzed. The quantitative values of total hemoglobin concentration were in accord with the physiological and pathological nature of differe nt tumors. For most tumors we detected, total hemoglobin concentration was larger co mpared to its surrounding tissues (except for one lymph node case and one fibro-glandular case); and especially for malignant cancer tumors, evident increases could be observed mo re distinct than most of the benign tumors. Furthermore, when combined with the oxygen saturation, it seems pr omising that this optical image technique using continuous-wav e will allow us to distinguish malignant tumors from benign. From this study, we can explore approaches using mathematical

PAGE 66

56 classification from all these opt ical and physiological coefficien ts that we obtained, or a possibility of carcinomas diagnos is assistance could be found. Both water content and scattering power were obtained, and tumors were visible in positions in accordance with mammography an d ultrasound examination reported in the corresponding images, but deeper quantitative analysis is a remaining problem. Hence, algorithms accounting for each of these will re quire further improvements before they can be used.

PAGE 67

57 LIST OF REFERENCES 1. M. Cutler, “Transillumination as an aid in the diagnosis of breast lesions,” Surg. Gynecol. Obstet. 48 (1929) 721-728. 2. E. Carlsen, “Transillumination Light Scanning,” Diagnostic Imaging 4 (1982) 2834. 3. K. D. Paulsen and H. Jiang, “Spatiall y varying optical property reconstruction using a finite element diffusion equati on approximation,” Med. Phys. 22 (1995) 691-702. 4. Y. Xu, X. Gu, T. Khan, and H. Jiang, “Absorption and scattering images of heterogeneous scattering media can be si multaneously reconstructed by use of dc data,” Appl. Opt. 41 (2002) 5427-5437. 5. S. Thomsen and D. Tatman, “Physiological and Pathological Factors of Human Breast Disease That Can Influence Optical Diagnosis,” Annals of the New York Academy of Sciences 838 (1998) 171-193. 6. V. G. Peters, D. R. Wyman, M. S. Patters on, and G. L. Frank, “Optical properties of normal and diseased breast tissues in th e visible and near-infrared,” Phys. Med. Biol. 35 (1990) 1317-1334. 7. A. E. Cerussi, D. Jakubowski, N. Shah, F. Bevilacqua, R. Lanning, A. J. Berger, D. Hsiang, J. Butler, R. F. Holcombe, and B. J. Tromberg, “Spect roscopy enhances the information content of optical mammogr aphy,” J. Biomed. Opt. 7 (2002) 60–71. 8. B. W. Pogue, S.g Jiang, H. Dehghani, C. K ogel, S. Soho, S. Srinivasan, X. Song, Tor D. Tosteson, S. P. Poplack, and K. D. Paulsen, “Characterization of hemoglobin, water, and NIR scattering in breast tissue: analysis of intersubject variability and menstral cycle changes,” J. Biomed. Opt, 9 (2004) 541-522. 9. B. J. Tromberg, O. Coquoz, J. B. Fishkin, T. Pham, E. R. Anderson, J. Butler, M. Cahn, J. D. Gross, V. Venugopalan, and D. Pham, “Non-invasive measurements of breast tissue optical properti es using frequency-domain photon migration,” Phil. Trans. R. Soc. Lond. B. 352 (1997) 661-8. 10. A. Corlu, T. Durduran, R. Choe, M. Schweiger, E. M. Hillman, S. R. Arridge, and A. G. Yodh, “Uniqueness and wavele ngth optimization in continuous-wave multispectral diffuse optical tomogr aphy,” Opt. Lett. 28 (2003) 2339-2341.

PAGE 68

58 11. A. Corlu, R. Choe, T. Durduran, K. Lee, M. Schweiger, S. R. Arridge, E. M. Hillman, and A. G. Yodh, “Diffuse optic al tomography with spectral constraints and wavelength optimization,” Appl. Opt. 44 (2005) 2082-2093. 12. S. J. Matcher, M. Cope, and D. T. Delpy, “In vivo measurements of the wavelength dependence of tissue-scattering coeffici ents between 760 and 900 nm measured with time-resolved spectroscopy,” Appl. Opt. 36 (1997) 386-396. 13. A. E. Cerussi, A. J. Berger, F. Bevilacqua, N. Shah, D. Hakubowski, J. Butler, R. F. Holcombe, B. J. Tromberg, “Sources of ab sorption and scattering contrast for NearInfrared optical mammography,” Acad Radiol 8 (2001) 211-218. 14. C. Li and H. Jiang, “A calibration method in diffuse optical to mography,” J. Opt. A: Pure Appl. Opt. 6 (2004) 844-852. 15. B. W. Pogue, S. P. Poplack, T. O. McBride, W. A. Wells, O. K. S., U. L. Oserberg, and K. D. Paulsen, “Quantitative he moglobin tomography with diffuse nearinfrared spectroscopy: p ilot results in the breast ,” Radiology 218 (2001) 261-6. 16. S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, C. Kogel, S. Soh, J. J. Gibson, T. D. Tosteson, S. P. Poplack, and K. D. Paulsen, “Interpreting hemoglobin and water concentration, oxygen saturation a nd scattering measured in vivo by nearinfrared breast tomogra phy,” PNAS 100 (2003) 12349-12354. 17. M. Hockel, P. Baupel, “Tumor hypoxia: defi nitions and current clinical, biologic, and molecular aspects,” J. Natl. Cancer Inst. 93 (2001) 266-276. 18. K. T.Moesta, S. Fantini, H. Jess, S. Totkas, M. A. Franceschini, M. Kaschke, and P. M.Schlag, “Contrast features of breast can cer in frequency-domain laser scanning mammography,” J. Biomed. Opt. 3 (1998) 129–36. 19. B. J. Tromberg, N. Shah, R. Lanning, A. Cerussi, J. Espinoza, T. Pham, L. Svaasand, and J. Butler, “Non-invasive in vivo characterization of breast tumors using photon migration spectrosc opy,” Neoplasia 2 (2000) 26–40. 20. V. Chernomordik, D. W. Hattery, D. Gros enick, H. Wabnitz, H. Rinneberg, K. T. Moesta, P. M. Schlag, and A. Gandjbakhche “Quantification of optical properties of a breast tumor using random walk theo ry,” J. Biomed. Opt. 7 (2002) 80–7. 21. T. O.McBride, B. W. Pogue, S. Poplack, S. Soho, W. A. Wells, S. Jiang, U. L. Oserberg, and K. D. Paulsen, “Multisp ectral near-infrared tomography: a case study in compensating for water and lipid content in hemoglobin imaging of the breast,” J. Biomed. Opt. 7 (2002) 72–9. 22. S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, and K. D. Paulsen, “Spectrally constrained NIR tomography fo r breast imaging: simulati ons and clinical results,” SPIE 5693 (2005) 293-300.

PAGE 69

59 23. D. Grosenick, H. Wabnitz, K. T. Moesta, J. Mucke, M. Moller, C. Stroszczynski, J. Stossel, B. Wassermann, P. M. Schlag and H. Rinneberg, “Concentration and oxygen saturation of haemoglobin of 50 breas t tumours determined by time-domain optical mammography,” Phys. Med. Biol. 49 (2004) 1165-81. 24. D. Grosenick, K. T. Moesta, H. Wabnitz, J. Mucke, C. Stroszczynski, R. Macdonald, P. M. Schlag, and H. Rinneberg, “Time-domain optical mammography: initial clinical results on detection and characterization of breast tumors,” Appl. Opt. 42 (2003) 3170-3186.

PAGE 70

60 BIOGRAPHICAL SKETCH My name is Lin Chen. I was born on March 14, 1981, as a girl in Jiangjing, a little town of Chongqing, in southwest China. Af ter accomplishing high school with a brilliant record, I fought my way out of millions of candidates into Peking University, one of the most prestigious and selective universities in China. As an undergraduate in Peking University (1999.7.-2003.6.), I intended to ma jor in electronics engineering. During this period, I enrolled in some courses and proj ects on optics and signal processing, and I found myself enamored with the area, especial ly the optical imaging. After graduation, I pursued graduate school at my advisorÂ’s sugge stion, and then enrolled in the masterÂ’s program of Clemson University At the Physics and Astrono my Department in Clemson University, my advisor, Prof. Huabei Jiang, brought me into the research for biomedical optical imaging of human breast. Under his gu idance, I began to do clinical experiments and build mathematical models for tissue functional information reconstruction and analysis. In 2005, I transferred to the Univer sity of Florida with the whole Biomedical Optics Laboratory, and continued my masterÂ’s Program and researches in the Biomedical Engineering Department here. I enrolled in courses on anatomy and physiology, which helps a lot in my research. Now, I have ach ieved some approaches on the differentiation of malignant from benign breast lesions base d on functional diffuse optical tomography.