Breast Cancer Detection Based on Incremental Biochemical and Physiological Properties of Breast Cancers

Article (PDF Available)inAcademic Radiology 12(8):925-33 · September 2005with48 Reads
DOI: 10.1016/j.acra.2005.04.016 · Source: PubMed
Abstract
To demonstrate that near-infrared spectroscopy would achieve sufficient sensitivity and specificity in human breast cancer to reach ROC/AUC values in the 90s and yet to warn of the potential liabilities of introduction of a novel technology in this field. 116 subjects from two nations (44 were cancer-verified by biopsy and histopathology) were reviewed. NIR spectroscopy of total hemoglobin and its relative oxygenation were monitored in breast cancers and compared to their contralateral breast in a 2D nomogram for diagnostic evaluation. A novel handheld NIR breast cancer detector pad with a 3-wavelength LED and 8 detectors with 4 cm separation between source and detectors was placed on the subject's breast. The method is convenient, rapid, and safe and has achieved high patient compliance with minimal patient apprehension of compression, confinement, or radioactivity. The absorbance increments of the cancerous region are referred to the mirror image location on the contralateral breast. The two metrics are increased hemoglobin concentration due to angiogenesis and decreased hemoglobin saturation due to hypermetabolism of the cancer. The 2D nomogram display of these two metrics shows Zone 1 contains verified cancers and Zone 2 contains noncancers. ROC evaluation of the nomogram gives 95% AUC for the two sites, Philadelphia and Leipzig. A simple, economical breast cancer detector has achieved high patient compliance and a high ROC/AUC score for a population which involved a range of tumors down to and including those of 0.8-1 cm in diameter.
Breast Cancer Detection Based on Incremental
Biochemical and Physiological Properties
of Breast Cancers:
A Six-Year, Two-Site Study
1
Britton Chance, PhD, DSc, Shoko Nioka, MD, PhD, Jun Zhang, MS, Emily F. Conant, MD, Emily Hwang, MS,
Susanne Briest, MD, Susan G. Orel, MD, Mitchell D. Schnall, MD, PhD, Brian J. Czerniecki, MD, PhD
Rationale and Objectives. To demonstrate that near-infrared spectroscopy would achieve sufficient sensitivity and speci-
ficity in human breast cancer to reach ROC/AUC values in the 90s and yet to warn of the potential liabilities of introduc-
tion of a novel technology in this field.
Materials and Methods. 116 subjects from two nations (44 were cancer-verified by biopsy and histopathology) were re-
viewed. NIR spectroscopy of total hemoglobin and its relative oxygenation were monitored in breast cancers and com-
pared to their contralateral breast in a 2D nomogram for diagnostic evaluation. A novel handheld NIR breast cancer detec-
tor pad with a 3-wavelength LED and 8 detectors with 4 cm separation between source and detectors was placed on the
subject’s breast. The method is convenient, rapid, and safe and has achieved high patient compliance with minimal patient
apprehension of compression, confinement, or radioactivity.
Results. The absorbance increments of the cancerous region are referred to the mirror image location on the contralateral breast.
The two metrics are increased hemoglobin concentration due to angiogenesis and decreased hemoglobin saturation due to hyper-
metabolism of the cancer. The 2D nomogram display of these two metrics shows Zone 1 contains verified cancers and Zone 2
contains noncancers. ROC evaluation of the nomogram gives 95% AUC for the two sites, Philadelphia and Leipzig.
Conclusion. A simple, economical breast cancer detector has achieved high patient compliance and a high ROC/AUC
score for a population which involved a range of tumors down to and including those of 0.8 –1 cm in diameter.
Key Words. ROC/AUC; angiogenesis; hypermetabolism; breast cancer.
©
AUR, 2005
This article describes the results of incremental biochemi-
cal and physiological properties of breast cancers with a
multi-wavelength, multi-detector breast cancer device.
Whereas conventional studies emphasize incremental
structural features such as spiculation and lobulation for
diagnostics, little attention has been paid to incremental
biochemical properties except using
18
FDOG in PET,
which follows Otto Warburg’s hypothesis of the predomi-
nance of glycolysis in cancer (1,2). Here we follow a de-
rivative of the Warburg hypothesis emphasized by our-
selves (3), namely that cancer has a normal complement
of citric acid cycle and mitochondrial oxidative phosphor-
ylation capacity but insufficient oxygen is delivered to the
cancer, rendering it relatively hypoxic in spite of vigorous
angiogenesis (4). In order to explore this hypothesis, we
have developed a series of multi-wavelength, multi-detec-
Acad Radiol 2005; 12:925–933
1
From the University of Pennsylvania, Departments of Biochemistry and
Biophysics, 250 Anatomy/Chemistry Building, Philadelphia, PA 19104-6059
(B.C., S.N., J.Z., E.H.); Hospital of the University of Pennsylvania, Depart-
ment of Radiology, Philadelphia, PA 19104 (E.F.C., S.G.O., M.D.S.); Univer-
sity of Leipzig, Department of Gynecology and Obstetrics, Leipzig, Ger-
many (S.B.); University of Pennsylvania, Department of Surgery, Philadel-
phia, PA (B.J.C.). Received February 11, 2005; revision received April 28,
2005; revision accepted April 29. Supported by National Institutes of
Health, National Cancer Institute No. CA87137; Pennsylvania Department
of Health. Address correspondence to: B.C. e-mail:
chance@mail.med.upenn.edu
©
AUR, 2005
doi:10.1016/j.acra.2005.04.016
925
Original Investigations
tor NIR devices (5,6) and for the past 6 years have dedi-
cated one particular version of this development to breast
cancer detection.
The optical method was used as early as 1929 by M.
Cutler (7,8) to make shadow images of the breast (diaph-
anography) assuming that the cancer had different optical
properties from the normal tissue and that a “shadow-
gram” could be made. This was a sound theory, but the
differences are so small that the method was discredited
in clinical trials. However, the understanding of the dual
wavelength method (9 –12), the photon diffusion in tissue
(13) and the availability of monochromatic laser diodes,
LED’s and sensitive silicon diode detectors for the near
infrared (NIR) regions (680-900 nm) has made possible
quantification of the specific absorption of oxygenated
and deoxygenated hemoglobin and of the total amount of
hemoglobin within the breast cancer by reflectance spec-
troscopy (5,14).
Developing novel methods of detecting early, small
cancers is the goal of NIH’s cancer program (15), and
this goal requires not only methods of high sensitivity/
specificity in order to minimize undesirable false signals
but also to afford a convenient, rapid, safe method with
high patient compliance and minimal patient apprehen-
sions of compression, confinement, and radioactivity. The
development of NIR spectroscopy and imaging has pro-
ceeded rapidly since our early efforts (16) in this and
other laboratories (17) and its usefulness has been tested
in studies of the breast (18), brain (19), and muscle (20).
Brain and muscle studies permit activation protocols in
which the baseline is a low activity state for brain and a
resting state for muscle. Baseline problems are minimized
and incremental values of functional activation are ob-
tained, particularly in event-related protocols (20). In the
case of human breast tissue, functional activation is not
available and instead spatial differences must be em-
ployed. Localization and characterization of the incremen-
tal changes of optical properties with respect to position
changes seem necessary. Thus, the problem is significant
and absolute values of optical properties involve problems
of the physiological and biochemical baselines for distin-
guishing pathologies. In order to ameliorate that problem,
we began to gather breast cancer data as an incremental
study referenced to the mirror image location on the con-
tralateral cancer-free breast (21,22) or, if necessary, a
breast model (23,24).
In this investigation we used an NIR method which is
capable of rapidly acquiring data from the human breast
with a handheld puck (taking advantage of opto-electron-
ics miniaturization and retaining high quality functionality
such as in cell phone technology (25)) that measures rela-
tive increases of tumor hemoglobin concentration and
relative hemoglobin desaturation, all data being taken on
a relative basis using generally the mirror image site on
the contralateral breast, substantially mitigating the multi-
ple effects of variable demographic and structural features
of the human breast by measuring relative hemoglobin
concentration and oxygenation compared to a normal
breast within subjects. In analyzing relative hemoglobin
concentration against relative saturation, a nomogram dis-
play conveniently segregates verified cancers from cancer-
free breasts over a wide range of tumor sizes and types.
received operating characteristic/area under the ROC
curve (ROC/AUC) curves, positive predictive value
(PPV), negative predictive value (NPV) and other infor-
mation are displayed.
MATERIALS AND METHODS
Subjects
This study includes two clinical centers: the Abramson
Family Cancer Research Institute/Department of Radiol-
ogy of the Hospital of University of Pennsylvania (HUP),
and the Department of Gynecology of Leipzig University.
The population targeted at HUP is those who have come
for possible biopsy and for radiology. The second group
has come to the Breast Cancer Clinic at Leipzig. HUP
and Leipzig provided 24 and 20 cancer patients, respec-
tively, and 64 and 8 noncancer disease patients, respec-
tively. Informed consent was obtained from all subjects
participating in the study approved in the institutional
review boards (IRB) of both the University of Pennsylva-
nia and the University of Leipzig.
Apparatus: NIR Spectrometer
Apparatus.—We have used a continuous wave (CW)
near-infrared spectrometer (NIRS) (Fig. 1A, B). In the
center of the probe (Fig. 1B) is a 3-wavelength light-
emitting diode (LED). The LED intensity is low, 10 –15
mA. LEDs are not regulated by the FDA. This apparatus
was used with a manual gain control for the first three
years of the study and a digital gain control thereafter,
but the device is considered to be substantially equivalent
over the 6 years of study (27–29). The probe consisted of
one multi-wavelength LED as a light source and 8 silicon
diodes as detectors (Fig. 1). These 8 detectors surround
CHANCE ET AL
Academic Radiology, Vol 12, No 8, August 2005
926
the LED at 4 cm distance, so that 8 locations over a 9-cm
diameter area from a breast can be measured. The 4-cm
source-detector distance gives a banana-shaped pattern
revealed by Monte Carlo simulation (5) which demon-
strates a high probability of photons at a 4-cm depth in
which the diffusion pattern is 20 cm long on the aver-
age and gives a significant probability of tumor detection
at 4 or more cm depth, as has been verified by CT scans
in the case of human brain hematomas (30) and in trans-
abdominal detection of fetal brain saturation in utero (31)
in which the fetus was 5 cm deep in a variety of sub-
jects with varying layers of overlying tissue and most
recently in a transthoracic determination of myocardial
saturation (31,32). The nomogram display of Figure 2
below illustrates the wide variety of tumor diameters in a
study in which no data were rejected on account of low
signal-to-noise ratio. Thus, we conclude that the tumor
detectability over the study of 44 cancers and 116 sub-
jects was not flawed by lack of cancer detectability with
tumor-to-tissue signal ratio of 2 X for the angiogenesis
signal.
The light sources are flashed alternately 20 times per
second and electronic circuits are designed to amplify and
time-separate the signals in a “sample and hold” circuit
which integrates the signals over an interval of a few sec-
onds sufficiently rapidly to follow the movements of the
sensor over the breast, allowing usually 10 seconds for
any particular sensor position. The data are then digitized
and presented as a running time display so that the opera-
tor can be sure that stable readings are reached at each
Figure 1. A photograph of the whole apparatus (a) illustrates the handheld puck or probe, the coupling to the circuit box which con-
tains the drivers for the LED, the amplifiers for the OPTI-101s, the digitally controlled gain adjustment amplifier, the electronic switch
which decodes the light pulses and stores the information in a memory capacitor, the second set of switches which sample the memory
capacitor at a rate compatible with the computer ADC, the software for computing blood concentration and blood saturation and the dis-
play on the computer which serves to normalize the signals through the digitally controlled amplifier (b). Handheld puck (c), which has 3
elements in series, gives a considerably narrow band wavelength at the appropriate wavelengths for measuring differentially the increase
of absorption at 760 nm; the decrease of absorption at 850 nm on deoxygenation of hemoglobin rejecting the common mode change
when the difference of the two is taken; and 805 nm is used for blood volume monitoring (26).
Academic Radiology, Vol 12, No 8, August 2005
NIR DETECTION OF BREAST CANCER
927
position of the sensor (19,32). The particular version of
the apparatus shown in Fig. 1 includes pressure transduc-
ers so that the effect of pressure on hemodynamics can be
precisely monitored and it is expected that this will de-
crease the variability of the results obtained with manual
estimations of pressure (28).
Methods of Procedure
Data Acquisition.—Skill was required to ensure that all
sensors were pressed upon the breast with equal pressure
(3 mmHg) (28,33). Patients were made to lie on their
backs and the puck was placed on the breast in such a
way that coronal breast scan data was acquired (perpen-
dicular to gravity).
The Cancer-Free Reference Signals.—The light inten-
sity from the 8 detectors was adjusted to be near 1 volt
by gains set and calibrated with a phantom with known
absorption and scattering coefficients (
a
0.04 to 0.07
and
s
8cm
1
). The puck was then transferred to the
contralateral breast to include the mirror image location
of the suspected cancer. The signal outputs from all
source-detector combinations were recorded. The probe
was then moved from the contralateral breast to the ipsi-
lateral breast suspected of cancer. The sensors giving the
largest changes with respect to the mirror image position
on the contralateral breast were related to the suspected
cancer. The procedure required less than 10 minutes.
Data Analysis
In order to validate the utility of the method described
above, we have used an appropriate blood model for three
purposes:
1. to establish the blood concentration metric,
2. to establish that the saturation metric was approxi-
mate, and
3. to establish that there was minimal cross-talk be-
tween the two.
It was found that the following equations were needed for
point 3 as determined not only by the concentration and sat-
uration of hemoglobin but by increments of blood volume
added to a system set at 50% to 70% saturation (33) (the
expected nominal value for tumor oxygen saturation with the
blood concentration set in the region of 50
M).
This calibration resulted in the following two equations
for blood concentration and oxygen saturation with mini-
mal cross-talk between the two. The concentrations may
be computed using ⌬␧ 1 cm
1
and L 4 cm 5 for
pathlength factors of 5, OD ⫽⌬·C·L where OD
optical density, extinction coefficient, C concen-
tration, and L the mean pathlength of photons.
BV 0.3 · OD730 ⫹⌬OD850 (1)
Deox y 1.3 · OD730 ⫺⌬OD850 (2)
Figure 2. A two-dimensional nomogram display of currently
available NIR breast cancer data. The abscissa is relative incre-
ments of blood concentration in units of micromolar concentration
change with respect to the average cancer-free value. The ordi-
nate represents incremental change with respect to the average
cancer-free value in percent change of hemoglobin saturation.
The reference values (zero for saturations and blood volumes) are
based on the contralateral cancer-free breast as represented by a
congruence. The wavelengths are at 760 and 850 nm, and the
incremental concentrations are appropriately processed for the
weighted sum and difference quantities for blood concentration
and saturation change, respectively, with 70% and 50
M the ref-
erence points (0) for the two metrics. The data cover studies from
1997 to 2003. Verified cancers are indicated with a dot and can-
cer-free breasts plotted with an X. Data from Leipzig are marked
L, data from Philadelphia are not marked. The nomogram is di-
vided into two parts, one containing the verified cancers (I) and
the other containing the cancer-free breasts (II).
Figure 3. ROC/AUC of the Cancer Zone (Zone I) in Figure 2.
CHANCE ET AL
Academic Radiology, Vol 12, No 8, August 2005
928
In order to demonstrate the utility of the two proposed
parameters, relative angiogenesis and relative hypermetab-
olism, we have followed processes of data analysis and
calibration procedures to quantitate blood volume concen-
tration (BV) and % relative oxygenation (%Deoxy). The
concentrations of Hb (deoxy hemoglobin) and HbO
2
(Oxy
hemoglobin) were calculated by a modified Beer-Lambert
Law, OD log
I
o
I
n
where I is light intensity after ab-
sorption and scattering and I
0
is the baseline taken from
the contralateral breast, using known extinction coeffi-
cients of Hb, HbO
2
and differential pathlength factors
(DPF) of 7-8 (34). BV and %Deoxy were calculated
by the proportionalities
BV ⬀⌬[Hb] ⫹⌬[HbO
2
] (3)
Deox y ⬀⌬[HbO
2
] ⫺⌬[Hb] (4)
Note that BV and Deoxy were based on a lipid
blood oxygen model. Thus the increments of BV and De-
oxy are relative to the contralateral breast:
BV
tumor
⫺⌬BV
contra
(5)
Deoxy ⫽⌬Deoxy
tumor
⫺⌬Deoxy
contra
(6)
where BV
tumor
, BV
contra
are BV in the tumor breast
and the mirror image position of the contralateral breast,
respectively, and Deoxy
tumor
, Deoxy
contra
are Deoxy in
the tumor breast and the mirror image position of the
contralateral breast, respectively. We further converted
oxygenation to % relative oxygenation (see above values
of ,C,L).
The average value of hemoglobin saturation for non-
cancer tissue is taken to be 70% (32,33). However, the
nomogram display allows the merging of datasets based
upon the mean value of the cancer-free breast. This value
of 70% is marked zero on the ordinate of Figure 2 and is
an approximate average value for the cancer-free breasts
shown in Figure 2. Both “zero references” for cancer-free
subjects can exhibit demographic variation. An average
value as determined optically (mainly arteriolar, capillary,
venular vessels) is approximately 25
M and this value is
taken to be the Zero point on the abscissa. However, in
both axes, the cancer-free breast tissue optical values are
set to zero (usually the contralateral mirror image tissue
value) and deviations are due to angiogenesis and hyper-
metabolism of the cancer. In this way the effect of de-
mographic and other changes upon the cancer detection
are minimized and an otherwise obscured diagnostic
capability emerges based on the incremental cancer
values with respect to cancer-free values (35).
Sensitivity, specificity and AUC/ROC (area under the
ROC curve) were calculated using 2D angiogenesis/hy-
permetabolism nomograms by using incremental BV and
%Sat (36).
RESULTS
All verified cancers covering studies from 1997 to
2003 are plotted in the 2D nomogram display of BV
%Sat relationship (Figure 2A).
We have been able to summarize our data in the ROC
diagram (Fig. 3). The calculated sensitivity, specificity,
PPV and NPV are appended in Table 1 (36).
To accommodate data from others’ studies, this nomo-
gram is in relative terms with 0 for incremental blood
volume and decremental saturation based upon the mean
value of the normal population. With this formulation,
angiogenesis and hypermetabolism would identify the
cancers and segregate them in the upper right portion of
the chart and the noncancers on the lower left portion. It
is seen that this survey appears to be in accordance with
that segregation and therefore the upper right portion has
been identified by a box enclosing nearly all of the can-
cerous breasts, leaving the bulk of the noncancerous
breasts in the lower left portion of the diagram.
False Positives and Negatives
All 6 “out of box” readings are within the zonation of
the nomogram. The cancer box contains 4 noncancer pa-
tients and the noncancer box contains 2 cancer patients.
The 4 false positives in the cancer box are very near the
margins of the noncancer box and 2 are low blood vol-
ume but high desaturation. A small error in determination
of the blood volume increment of roughly 0.2 micromolar
is involved. The two false negatives are identified as
small cancers, 6 and 7 mm, in which the blood volume
increment is significantly underestimated while the deoxy-
genation is significant.
Population Differences
The Leipzig population (marked with L) shows a wide
blood volume distribution and a small increment of de-
saturation. In another article we showed correlations of
Academic Radiology, Vol 12, No 8, August 2005
NIR DETECTION OF BREAST CANCER
929
the optical determination of desaturation with the Eppen-
dorff needle electrode determinations of tissue oxygen
concentration in collaboration with Dr. P. Vaupel (38,39,
and unpublished data ). The zero point of blood concen-
tration is 50
M blood and the zero point of saturation
is 70%.
Operator Differences
Both in Philadelphia and in Leipzig, the data were ac-
quired by more than one operator; the instruments were
electronically identical.
Breast Size
In order to give an indication of the range of breast
cancer sizes, Fig. 4 provides a histogram display of the
diameters of the tumors examined as obtained from mam-
mography and histopathologically verified cancers.
Figure 4 displays the distribution of sizes of the de-
tected cancers for which size data are available. Most
were under 2 cm in diameter. It is limited at the low end
by the fact that the mammography does not readily detect
cancers below 8 mm in diameter and in the future, MRI
will be used in order to define the lower limiting size for
the detection of angiogenesis and hypermetabolism by the
NIR system with source-detector separations that may
exceed the 4 cm used here.
The high blood volume content of cancers is in agree-
ment with Folkman’s ideas of angiogenic activities of
tumors (3) and the hypoxic regions of tumors are in ac-
cordance with Vaupel’s oxygen electrode data (38). Intes
recently published similar studies but reported data on an
absolute basis (35). It should be emphasized that special
care has been taken here to employ blood lipid models
with variable blood volume and saturation to ensure mini-
mizations of crosstalk between increase of blood volume
and decrease of saturation.
DISCUSSION
This multicenter study has demonstrated a simple,
economical, and patient-friendly NIR breast cancer de-
tector with a multi-wavelength, multi-detector system
that gives incremental values of angiogenesis and hy-
permetabolism, and that is proven to give an ROC/
AUC of 95%. The 2-parameter nomogram display of
Figure 4 illustrates the total databank obtained over the
several years of studies at the University of Pennsylva-
nia in Philadelphia and at Leipzig. The nomogram dis-
play contains two zones, Zone 1 in which relative an-
giogenesis/hypermetabolism predominates, and Zone II,
cancer-free breasts in which neither metric predomi-
nates. The usefulness of this 2D nomogram in compar-
ing data over a range of demographics is also demon-
strated by other laboratories (35,39).
A novel concept of angiogenesis and hypermetabo-
lism which is complementary to the focus on morpho-
logical characteristics of cancer employed by the ana-
tomical approach of ultrasound, X-ray mammography,
and MRI is to focus on two biochemically, physiologi-
cally based characteristics of cancer not heretofore em-
ployed as a diagnostic, namely the spatial correlation
of incremental values of total hemoglobin and deoxy-
genation of hemoglobin, predominantly in the arterio-
lar, venular, and capillary bed of breast cancers that
are incremental with respect to the mirror image loca-
tion on the contralateral breast. The NIR optical infor-
mation is in the form of magnitudes rather than shapes,
a considerable distinction between the NIR optical
method (31,32,34) and the morphologically based X-
ray mammography (41), ultrasound (42), and MRI (40).
Figure 4. Histogram of tumor size for those verified breast can-
cers for which size data are available, 38 breast cancers detected
by NIR hand held “puck.”
Table 1
Summary of Statistics of Relative Oxygenation/Blood Volume
Concentration for Figure 2
ROC/AUC 95%
Sensitivity 96%
Specificity 93%
PPV 89%
NPV 97%
CHANCE ET AL
Academic Radiology, Vol 12, No 8, August 2005
930
With regard to hypermetabolism NIRS can detect with
low oxygen concentration, and PET detection with
FDG (43) uses the same concept of lack of oxygen me-
tabolism. Taking into account the work of Vaupel (38)
and others indicating that, in spite of intense angiogen-
esis, a growing cancer may still be underserved with
oxygen because of the faulty nature of the vessels and
the oxygen demand of growth, to the extent that glyco-
lytic activity may be stimulated as originally observed
by Otto Warburg (1) and studied further by us as
Krebs’ Crabtree’s effect. It seems therefore reasonable
that the increment of the blood flow might be corre-
lated with the decrease of oxygen saturation and that a
coordinated plot of the two might be useful. On this
basis, relative rather than absolute values may be ade-
quate in NIRS and previous study suggests excessive
between-subject effects (37). Large variables of the
bulk optical property of the breast suggest the use of
internal reference rather than absolute values of the
optical properties (45).
Although by using selected wavelengths sufficient
sensitivity and specificity are obtained in these studies
to provide a highly significant ROC/AUC score, this
CWS method is less sensitive to the many demographic
factors which influence the properties of the human
breast (46) yet which are known in general to influence
both breasts equally. One of the key problems of the
breast is the wide variety of demographic size, texture,
chemical composition, and age (Tromberg, personal
communication). Thus, a secure tissue baseline is not
available unless it is taken from the adjacent tissue,
which is found to be a satisfactory reference and is a
feasible standard for cancer detection. One may choose
either a portion of the same breast or the mirror image
portion of the contralateral breast. The latter may seem
to be better since the majority of breasts are highly
symmetrical. For this reason we propose as our stan-
dard baseline metric, mirror image portion of the con-
tralateral breast. If a priori information is available, for
example, by palpation, ultrasound, or mammography,
the NIR device may be placed directly on the surface
projection of the cancer. The work of others, particu-
larly Tromberg (47), Cubbedu (47), Vaupel (49), and
Yamashita (49) have suggested that the use of other
metrics such as water, lipid, and light scattering could
afford increases in AUC multi-dimensional plots of the
type used here. However, none of these is known to be
directly related to the biochemical, physiological prop-
erties of tumor growth as are the two parameters se-
lected here. However, other biochemical characteristics
may be employed, for example the concentration of
intracellular compounds related to lipid biosynthesis or,
indeed, the impact of growth and its concomitant en-
ergy demand upon the cancer bioenergetics (48).
The addition of further metrics, combined with
higher spatial resolution, should increase further the
ROC/AUC values and at the same time give better
quantitation of the optical characteristics of deep small
tumors. However, the high throughput, simplicity of
operation, and the convenience of this device make it
worthwhile to proceed with further studies to improve
the breast sensor contact problem by contact pressure
sensing or eliminate the problem by flying spot non-
contact, remote sensing of the optical signals (17). In
the later case, higher resolution imaging can be readily
obtained by flying spot and phased array systems
(17,23,24) from all geometric aspects of the breast.
Such a manifold of source and detector positions pro-
posed must decrease the throughput and increase the
complexity of the device. Further, the frequency of oc-
currence of small cancers and precancers is low, e.g.,
in the Hospital of the University of Pennsylvania
(HUP) BRCA1/2 study, fewer than 1% of those in the
high-risk category initially exhibit a detectable cancer
(44).
Each breast cancer detection method (MRI, ultrasound,
optical tomography) may have its own niche in health care
delivery, the larger more complex devices as a part of Ra-
diological Services and the small devices for clinics special-
izing in early detection of cancer, and even for personal
healthcare as befits underserved, noncompliant populations
(50). There are a number of intrinsic signals that are under
investigation elsewhere that may be quite useful. Water con-
tent is readily accessible optically, as are scattering, lipid,
and pigment. If exogenous probes are to be considered, ICG
affords not only increased tumor-to-tissue ratio, but also a
measure of diagnostic capability. Finally, molecular beacons
responsive to the energy demands of tumor growth, particu-
larly fluorescent glucose, by analogy to radioactive glucose,
may add to the sensitivity and specificity of cancer detection
on a biochemical/physiological basis. Thus, a very large
number of factors will increase the sensitivity and specificity
of cancer detection by optical means. While incremental
changes of angiogenesis and hypermetabolism are small,
much larger tumor-to-tissue ratios can be expected of extrin-
sic signals, ICG, fluorescent glucose, and genetically directed
molecular beacons.
Academic Radiology, Vol 12, No 8, August 2005
NIR DETECTION OF BREAST CANCER
931
Relation to Other Studies
While a general consensus has been reached that an-
giogenesis gives a high blood volume signal, results seem
to be inconsistent with respect to the blood saturation
level of cancer. We believe due to the lack of blood mod-
els that vary both blood volume and blood O
2
saturation,
that separation of the blood volume signal from the satu-
ration signal is important, and that an increase of blood
volume, causing an increase of oxy hemoglobin, will ad-
versely affect the calculation of desaturation of hemoglo-
bin. In fact, the usual equations for calculating saturation
do not take into account the possibility of crosstalk from
other signals as does our empirical equation (13,35,39).
Effect of Cancer Type
Since our approach is based upon two biochemical/
physiological properties of breast cancers, the structural,
architectural, and histopathological features which may be
used to classify cancers have been found less well-corre-
lated with the Figure 3 nomogram. As might be expected,
these features are de-emphasized in this survey of bio-
chemical and physiological differences of growing can-
cers relative to the corresponding cancer-free tissue. Thus,
commonality of factors involved in the biochemistry of
tumor growth is expected from these studies while struc-
tural features of different kinds of tumors would be de-
emphasized by the use of relative values for the baseline,
emphasizing the difference between this approach and one
based upon morphological characteristics involving spicu-
lation, lobulation, and other demographic features of the
cancers (41,42,46).
Future Developments
A measure of success based upon a nomogram of two
intrinsic features of breast cancer can be improved by
other metrics that correspond physiologically or biochemi-
cally to growth stimuli (26,51).
REFERENCES
1. Warburg O. Biochem Z 1926; 177:471– 486.
2. Zhao Z, Zhang J, Nioka S, Dong L, Du J, Wen S, Chance B. Blood flow
and oxygen saturation changes according to pressure effect in NIR
breast cancer imaging study. SPIE Proceedings Vol. 5693 Optical to-
mography and spectroscopy of tissue VI (Chance B, Alfano RR, Trom-
berg BJ, Tamura M, Sevick-Muraca EM, eds.) Bellingham, WA: SPIE,
2005; 271-277.
3. Chance B, Castor LN. Some patterns of the respiratory pigments of
Ascites tumors of mice. Science 1952; 116:200 –202.
4. Folkman J. Tumor angiogenesis: therapeutic implications. N Engl
J Med 1971; 285:1182–1186.
5. Lin Y, Lech G, Nioka S, Intes X, Chance B. Noninvasive, low-noise, fast
imaging of blood volume and deoxygenation changes in muscles using
light-emitting diode continuous-wave imager. Rev Sci Instrum 2002;
73:3065.
6. Patterson MS, Chance B, Wilson BC. Time resolved reflectance and
transmittance for the noninvasive measurement of tissue optical prop-
erties. J Appl Optics 1989; 28:2331–2336.
7. Cutler M. Transilumination of the breast. Surg Gynecol Obstet 1929;
48:721–727.
8. Girolamo RF, Gaythorpe JV. Clinical diaphanography—Its present per-
spective. Crit Rev Oncol Hematol 1984; 2:1–31.
9. Millikan GA. Experiments on muscle haemoglobin in vivo; The instanta-
neous measurement of muscle metabolism. Proc Roy Soc London B
1937; 129:218 –241.
10. Millikan GA. The oximeter, an instrument for measuring continuously
the oxygen saturation of arterial blood in man. Rev Sci Instrum 1942;
13:434.
11. Chance B. Stable spectroscopy of small density changes. Rev Sci In-
strum 1947; 18:601– 609.
12. Chance B, Williams GR. Respiratory enzymes in oxidative phosphoryla-
tion. II. Difference spectra. J Biol Chem 1955; 217:395– 408.
13. Yodh A, Chance B. Spectroscopy and imaging with diffusing photons.
Phys Today 1995; 48:34 40.
14. Cheng Z, Mao JM, Bush R, Kopans DB, Moore RH, Chorlton M. Breast
cancer detection by mapping hemoglobin concentration and oxygen
saturation. Appl Opt 2003; 42:6412– 6421.
15. Jakubowski DB, Cerussi AE, Bevilacqua F, Shah N, Hsiang D, Butler J,
Tromberg BJ. Monitoring neoadjuvant chemotherapy in breast cancer
using quantitative diffuse optical spectroscopy: a case study. J Biomed
Opt. 2004 Jan-Feb;9(1):230-8.
16. Hamaoka T, Albani C, Chance B, Iwane H. A new method for the eval-
uation of muscle aerobic capacity in relation to physical activity mea-
sured by near-infrared spectroscopy. In Saito Y, Portmans J, Hashi-
moto I, Oshida Y, eds: Integration of Medical and Sports Sciences
Basel: Karger, 1992, vol. 37, p. 421– 429.
17. Gratton E, Mantulin WW, vandeVen MJ, Fishkin JB, Maris M, Chance
B. The possibility of a near-infrared optical imaging system using fre-
quency domain methods. Proceedings of the Third International
Conference: Peace through Mind/Brain Science (Aug. 5-10),
Hamamatsu, Japan, 1990; pp. 183-189.
18. Kang KA, Chance B, Zhao S, Srinivasan S, Patterson E, Troupin R.
Breast tumor characterization using near-infra-red spectroscopy. SPIE
Proceedings Vol. 1888 Photon Migration and Imaging in Random Me-
dia and Tissues (Chance B., Alfano RR, eds.) Bellingham, WA: SPIE,
1993; 487– 499.
19. Chance B, Tamura T. Spatial and temporal resolution of frontal cogni-
tive response using NIR tomography. International Symposium of Brain
Mapping OISO ‘96 (Sept. 13-15), Oiso, Japan, 1996; p. S-III-4.
20. Chance B, Dait MT, Chang C, Hamaoka T, Hagerman F. Non-invasive
evaluation of deoxygenation during intense exercise and recovery of
oxygenation of hemoglobin and myoglobin in the quadriceps muscles
of elite competitive rowers. Am J Physiol 1992; 262:C766 –C775. BC
brain.
21. Chance B. NIR responses of the human forebrain (Brodman’s 9 and
10) to physiological function. SPIE Proceedings Vol. 5693 Optical to-
mography and spectroscopy of tissue VI (Chance B, Alfano RR, Trom-
berg BJ, Tamura M, Sevick-Muraca EM, eds.) Bellingham, WA: SPIE,
2005; 166-171.
22. Chance B, Onaral B, Pourrezai, Herr M. A new approach to early de-
tection for women’s medicine. Era of Hope Department of Defense
Breast Cancer Research Program Meeting Proceedings. Philadelphia,
PA: Department of the Army, 2005; 48 .
23. Zhang J, Lin Y, Nioka S, O’Connor N, Czemiecki B, Conant EF,
Chance B. Application of LED device for breast cancer diagnosis, Proc.
SPIE 2002; 4916:30 –36.
24. Chance B, Kang K, He L, Weng J, Sevick E. Highly Sensitive object
location in tissue models with linear in-phase and anti-phase multi-ele-
ment optical arrays in one and two dimensions. 1993. Proc Natl Acad
Sci USA 1993; 90:3423–3427.
25. Chen Y, Zheng G, Zhang ZH, Blessington D, Zhang M, Li H, Liu Q,
Zhou L, Intes X. Metabolism-enhanced tumor localization by fluores-
cence imaging: in vivo animal studies. Optics Let 2003; 28:2070 –2072.
CHANCE ET AL
Academic Radiology, Vol 12, No 8, August 2005
932
26. Vo-Dinh T. Biomedical Photonics Handbook. Boca Raton, FL: CRC
Press. 2003.
27. Chance B. The optical method. Annu. Rev Biophys Biophys Chem
1991; 20:1–18.
28. Zhou S, Huang S, Xie C, Long H, Nioka S, Chance B. Optical imaging
of breast tumor by using dual wavelength amplitude cancellation sys-
tem (phased array). In: Sevick-Muraca EM, Izatt JA, Ediger MN, eds.
OSA Trends in Optics and Photonics, Vol. 22. Biomedical Optical
Spectroscopy and Diagnostics/Therapeutic Laser Applications. Wash-
ington, DC: Optical Society of America, 1998; 197–200.
29. Nioka S, Wen S, Zhang J, Du J, Intes X, Zhao Z, Chance B. Simulation
study of breast tissue hemodynamics during pressure perturbation.
Adv Exp Med Biol. 2005 in press.
30. Fukui Y, Ajichi Y, Okada E. Monte Carlo prediction of near-infrared
light propagation in realistic adult and neonatal head models, Appl Opt
2003; 42:2881–2887.
31. Gopinath SP, Robertson CS, Contant CF, Narayan RK, Grossman RG,
Chance B. Early detection of delayed traumatic intracranial hematomas
using near-infrared spectroscopy. J Neurosurg 1995; 83:438 44.
32. Choe R, Durduran T, Yu G, Nijland MJ, Chance B, Yodh AG, Ramanu-
jam N. Transabdominal near infrared oximetry of hypoxic stress in fetal
sheep brain in utero. Proc Natl Acad SciUSA.2003;
100:12950 –12954.
33. Chance B, Nioka S, Cappola T, Margulies K. NIR imaging of fetal brain
and adult myocardium. The 46
th
Experimental Nuclear Magnetic Reso
-
nance Conference, April 10 –15, 2005, Rhode Island Convention Cen-
ter, Providence, RI.
34. Smith DS, Levy W, Maris M, Chance B. Reperfusion hyperoxia in brain
after circulatory arrest in humans. Anesthesiology 1990; 73:12–19.
35. Fantini S, Hueber D, Franceschini MA, Gratton E, Rosenfeld W,
Stubblefield PG, Maulik D, Stankovic MR. Non-invasive optical moni-
toring of the newborn piglet brain using continuous-wave and fre-
quency-domain spectroscopy. Phys Med Biol 1999; 44:1543–1563.
36. Intes X, Djeziri S, Ichalalene Z, et al. Time-domain optical mammogra-
phy SoftScan
©
: initial results. Acad Radiol 2005; In Press.
37. Campbell MJ, Machin D. Medical Statistics 3rd ed. Wiley, 1999, pp.
43-46.
38. Thews O, Li Y, Kelleher DK, Chance B, Vaupel P. Microcirculatory
function, tissue oxygenation, microregional redox status and ATP distri-
bution in tumors upon localized infrared-A-hyperthermia at 42 degrees
C. Adv Exp Med Biol 2003; 530:237– 47.
39. Vaupel P, Schlenger K, Knoop C, Hockel M. Oxygenation of human
tumors: evaluation of tissue oxygen distribution in breast cancers
by computerized O
2
tension measurements. Cancer Res 1991; 51:
3316 –3322.
40. Grosenick D, Wabnitz H, Moesta KT, et al. Concentration and oxygen
saturation of haemoglobin of 50 breast tumours determined by time-
domain optical mammography. Phys Med Biol 2004;7:49, 1165-81.
41. Ntziachristos V, Yodh A, Schnall M, Chance B. MRI-Guided diffuse op-
tical spectroscopy of malignant and benign breast lesions. Neoplasia
2002; 4:347–354.
42. Li A, Miller E, Kilmer M, et al. Tomographic optical breast imaging
guided by three-dimensional mammography. Appl Opt 2003; 42:5181–
5190.
43. Zhu Q, Huang MM, Chen NG, et al. Ultrasound-guided optical tomo-
graphic imaging of malignant and benign breast lesions: initial clinical
results of 19 cases. Neoplasia 2003; 5:379 –388.
44. Alavi A, Kung JW, Zhuang H. Implications of PET based molecular im-
aging on the current and future practice of medicine. Semin Nucl Med
2004; 34:56 69.
45. Pogue BW, Jiang S, Dehghani H, et al. Characterization of hemoglobin,
water, and NIR scattering in breast tissue: analysis of intersubject vari-
ability and menstrual cycle changes. J Biomed Opt 2004; 9:541–552.
46. Durduran T, Choe R, Culver JP, et al. Bulk optical properties of healthy
female breast tissue. Phys Med Biol (2002).; submitted
47. Tromberg BJ, Shah N, Lanning R, et al. Non-invasive in vivo character-
ization of breast tumors using photon migration spectroscopy. Neopla-
sia 2000; 2(1-2): 26 40.
48. Torricelli A, Spinelli L, Pifferi A, Taroni P, Cubbedu R. Use of a nonlin-
ear perturbation approach for in vivo breast lesion characterization by
multiwavelength time-resolved optical mammography. Opt Express
2003;11 p. 853.
49. Vaupel P, Hockel M. Blood supply, oxygenation status and metabolic
micromilieu of breast cancers: characterization and therapeutic rele-
vance. Int J Oncol 2000; 17:869 879.
50. Suzuki K, Yamashita Y, Ohta K, Chance B. Quantitative masurement of
optical parameters in the breast using time-resolved spectroscopy:
Phantom and preliminary in vivo results. Invest Radiol 1994; 29:410 4.
51. Armstrong K, Weber BL. Breast cancer screening for high-risk women:
too little, too late? J Clin Oncol 2001. Feb 15;19:919 –20.
Academic Radiology, Vol 12, No 8, August 2005
NIR DETECTION OF BREAST CANCER
933
    • "The latter application benefits from the sensitivity of NIRS and DOI to important breast tissue chromophores, namely oxy-hemoglobin, deoxy-hemoglobin, water, and lipids, and leads to the discrimination of benign and malignant breast tumors [5,6], and monitoring response to neoadjuvant chemotherapy [7][8][9][10]. More specifically, optical mammography has shown its potential in detecting angiogenesis [11], hypoxia [12,13], and collagen content as a measure of increased breast tissue density [14], which are all relevant physiological parameters for the diagnosis and characterization of breast cancer [15][16][17]. To enhance the spatial information content of DOI, it is desirable to achieve spatial resolution in three dimensions. "
    [Show abstract] [Hide abstract] ABSTRACT: We present a method for depth discrimination in parallel-plate, transmission mode, diffuse optical imaging. The method is based on scanning a set of detector pairs, where the two detectors in each pair are separated by a distance δDi along direction δ D i within the x-y scanning plane. A given optical inhomogeneity appears shifted by αi δ D i (with 0≤ αi ≤1) in the images collected with the two detection fibers of the i-th pair. Such a spatial shift can be translated into a measurement of the depth z of the inhomogeneity, and the depth measurements based on each detector pair are combined into a specially designed weighted average. This depth assessment is demonstrated on tissue-like phantoms for simple inhomogeneities such as straight rods in single-rod or multiple-rod configurations, and for more complex curved structures which mimic blood vessels in the female breast. In these phantom tests, the method has recovered the depth of single inhomogeneities in the central position of the phantom to within 4 mm of their actual value, and within 7 mm for more superficial inhomogeneities, where the thickness of the phantom was 65 mm. The application of this method to more complex images, such as optical mammograms, requires a robust approach to identify corresponding structures in the images collected with the two detectors of a given pair. To this aim, we propose an approach based on the inner product of the skeleton images collected with the two detectors of each pair, and we present an application of this approach to optical in vivo images of the female breast. This depth discrimination method can enhance the spatial information content of 2D projection images of the breast by assessing the depth of detected structures, and by allowing for 3D localization of breast tumors.
    Full-text · Article · Mar 2013
    • "In this study, the goal is to increase the lesion visibility in DOT by combining fluorescence and optical absorption data at the voxel level in one single graph, a scatterplot. This concept was introduced by Chance et al. in optical imaging of breast cancer [11]. They plotted the mean percentage of oxygen desaturation of blood versus the mean blood volume of their patients into a graph. "
    [Show abstract] [Hide abstract] ABSTRACT: Using scatterplots of 2 or 3 parameters, diffuse optical tomography and fluorescence imaging are combined to improve detectability of breast lesions. Small or low contrast phantom-lesions that were missed in the optical and fluorescence images were detected in the scatterplots. In patient measurements, all tumors were visible and easily differentiated from artifacts and areolas in the scatterplots. The different rate of intake and wash out of the fluorescent contrast agent in the healthy versus malignant tissues was also observed in the scatterplot: this information can be used to discriminate malignant lesion from normal structures.
    Full-text · Article · Apr 2011
    • "However , the DOI literature has contrasting reports on StO2 in cancerous tissue. A decrease of StO2 has been reported by some groups474849 , whereas no statistically significant changes have been reported by others [19,20,42,45,50] . This discrepancy could be attributable to the fact that tissue oxygenation differs depending on cancer stage and type. "
    [Show abstract] [Hide abstract] ABSTRACT: Existing imaging modalities for breast cancer screening, diagnosis and therapy monitoring, namely X-ray mammography and magnetic resonance imaging, have been proven to have limitations. Diffuse optical imaging is a set of non-invasive imaging modalities that use near-infrared light, which can be an alternative, if not replacement, to those existing modalities. This review covers the background knowledge, recent clinical outcome, and future outlook of this newly emerging medical imaging modality.
    Full-text · Article · Jan 2011
Show more