Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
O1%24%O1 11:08 FAQ 613 234 3583 ilacRae & Co. ~ 005
_ . ~ ~~
. : _ W,
4
b~T80D OF DETECTION OF CANCEROUS
LESIONS HY TBEIR LFp'$CT ON ~ SPATIAL
DZSTRIBLiTION OF MODULATION OF TE'~dPERATURE
AND HOMOGENEITY OF TISSUE
CROSS-REFERENCE TO RELATED A..ppLICATION
The present application is a continuation-in-part
of application Serial No. 08/368,161, filed January 3,
1995.
BACKGROUND OF THE INVENTION
The present invention relates generally to cancer
detection and, more particularly, to~a cancer detection
method involving the measurement, recording and analysis
of temporal periodic perfusion changes associated with
dysfunction of neuronal control of the vasculature
surrounding cancerous lesions. While the present
invention has application to cancer detection throughout
the human body, it za particularly applicable to a
breast cancer screening teat involving the measurement
of temporal changes in perfusion over large areas of the
breasts to identify cancer. Specifically, cancer
detection is derived from a time series of infrared
images of an area o.f interest of a human breast. The
24 infrared images relate to the temporal periodic
perfusion changes and are converted by a computer to
time series. average temperature and standard deviation
of each of a plurality of subareas. The data is then
analyzed to identify clusters having abnormal
temperature dependent frequencies indicative of cancer.
SUMMARY OF THE INVENTION
Thermoregulatory frequencies of the processe$ that
control akin temperature over different areas of the
body are derived from the periodic changes in
temperature distribution over those skin areas. As
shown by Dr. Anbar in the European J Thermology 7:105-
118, 1997, under conditions of hyperperfusion, the
CA 02333406 2000-11-24
O1%2401 11:08 FAQ 613 23:~ 3583 MacRae & Co. f~ 008
~ s
- 2 -
homogeneity of the skin temperature distribution reaches
a maximum and the amplitude of its temporal modulation
is at a minimum. In other words, periodic changes in
the spatial homogeneity of akin temperature (HST) are
dictated by the processes that control the satuz~ation of
the cutaneous capillary bed in a particular area of the
akin. Accordingly, skin temperature and HST are
independent physiological hemodynamic parameters that
are determined by the structure of the cutaneous
vasculature and by its heat dissipatory activity.
However, unlike average temperature,~HST is affected
mainly by the behavior of the cutanevus capillaries and
to a much lesser extent by the blood flow in
subcutaneous vessels. As perfusion is enhanced, more
capillaries are recruited as blood conduits and HST
increases. The neuronal control of HST is, therefore,
different from that of skin temperature. Gonseqvently,
the spatial distribution of changes in skin temperature
and homogeneity, HST, of tissue derived from a time
series of infrared images of a skin area according to
the present invention is more informative than a
classical thermogram. HST is the reciprocal of the
spatial coefficient of variation of temperature in small.
( T~micro" ) areas of skin (< 100 mmz) : HST = average
temperature divided by the standard deviation of the
average temperature (HST is a dimensionless parameter).
The concept of HST has been fully described by,.Dr.
Michael Anbar in Biomedical Thermology, 13:173-187,
1994.
These and other objects of the present invention,
will become increasingly more apparent to those skilled
in the art by reference to the following description.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Although skin temperature may vary over a wide
range, depending upon the environment and on the level
of metabolic activity, it is regulated under normal
CA 02333406 2000-11-24
O1%24!01 11:09 FAQ 613 234 3563 hlacRae & Co. f~ 007
. , ,, ,
- 3 -
conditions. This regulation may be occasionally less
stringent, like during sleep but even then, some skin
regulation is retained. In the human body, skin
temperature maintenance is due, in substantial part to
S neuronal thermoregulation of vasoconstriction and
vasodilation of the vasculature, thereby causing a
characteristic mvdulativn of blvvd perfusion, unless the
neuronal thermoregulation is inhibited or taken over by
a non-neuronal thermoregulatory control, such as nitric
oxide (NO). Modulation of skin temperature is on the
order of 10 to 50 millidegrees Kelvin.
NO has been recognized as an ubiguitous
vasodilatory chemical messenger. Its main role appears
to be synchronisation of intercellular and inter organ
functions, because it diffuses freely in the
interstitial space. This has been discussed extensively
by Dr. ~lnbar in J. Pain Symptom Management 14:225-254,
1997. Under certain conditions, such as in breast
cancer, autocatalytic production of NO may occur, which
results in oscillatory vasodilation, independent of and
substantially different from the temperature
vacillations of perfusion caused by the neuronal
thermvregulatory system. NO can, therefore, inhibit
autonomic vasoconstrictive control in substantial
regions of the microvasculature and cause regional
hyperperfusion. Subcutaneous and cutaneous
hyperperfusion are manifested as hyperthermia of the
overlying skin.
Thus, cancer associated breast hyperthermia results
from impaired neuronal thermoregulation caused by
excessive production of NO by breast cancer cells and in
macrophages that react to the neopJ~astic tissue. The
latter activity is an expression of the immune response
because NO generated by macrophages is a major factor in
killing of microorganisms or of mammalian cells
recognized as alien. The mechanism for this activity
is described fully in a 1994 publication of Dr. Michael
CA 02333406 2000-11-24
O1!2~%O1 11:09 FAX B13 234 3563 MacRae & Co. ~ 008
- 4 -
Anbar, Cancer Letters 84:2 329, 1994. The rate of
production of NO is further enhanced by the presence of
ferritin, the level of which is significantly elevated
in breast cancerous tissue. Fe~+ released from ferritin
S is used to produce more NO synthase (NOS), an iron
carrying enzyme which produces NO from arginine, and
thus results in a further increase in the rate of
production of NO. Furthermore, NO has been shown to
release Fey'' from ferritin by forming an NO-ferritin
complex which consequently results in an autocatalytic
production of NO. Fey' also eliminates superoxide
radicals (HOZ) which are normally the major scavengers of
NO. Elimination of H02 maintains the local high level of
NO, and hyperperfusion of the tissues surrounding the
cancerous lesion. The ferritin dependent enhancement of
NO production seems to be specific to neoplastic cells
and is less likely to occur in other inflammatory
situations, including those induced by microoxganzsms.
Therefore, the freguencies of temperature and HST
oscillations observed over a cancerous breast differ
substantialJ.y from those observed over a non-cancerous
(normal) breast. The oscillations over a normal breast
are caused by the neuronal thermoregulatory processes,
which follow several characteristic bands of
frequencies. The cancerous area of the breast, on the
other hand, which loses its neuronal thermoregulatory
control due to the over-production of NO, is
characterized by the disappearance of the neuronal
osciilations and the appearance of oscillations due to
the autocatalysis of NO production, with their typical
frequency bands. The latter autocatalytic processes,
controlled by the temporary local depletion of one of
the precursors of NO, are utterly different in nature
from the neurological feedback processes manifested in
the neuronal frequency bands. Consequently, the
disappearance of neuronal frequencies over substantial
CA 02333406 2000-11-24
O1%24%O1 11:09 FAQ 613 234 3563 MacRae & Co. f~ 009
- 5 -
parts of the cancerous breast is sufficient to identify
pathology. Furthermore, the appearance of the
autoeatalytic freguencies characteristic o~ NO
over-production is, by itself, sufficient to identify
pathology. The substitution of one set of frequeriCy
bands by the other is an even more strict cz~iterion of
pathology in a human breast. Hemodynamic modulation is
generally in the range of 0.8 to 2 Hz, whereas
modulation by the autonomic nervous system ranges
generally between 10 and 700 mHz.
It is the detection of these freguencies, whether
neuronal or non-neuronal, which is the basis o~ the
detection methods of the present invention.
Specifically, under conditions of extravascular NO
overproduction and its consequent hyperperfusion, skin
temperature and the homogeneity of skin temperature,
HST, reach values that oscillate at frequencies
dependent on the modulated autocatalytic rate of NO
production. The frequency bands v~ the modulation of
temperature and HST are, according to the present
detection methods, independent criteria of pathology.
It should be understood that while the present
detection methods are described herein with respect to a
human breast, it is not intended to be so limited.
Detection o~ cancerous lesion$ in the hands, arms,
chest, Legs, buttocks, back and literally any other body
part is possible with the present methods.
Thus, the screening techniques of the present
invention use the spatial distribution of the
characteristic oscillations or modulations in the
temporal behavior of blood perfusion caused by enhanced
NO production by cancerous cells and in macrophages and
amplified by ferritin to detect an immune response
induced by neoplastic disease. The temperature
vscillat~.on of blood perfusion associated with the
autocatalytic production of NO, as well as the
diminution or disappearance of the neuronal
CA 02333406 2000-11-24
O1%24%O1 11:10 FAQ 813 234 3583 MacRae & Co. (71010
- 6 -
thermoregulatory frequencies, TRFs, are used as the
diagnostic parameters. The neuronal and autocatalytic
oscillations are measured by fast Fourier transform
(FFT) analysi$, an analysis method well known in the
art, of the temporal behavior of breast perfusion
(manifested in the tempvrax behavior of breast
temperature and of HST). The TRFs of HST are,
therefore, addztianal independent diagnostic parameters.
Accordingly, the present invention provides for the
accumuiation of hundreds of sequential thermal images
that are then subjected to FFT to extract the
frequencies and amplitudes of periodic changes at each
pixel of the image. To measure the spatial distribution
of modulation of temperature and HST, an infrared camera
is positioned to provide infrared flux images of a part
of the human body, for example a human breast. A
preferred camera is equipped with a 256 x 256 focal
plane array (FPA) GaAs quantum-well infrared
photodetectvr (QWIP). Such a camera can record
modulation of skin temperatuz~e and its homogeneity with
a precision greater than f 1 millidegrees C, i.e., less
than 1/10 of physiological modulation of temperature and
of homogeneity of human skin. Another camera that is
useful with the present invention i$ a HgCdTi infrared
photodetector.
While successful results can be achieved by
analyzing the temporal behavior of at least 7.28 thermal
images, it is most preferred to measure 1024 consecutive
thermal images over a period o~ 10 to 60 seconds. The
infrared images are transmitted to a CPU having an
analog/digital converter which processes the recorded
infrared flux information and outputs digital infxexed
flux data.
To determine whether the breast is normal or
cancerous, the CPU factors the digital data to
equivalent temperature readings for a designated spot
subarea or group of spots of the breast (each spot
CA 02333406 2000-11-24
O1%24!01 11:10 FAR 813 23:1 3583 MacRae & Co. C~ 011
_ _ 7 _
subarea corresponds to about 25 to 100 mm~ of skin based
on a 256 x 256 pixel field, as will be described
hereinafter), which are later to be analyzed for the
presence of lesions. A spot is typically about four to
sixteen pixels and can have the shape of a circle,
square or other shapes. The temperature values of the
pixels in each subarea of the image are averaged. The
variance of the average temperature is used to calculate
the HST of each subarea. The accumulated images are
then analyzed by FFT to extract the corresponding
frequencies of the average temperature and standard
deviation of average temperature. The FFT yield$ the
frequency spectra of each pixel together with the
relative amplitude o~ each TRF. The software then
1.5 tabulates, or displays as color-coded bitrnaps, the
spatial distribution of the TRFs within a given range of
relative ampzitudes over the image. The same procedure
is followed with the HST data.
When TRFs in a cluster of subaxeas are displayed
with amplitudes above a given threshold, a subset of
characteristic neuronal frequencies over areas of
breasts free from cancer~enhanced immune response is
identified. A cluster is defined as at least $ix
congruent subareas of four to sixteen pixels per
subarea. Such TRFs are significantly attenuated or
completely absent in areas overlying breasts with
neoplastic lesions. The latter areas are characterized
by non-neuronal thermoregulatory behavior which manifest
substantially different TRFs caused by the autocatalytic
production of NO. Also the latter areas are, therefore,
characterized by aberrant modulation of blood perfusion
and aberrant temperature oscillations. A hard copy in
the form of a colored bitmap of the imaged skin area is
! then generated to allow an expert to anatomically
identify the location of the aberrant area, or areas.
The computer algorithms that facilitate this
computation are as follows:
CA 02333406 2000-11-24
--__01/24/01 11:10 FAQ 813 234 3583 MacRae & Co. 0 012
..
- 8 -
A. Use of temperature values of individual pixels and
the computation of TRFs.
1. The computerized camera takes an image of the
infrared flux (256 x 256 pixels) of an area of
interest of the skin, such as of a human
breast, and converts it into a digital thermal
image where each pixel has a certain
temperature value. This process is repeated,
preferably thirty times a second until 1024
thermal images have been accumulated and
stored in the computer.
2. The infrared image is subdivided into subareas
of 4 to 16 pixels, each subarea corresponding
to about 25 to 100 square millimeters of skin,
depending on the optics of the camera and the
geometry of the measurement. For example, in
a 256 x 256 pixel field, each pixel. represents
about 2.4 mm2. Higher resolution camera will
have greater skin area detail.. The computer
then converts each infrared image into a
digital thermal image where each pixel has a
certain temperature value.
3. The computer calculates the average
temperature and the temperature standard
deviation of each of the subareas in each of
the 1024 images. The average temperature
values of each subarea constitutes a'single
time series that is theca subjected to ~'FT
analysts to extract the contributing
i frequencies of each image of the series and
' their relative amplitudes. The computer
stores the resulting FFT spectrum for the
calculated subarea related to the subdivided
area of the image_ The computez~ repeats the
same procedure far each of the subareas of the
image of the skin. Additionally, or in the
CA 102333406 2000-11-24
01!24%O1 11:10 FAQ 813 234 3563 MacRae & Co. ~ 013
_ g _
altez-native, the temperature standard
deviation value of each subarea can be used to
calculate a time series that is subjected to
FFT analysis tv extract the contributing
frequencies of each image of the series and
their relative amplitudes.
4. The computer selects the FFT spectra of
each of the subareas and displays colored
bitmaps of the relative amplitudes
derived from the FFT analysis in any
exhibited range of frequencies to thereby
identify clu$ters of subareas having
frequencies with abnormal amplitudes.
5. Tf procedure #4 does not identify definitely
aberrant clusters, the computer determines
that the test is negative for cancer and the
patient is normal. A message to this effect
is then output by the computer. Otherwise,
the computer proceeds with procedure #6.
6. The computer examines all the pixels i.n the
aberrant clusters identified in procedure #4
tv identify frequencies with amplitudes that
are Characteristic of cancer, i.e., at least
six congruent ~eubareas wherein a spatial
distribution of the amplitude value in the
congruent subareae is less than about 10%, and
preferably less than about 20%, yr more than
100% of the average amplitudes of the
temperature values of all of the surrounding
subareas.
7. If procedure #4 identifies a definitely
aberrant cluster, while procedure #6 turns out
negative, the computer prints out a color
image v~ the breast. If procedure #6 yield a
confirmation, the computer prints out another
color image with the aberrant clusters.
i
CA 02333406 2000-11-24
01!24%O1 11:11 F:~Y 613 234 3563 MacRae & Co. ~ 014
' - 10 -
B. Use of HST values and the computation of HST
TR.Fs .
1. The computerized camera takes on image of the
infrared flux (256 x 256 pixels) of an area of
interest of the skin, such as a human breast,
and Converts it into a thermal image where
each pixel has a certain temperature value_
This process is respected, preferably thirty
times a second until 1024 thermal images have
been accumulated and stored in the computer.
2. The infrared image is subdivided into subareas
of 4 to 16 pixels, each subarea corresponding
to about 25 to 100 square millimeters of skin,
based on the 256 x 256 optics resolution of
the camera. The computer then converts each
infrared image into a digital thermal image
where each pixel has a certain temperature
value.
3. The computer caJ.culates the average
temperature (AVT) value and temperature
standard deviation (SD) of each subarea. The
computer then calculates the HST value for
each subarea: HST = AVT/SD. The HST values
are analysed by FFT to extract the
contributing frequency of each subarea to
yield HST TRFs, and the relative amplitude Qf
each frequency.
4. The computer selects the FFT spectra.-'of each
of the subareas and displays colored bitmape
3~ of the relative amplitude derived from the FFT
analysis in any exhibited range of frequencies
to thereby identify clusters of subareas with
frequencies having abnormal amplitudes.
s. If procedure #4 d4es not identify definitely
aberrant clusters, the computer determines
that the test is negative for cancer and the
patient is normal. A message to this effect
CA 102333406 2000-11-24
01!24!01 __11:11 FAQ 613 234 3583 MacRae & Co. f~015
- 11 -
is then output by the computer. Otherwise,
the computer proceeds with procedure #6.
6. The computer examines all the subareas in the
aberrant clusters identified in procedure #4
to identify frequencies that are
characteristic of cancer, i.e., at least six
congruent subareas wherein a spatial
distribution of the amplitude value in the
congruent aubareas is less than about 10%, and
preferably less than about 20%, yr more than
100% of the amplitudes of the HST values of
all of the surrounding subareas.
7. If procedure #4 identifies a definitely
aberrant cluster, while procedure #6 tuz~ne out
negative, the computer prints out a color
image of the breast. If procedure #6 yields a
confirmation, the computer prints out another
color image with the aberrant clusters.
8. A match between the findings of procedure A
and B increases the diagnostic certainty of
the detection method of the present invention.
According to a further aspect of the present
invention, the difference between normal and cancerous
breasts is accentuated by a thermal challenge (cooling
or warming) of the breasts. Such a thermal challenge
affects only the neuronal thermoregulatory system and
therefore affects anly TRFs in areas that are not
vasodilated by excessive extravascular NO production.
The computer is programmed to look for the frequency
bands of the neuronal and the NO controlled TRFs in
every subset of pixels (e.g., 4 to 16 pixels) of the FFT
processed image. If the computer does not find any
clusters with neuronal TRFs having exceptionally low
amplitude (except in the periphery of the image which
does not depict skin), and no clusters are found to have
the NO controlled autacatalytic TRFs with a significant
CA 02333406 2000-11-24
01!24%O1 11:11 Fad 813 234 3583 MacRae & Co. 1~J016
- 12 -
amplitude, the findings of the test are declared as
negative (i.e., normal). This finding is then confirmed
by computing and analyzing the HST TRFs. If the
computer finds certain clusters with exceptionally low
neuronal TRF$ and especially if those clusters exhibit
the NO controlled autocatalytic TRFs, the test findings
are classified as pathological. This finding is then
confirmed by analyzing the HST data, as described for
the uncooled or unwarmed breast. Cooling or warming of
the breasts (by a mild flow of forced air? attains
maximal sensitivity and specificity. .Such additional
testing is administered as a confirmatory test only to
patients who show a positive result on the uncooled
test.
Although the invention has been described zn detail
for the purpose of illustration, it is to be understood
that such detail is solely for that purpose and that
variations can be made therein by those skilled in the
art without departing from the spirit and scope of the
invention except as it may be limited by the claims.
CA 02333406 2000-11-24
