Note: Descriptions are shown in the official language in which they were submitted.
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
. _1-
S~S~'EMS AND METHO,QS~~' MOLECULAR SPECTROSCOPY
TO PROVIDE FOR THE DIAGNOSI~~F TISSUE .
Related to U.S. Agslication
This application is a continuation-in-part of
"Systems and Methods of Molecular Spectroscopy to
Provide for the Diagnosis of Tissue", U.S. Serial
No. 07/661,077 filed on February 26, 1991 by Richard
P. Rava, Joseph J. Baraga, and Michael S. Feld, and
is incorporated herein by reference. This
application is also related to "Devices and Methods
For Optical Diagnosis of Tissue" filed on February
26, 1991 by G. Sargent Janes and Gary B. Hayes which
corresponds to U.S. Serial No. 07/661,072 and is
incorporated herein by reference.
Government Sun~rt
Funding for research conducted in connection
with the subject matter of the present application
was provided under NIH Grant No. RR 02594.
Backdy~~nd of the Invention
In the United States heart attacks, almost
entirely attributable to coronary atherosclerosis,
account for 20-25% of all deaths. Several medical
and surgical therapies are available for treatment
of atherosclerosis; however, at present no in situ
methods exist to provide information in advance as
to which lesions will progress despite a particular
~ medical therapy.
Objective clinical assessments of
atherosclerotic vessels are at present furnished
almost exclusively by angiography, which provides
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-2-
anatomical infarmation regarding plaque size and '
shape as well the degree of vessel stenosis. The
decision of whether an interventional procedure is '
necessary and the choice of appropriate treatment
modality is usually based on this information.
However, the histological and biochemical
composition of atherosclerotic plaques vary
considerably, depending on the stage of the plaque
and perhaps also reflecting the presence of multiple
etiologies. This variation may influence both the
prognosis of a given lesion as well as the success
of a given treatment. Such data, if available,
might significantly assist in the proper clinical
management of atherosclerotic plaques, as well as in
the development of a basic understanding of the
pathogenesis of atherosclerosis.
At present biochemical and histological data
regarding plaque composition can only be obtained
either after treatment, by analyzing removed
ZO material, or at autopsy. Plaque biopsy is
contraindicated due to the attendant risks involved
in removing sufficient arterial tissue for
laboratory analysis. Recognizing this limitation, a
number of researchers have investigated optical
spectroscopic methods as a means of assessing plaque
deposits. Such "optical biopsies" are non-
destructive, as they do not require removal of
tissue, and can be performed rapidly with optical
fibers and arterial catheters. With these methods,
the clinician can obtain, with little additional
risk to the patient, information that is necessary
to predict which lesions may progress and to select
the best treatment for a given lesion.
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-3-
= Among optical methods, most attention has
centered on ultraviolet and/or visible fluorescence.
~ Fluorescence spectroscopy has been utilized to
diagnose disease in a number of human tissues,
including arterial wall. In arterial wall,
fluorescence of the tissue has provided for the
characterization of normal and atherosclerotic
artery. However the information provided is limited
by the broad line width of fluorescence emission
signals. Furthermore, for the most part,
fluorescence based methods provide information about
the electronic structure of the constituent
molecules of the sample. There is a need for non-
destructive real time biopsy methods which provide
more complete and accurate biochemical and molecular
diagnostic information. This is true for
atherosclerosis as well as other diseases which
affect the other organs of the body.
Summary of the Invention
The present invention relates to vibrational
spectroscopic methods using Fourier transform
infrared (FT-IR) attenuated total reflectance (ATR)
and near-infrared (IR) FT-Raman spectroscopy. These
methods provide extensive molecular level
information about the pathogenesis of disease. Both
of these vibrational techniques are readily carried
out remotely using fiber optic probes. In
particular, a preferred embodiment utilizes FT-Raman
spectra of human artery for distinguishing normal
and atherosclerotic tissue. Near IR FT-Raman
spectroscopy can provide information about the
tissue state which is unavailable from fluorescence
CA 02497575 1992-O1-17
WO 92/I5008 PGT/US92/00420
-4-
methods. ,~n,situ vibrational spectroscopic
techniques allow probing of the molecular level
changes taking place during disease progression.
The information provided is used to guide the choice
of the correct treatment modality.
These methods include the steps of irradiating
the tissue to be diagnosed with radiation in the
infrared range of the electromagnetic spectrum,
detecting light emitted by the tissue at the same
frequency, or alternatively, within a range of
frequencies on one or both sides of the irradiating
light, and analyzing the detected light to diagnose
its condition. Both the Raman and ATR methods are
based on the acquisition of information about
molecular vibrations which occur in the range of
wavelengths between 3 and 300 microns. Note that
with respect to the use of Raman shifted light,
excitation wavelengths in the ultraviolet, visible
and infrared ranges can all produce diagnostically
useful information. Near IR FT-Raman spectroscopy
is ideally suited to the study of human tissue.
Raman spectroscopy is an important method in
the study of biological samples, in general because
of the ability of this method to obtain vibrational
spectroscopic information from any sample state
(gas, liquid or solid) and the weak interference
from the water Raman signal in the "fingerprint"
spectral region. The FT-spectrometer furnishes high
throughput and wavelength accuracy which might be
needed to obtain signals from tissue and measure
small frequency shifts that are taking place.
Finally, standard quartz optical fibers can be used ,
to excite and collect signals remotely.
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
_5_
Near IR FT-Raman spectroscopy provides the
capability to probe biological substituents many
hundred microns below the tissue surface. In
particular, for atherosclerotic tissue, calcified
deposits below the tissue surface are easily
discerned. Thus, it becomes possible to detect
pathologic conditions which would not be apparent
using angioscopic methods, as well as to study
the
detailed molecular basis of the pathology.
In contrast with electronic techniques, the
bands in a vibrational spectrum are relatively
narrow and easy to resolve. Vibrational bands
are
readily assigned to individual molecular groups.
The ATR technique offers several features
especially suited to sampling of human tissue
in vivo. Being a surface technique, the ATR method
can non-destructively probe internal human tissue
either by direct contact in a hollow organ (e.
g.
artery), or by insertion of a needle probe. In
the
mid-IR region, strong water absorption dominates
the
spectra of highly hydrated samples such as arterial
tissue, obscuring the absorption from other tissue
components (see Figure 8). Accurate subtraction
of
the strong water absorption from FT-IR ATR spectra
is relatively easy and very reliable with the
high
dynamic range, linearity, stability, and wavelength
precision of available FT spectrometers.
Furthermore, high quality mid-IR spectra of aqueous
protein solutions can be collected with fiber
optic
ATR probes. Such probes are easily adaptable to
existing catheters for remote, non-destructive
measurements vo. The mid-IR ATR technique
allows clinicians to gather precise histological
and
- CA 02497575 1992-O1-17
WO 92115008 PGT/US92/00420
-6-
biochemical data from a variety of tissues during
standard catheterization procedures with minimal
additional risk. -
The present methods relate to infrared methods
of spectroscopy of various types of tissue and
disease including cancerous and pre-cancerous
tissue, non-malignant tumors or lesions and
atherosclerotic human artery. Examples of
measurements on human artery generally illustrate
l0 the utility of these spectroscopic techniques for
clinical pathology. Results obtained demonstrate
that high quality, reproducible FT-IR ATR spectra of
human artery can be obtained with relative ease and
speed. In addition, molecular level details can be
reliably deduced from the spectra, and this
information can be used to determine the biochemical
composition of various tissues including the
concentration of molecular constituents that have
been precisely correlated with disease states to
provide accurate diagnosis.
Another preferred embodiment of the present
invention uses two or more diagnostic procedures
either simultaneously or sequentially collected to
provide for a more complete diagnosis. These
methods can include the use of fluorescence of
endogenous tissue, Raman shifted measurements
and/or ATR measurements.
Yet another preferred embodiment of the present
invention features a single stage spectrograph and
charge-coupled device (CCD) detector~to collect NIR
Raman spectra of the human artery. One particular
embodiment employs laser light in the 810 nm range
to illuminate the tissue and thereby provide Raman
CA 02497575 1992-O1-17
WO 92/15008 PGT/U892/00420
spectra having frequency components in a range
suitable for detection by the CCD. Other
wavelengths can be employed to optimize the
diagnostic information depending upon the particular
type of tissue and the type and stage of disease or
abnormality. Kaman spectra can be collected by the
CCD at two slightly different illumination
frequencies and are subtracted from one another to
remove broadband fluorescence light components and
thereby produce a high quality Kaman spectrum. The
high sensitivity of the CCD detector combined With
the spectra subtraction technique allow high quality
Kaman spectra to be produced in less than 1 second
with laser illumination intensity similar to that
for the FT-Kaman system also described herein.
Brief Description of the Drawings
Figures lA-1C axe schematic illustrations of
preferred systems fox providing the spectroscopic
measurements of the invention.
Figure 2 graphically illustrates FT-Kaman
spectra of human aorta: a) normal artery;
b) atheromatous plaque; c) FT-Kaman spectrum solid
cholesterol (Sigma).
Figure 3 graphically illustrates FT-Kaman
spectra of normal human aorta: a) irradiated from
intimal side (spectrum multiplied by 3); and b)
irradiated from adventitial side (primary adipose
tissue). c) NIR FT-Kaman spectrum of triglyceride,
triolein.
Figure 4 graphically illustrates FT-Kaman
spectra from human aorta: a) fibrous plague; and
CA 02497575 1992-O1-17
WO 92/15088 PGT/US92/00420
-g-
b) atheromatous plaque. c) FT-Roman spectrum of
cholesterol monohydrate powder.
Figure 5 graphically illustrates FT-Roman
spectra of calcified human aorta: a) calcified with
fibrous cap; b) excised calcification from a
different plaque; c) spectra of the same tissue as
in a) taken from adventitial side.
Figure 6 graphically illustrates FT-Roman
spectra of calcified human aorta: a) calcified
plaque with a fibrous cap (spectrum multiplied by
8); and b) exposed calcification.
Figure 7 graphically illustrates the measured
NIR Roman intensity of the 960 coil band (A(960 cm'1)
indicates the area of this band) in a calcified
deposit as a function of depth below the irradiated
surface. The dashed curve corresponds to the fit of
an exponential function to the data with an exponent
of 2.94 main.
Figure 8 graphically illustrates FT-IR ATR
spectra (4000 - 700 cm'') of (a) normal aorta,
intimal surface; and (b) buffered saline (0.14M
NaCl,pIi 7.4) .
Figure 9 graphically illustrates FT-IR ATR
spectra (1800 - 800 cm-1) after water subtraction:
(a) Normal aorta, intimal surface; (b) Sub-
adventitial fat; (c) Saline rinsed from the intimal
surface of normal aorta.
Figure 10 graphically illustrates FT-IR ATR
spectra (I800 - 800 cm''): (a) Two consecutive
water-subtracted spectra of normal aorta, intimal
surface, collected immediately after placement on
ATR element (solid line) and 10 minutes later
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-g-
(dashed line); (b) Same two spectra as in (a) after
lipid subtraction, scaled to have equal maxima.
Figure 11 graphically illustrates FT-IR ATR
spectra (1800 - 800 cm''), water-and lipid-
s subtracted: (a) Normal aorta, media layer; (b)
Atherosclerotic plaque, intimal surface; (c)
Atheromatous plaque with intact fibrous cap, intimal
surface.
Figure 12 graphically illustrates FT-IR ATR
1o spectra (1800 - 800 cm''): (a) Necrotic core of
atheromatous plaque, water-and lipid-subtracted;
(b) Dry film of cholesterol.
Figure 13 graphically illustrates scatter plot
for all samples of the area, A(1050), of the 1050
15 cm'' cholesterol band (integrated from 1075 to 1000
cm') ratioed to the area, A(1550) of the
1548 cni' protein amide II band (integrated from 1593
to 1485 cm''). The intensities were calculated from
the water-and lipid-subtracted spectra. NORMAL
20 denotes normal aorta specimens, intimal side,
FIBROUS includes atherosclerotic and atheromatous
plaques with intact fibrous caps, and NECROTIC
includes exposed necrotic atheroma cores and
necrotic material isolated from atheromatous
25 plaques.
Figure 14 graphically illustrates FT-IR ATR
spectra (1800 - 800 cm''): (a) second derivative
spectrum of normal aorta intima (Figure 8a); (b)
Water-subtracted spectrum of same normal aorta
30 intima specimen (same as Figure 9a).
Figure 15 graphically illustrates a scatter
diagram for all the specimens of the area, A(1050)
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00410
-10-
of the 1050 cm'1 cholesterol band plotted versus the
area, A(1382), of the 1382 cm's cholesterol band.
Both cholesterol bands have been normalized to the
area, A(1050), of the protein amide II band. All
band intensities were calculated from the water-and
lipid-subtracted spectra. Tissue categories are the
same as in Figure i3. The solid line represents a
linear least squares fit to the data.
Figures 16A and 16B are additional preferred
embodiments of ATR probes adapted to make the
diagnostic measurements of the present invention.
Figure 17 is a schematic diagram of the system
of Figure lA modified to use the spectrograph/CCD
Roman detector of the present invention.
I5 Figure 18 is a schematic diagram of a preferred
system for implementing the spectrograph/CCD Roman
detector of the present invention.
Figure 19 graphically illustrates
spectrograph/CCD-Roman spectra of normal human
aorta: A) Roman plus fluorescence spectrum produced
by illuminating the tissue sample with 810 nm laser
light; B) Roman difference spectrum produced by
subtracting spectra produced by illuminating the
tissue sample with 810 and 812 nm laser light; C)
resulting Roman spectrum produced by integrating the
Roman difference spectrum of B).
Figure 20 graphically illustrates
spectrograph/CCD-Roman spectra of an atherosclerotic
plaque with a calcified deposit exposed at the
surface: A) Roman plus fluorescence spectrum
produced by illuminating the tissue sample with 810
nm laser light; B) Roman difference spectrum
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-11-
' produced by subtracting spectra produced by
illuminating the tissue sample with 810 and 812 nm
' laser light; G) resulting Raman spectrum produced by
integrating the Raman difference spectrum of B).
Figure 21 graphically illustrates a
spectrograph/GCD-Raman spectrum of adventitial
adipose tissue.
Detailed Descrig ion
The spectroscopic methods of the present
invention can be performed on a system such as that
for laser treatment of atherosclerosis which is
illustrated in Figure 1A. Figure lA includes
separate block diagrams for the system of the
invention which utilizes laser light for
spectroscopic diagnosis as well as for treatment
and/or removal of tissue. The ablation laser 225,
pulse stretcher 229 and the pulse filler/multiplexer
231, 233 produce an output laser ablation pulse of
sufficient energy and intensity to remove tissue and
sufficient pulse duration to propagate through a
fiber optic laser catheter delivery system without
damaging the fibers. These systems and methods are
more fully described in co-pending application U.S.
Serial No. 07/644,202 filed on January 22, 1991,
which is incorporated herein by reference.
To remove plaque, a device 219 is used to
contact the tissue such as multiple-fiber laser
_ catheter 10 of Figure iB With an optical shield.
The catheter 10 is inserted into the artery and the
distal end of the catheter is brought into contact
with the lesion. Next, a determination is made as
to the type of tissue at which each optical fiber
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-12-
2oa-c' is aimed. Only fibers aimed at diseased
tissue are activated. Thus, selective tissue
removal is obtained. Furthermore, this technique is
also applicable for guiding surgical procedures in
other organs and tissue types such as the colon and
bladder.
The present invention relates to systems and
methods of performing spectral diagnostics to
diagnose the tissue in front of each fiber. A
preferred embodiment a laser light source 207 that
is coupled to the fibers. The diagnostic light is
sent to the fiber of choice by the optical fiber
selector 217.
The diagnostic light exits the selected optical
fiber and falls on the tissue. The tissue absorbs
the light and a fraction of the absorbed Light is
re-emitted, by Rayleigh fluorescence, Raman or other
elastic or inelastic light scattering processes.
This light is returned to the optical fibers and
exits from selector 217, and is detected by a
photodiode, photomultiplier or other detector 203.
Returning light could use different optical fibers
than those employed for illumination. Diagnostic
subsystem produces the entire spectral signal which
is coupled to computer 80.
The computer stores the information in a memory
as a spectrum, which is a graph of light intensity
vs. wavelength. This can be displayed immediately
on the video display 82 or compared to an existing
spectrum stored in the computer and the difference
displayed on the spectral display 86. Temporal
display 88 can display corrections made for the
wavelength dependent sensitivities of the source.
CA 02497575 1992-O1-17
WO 92/I5008 , PCT/US92/00420
-13-
' Information from either the temporal or spectral
display can be stored in the computer 80. The
' comparative data is shown on numerical display 84 to
provide a quantitative measure of the health of the
tissue observed.
With a multichannel detector and a computer, or
with appropriate multiple filters and detectors, it
is possible to gather this information in a fraction
of a second. Thus, a spectral or numerical display
is provided which indicates the type of tissue at
which the fiber of interest is aimed. If the tissue
is plaque, and is to be removed, then fiber selector
217 will align this fiber with the output beam of
the high power laser 225. Then, the high power
laser 225 is turned on and an appropriate power
level is selected for a predetermined amount of time
to remove a certain amount of diseased tissue. The
beam of laser 225 is transmitted to pulse stretcher
229 and pulse filler/multiplexer 231, 233 to
properly adjust the beam fluence.
The procedure is repeated for different fibers.
Where diseased tissue is detected, it is quickly
removed. The laser catheter 10 nibbles away at the
plaque, leaving the healthy artery wall intact.
If the artery 30 makes a bend 31 as shown by
Figure 1B, the laser catheter 10 will tend to make
contact with artery wall 32 at the outside wall of
the bend. To prevent the catheter from contacting
the artery wall, the optical fiber 20c is not fired.
The lesion is removed asymmetrically. This allows
the laser catheter 10 to follow the lumen 39, 39a
around the bend. Thus, the artery wall 32 is not
irradiated and is not perforated. Additional
CA 02497575 1992-O1-17
WO 9Z/15008 PGT/US92/00420
e.
-14-
details of this fiber optic catheter 10 are
disclosed in U.S. Patent No. 4,913,142, the contents
of which are incorporated herein by reference.
In both Raman and ATR methods, information is
contained in the spectral lines which are observed
in their intensities, and also their linewidths and
center frequencies (and how they change in different
environments). Further, Raman and infrared ATR have
different "selection rules". Some vibrations seen
to in infrared won't show up in Raman, and vice versa.
In other cases the same vibration can be detected by
both techniques, but with different relative
intensities (e. g. a strong Raman line will be weak
in ATR). So in this sense the two techniques
provide complementary information and combining the
two techniques (or using either or both with laser
induced fluorescence) is valuable in diagnosing
pathology.
The methods can utilize Fourier transform
detection to observe the radiation thereby providing
improved signal/noise ratios. Other techniques
(e.g, diode array detection and CCD detection) can
also be used.
As described in more detail below contributions
from major tissue constituents can be "subtracted
out" to reveal information about molecules which are
present in small concentrations. For example, in
ATR water contributions are removed before the "dry"
tissue constituents could be studied. Also, ,
3o derivative spectroscopy is used to eliminate
background signals and low frequency filters. Note ,
that these techniques deconvolute the observed
spectra into its individual constituents, an
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-15-
essential step for optimal extraction of diagnostic
information.
While Raman can sample deeply into tissue, ATR
samples only a very thin layer (a few microns).
Thus, ATR is "naturally" suited to probe surface
disease, such as the superficial cancers of the
bladder and GI tract, whereas Raman is well suited
to providing information about conditions deep
inside tissue (such as breast cancer or stones).
This is important for 3D imaging. Furthermore, the
ATR tissue sampling depth can be controlled by
properly matching the probe surface material to the
tissue type.
Generally, the ATR signal is very sensitive to
the surface of the waveguide or probe. For example,
if the probe surface has an affinity for lipids in
the tissue, lipids can migrate to the probe surface,
creating a thin lipid layer and producing a large
signal. This can be a problem (it can give
misleading information if not properly recognized
and guarded against). Conversely, it can be used to
advantage: Probes with special surfaces can be
developed to prevent this effect or to promote it,
in order to search for particular substances in the
tissue.
In a preferred method one can adjust depth
probed by ATR by varying refractive index of ATR
probe. Alternatively as discussed below one can use
a "waveguide needle" to get subsurface information.
Raman diagnostic methods permit adjustment of
Raman depth by varying the wavelength. Penetration
depth is wavelength dependent, and can be varied by
choosing different excitation wavelengths between
CA 02497575 1992-O1-17
WO 92/15008 PCf/US92/00420
-16-
about ~=700nm and 2~m. Another potentially
important way of adjusting Roman depth is to vary
the collection angle. In the near IR, incident
(exciting) light is strongly scattered out of the
forward direction into larger angles, so Roman
signals sampled at smaller angles come from tissue
closer to the surface. Therefore, the Roman
sampling depth can be controlled to a large extent
by probe design.
to Depth information is important if one desires
to provide imaging by creating 3D images of small
tumors in the brain or breast. Differential
techniques based on the ideas of the preceding
paragraph might allow accurate localization of such
tumors in three dimensions. Near-IR Roman can be
combined with a sound wave technique (time of flight
or standing waves set up in the tissue)--the sound
wave would modulate the Roman signal emanating from
a point in the tissue when it passes that point, and
the modulated signal could be used to establish the
depth of the tissue producing the signal.
A system employed for the collection of Roman
spectral data from excised tissue samples is
illustrated in Figure iC. FT-Roman spectra were
measured from 0 - 400ocm~ below the laser excitation
frequency with a FT-IR interferometer 40 equipped
with a FT-Roman accessory. The accessory employed
at 180 back scattering geometry and a cooled (77K)
InGaAs detector 42.
A 1064 nm CW Nd:YAG laser 44 can be used for
irradiating a sample 46: utilizing 500 mW to 1 W
laser power in a 1.0 to 2.5 mm spot 48 at the sample
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-17-
46 to collect Raman data. Alternatively, a pulsed
laser source can also be employed. Laser 44
generated a beam 46 that is directed through plasma
filter 48, mirrors 50, 52, focussing lens 54 and
mirror or prism 56 before irradiating the sample 46.
The radiation received by sample 46 undergoes
various mechanisms of absorption, reflection and
scattering including Raman scattering. Some of the
light emitted by the tissue is directed toward lens
6o and then through one or more Rayleigh filters 62.
The collecting lens 60 collects this backscattered
light 64 and collimates it. The Rayleigh filters 62
removes the elastically scattered light and
transmits the inelastically scattered, frequency
shifted (Raman) light. The Raman scattered light
enters the interferometer 40. No visible sample
degradation was observed under these conditions.
Excised human aorta was chosen of
atherosclerotic artery tissue. Samples were
obtained at the time of post mortem examination,
rinsed with isotonic saline solution (buffered at pH
7.4), snap-frozen in liquid nitrogen, and stored at
-85C until use. Prior to spectroscopic study,
samples were passively warmed to room temperature
and were kept moist with the isotonic saline.
Normal and atherosclerotic areas of tissue were
identified by gross inspection, separated, and
sliced into roughly 8x8 mm pieces.
The tissue samples were placed in a suprasil
quartz cuvette with a small amount of isotonic
saline to keep the tissue moist, with one surface in
contact with the irradiated by the laser 44. The
spectra shown in Figures 2 through 6 were collected
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-18-
with 512 scans at 8 cm'1 resolution (approximately 35
minutes total collection time).
Human aorta is composed of three distinct
layers: intima, media, and adventitia. The intima,
normally less than 300 ~m thick, is the innermost
layer and provides a non-thrombogenic surface for
blood flow. It is mainly composed of collagen
fibers and ground substance. The medial layer,
typically about 500 ~Cm thick, is quite elastic and
l0 serves to smooth the gulsatile blood flow from the
heart. The structural protein elastin is the major
component of aortic media, with some smooth muscle
cells present as well. The outermost adventitial
layer serves as a connective tissue network which
loosely anchors the vessel in place, and is mainly
made up of lipids, lipoproteins and collagen.
During the atherosclerotic process, the intima
thickens due to collagen proliferation, fatty
necrotic deposits accumulate under and within the
2o collagenous intima, and eventually, calcium builds
up, leading to calcium hydroxyapatite deposits in
the artery wall.
Figure 2a shows the FT-Raman spectrum of a full
thickness section of aorta grossly identified as
normal. Laser irradiation was on the intimal side.
The dominant bands appear at lfi5s cm'1 and 1452 cm'1
and can be assigned to an amide I backbone and C-H
in-plane bending vibration from proteins,
respectively. Weaker bands at 1331 and 1255 cm'1 are '
assigned to C-H wagging and amide III vibrations
from proteins, respectively. The frequencies of '
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92100420
-19-
amide I and III are consistent with those observed
for disordered proteins.
Another example of a typical NIR FT Raman
spectrum from normal aorta is shown in Figure 3.
when irradiated from the intimal side, Figure 3a,
the major vibrational bands observed in normal aorta
are all attributable to protein vibrations: the
band at 1658 cm'' is assigned to the amide I
vibration of the polypeptide chain, the 1453 cml
l0 band to a C-H bending mode of proteins, and the 1252
cm'' band to the amide III vibration. The spectrum
of normal aorta is at least 25% weaker than any of
the pathologic samples. The peak frequency of the
C-H bending band, which averaged for all the normal
specimens is 1451~1 cm'', is specific to the protein
C-H bending mode (See below). The weak band near
1335 cm"', which appears as a shoulder in many of the
normal specimens, appears to be specific to elastin,
and the weak band at 1004 cm"' is likely due to
phenylalanine residues. In general, the relative
intensities of the bands in the region between 1250
and 1340 cm'' appears very much like that observed in
the FT Raman spectrum of elastin. This observation
is consistent with the thin collagenous intima in
normal aorta, the elastic nature of the media of
aorta, and the deep penetration depth of 1064 nm
radiation. Band assignments for all tissue types
presented here are listed in Table 2.
Figure 3b displays the NIR FT Raman spectrum of
3o the adventitial side of normal aorta. In this case,
the irradiated adventitial surface consisted of
several millimeters of visible adipose tissue. In
~
.
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-20-
contrast with the spectrum collected from the
intimal side, the bands observed in this adipose
material appear to be mainly due to lipid, and in .
particular triglyceride, with almost no contribution
from protein. This is not unexpected, as the
triglyceride content of adipose tissue is on the
order of 60%. The sharp band at 1655 cml is due to
stretching of C=C groups in unsaturated fatty acid
chains. This band is distinguished from amide I by
l0 its peak frequency and its width, which in this case
is 17 cml FWHM. Amide I, in contrast, is roughly 60
cni' wide. The dominant C-H bending band is shifted
to 1440 cm'', characteristic of lipids. This band is
about 3 times more intense in adipose tissue than in
normal intima, probably a result of the greater
number of C-H groups per unit volume in
triglycerides. The bands as 130i/1267 cml and 1080
cm'I are assigned to C-H bending and C-C stretching
vibrations of fatty acids, respectively.
The 1746 cm'I band, assigned to the C~
stretching mode of the triglyceride ester linkages,
indicates that most of the lipid observed in the
adventitial adipose tissue is of the triglyceride
form. Specifically, the integrated intensity of
this band relative to the C-H bending band at 1440
cm'1 is equal to 0.103, while this same ratio for
triolein is 0.136, which indicates that roughly 75%
of the C-H band is due to triglyceride. The NIR FT
Kaman spectrum of triolein (a triglyceride
containing fatty acid chains of 18 carbons and a
single double bond) is shown for comparison in
Figure 3c. Additional molecular information
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-21-
regarding the state of the fatty acid chains is
readily deduced from the adventitial adipose
spectrum. For example, the relative intensity of
the C=C band at 1655 cm'' indicates that there are on
average roughly 0.7 unsaturated double bonds per
fatty acid chain, assuming 16-18 carbon fatty acids.
In addition, the frequencies and structures of the
C-H bending and C-C stretching bands suggest that
most of the fatty acid chains are in the gauche
conformation. The sharp 1129 cm'' band,
characteristic of all-traps chains, is not observed
in the spectrum.
The FT-Raman spectrum obtained from a diseased
artery, an atheromatous plaque, with a thick fibrous
connective tissue cap and an underlying necrotic
core is shown in Figure 2b. The necrotic core of an
atheromatous plaque contains cellular debris as well
as large accumulations of oxidized lipids and
cholesterol. The band in the amide I region,
peaking at 1665 cm'', is distinctly narrower in this
spectrum compared to normal aorta. In addition, the
in-plane C-H bend, at 1444 cm'', is relatively more
intense and has a distinct shoulder at higher
frequency. The two weaker bands at 1307 and
1267 cm'' are shifted in frequency from those found
in the spectrum of normal aorta. The band
frequencies and shapes in the FT-Raman spectrum of
cholesterol, shown in Figure 2c, coincide with some
of those observed in the atheromatous plaque,
consistent with the expected composition of the
necrotic core.
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-22-
The NIR FT Raman spectra of other fibrous
plaque specimens exhibit a range of features as
shown in Figures 4 and 5. Figure 4a shows a
representative spectrum from one of the types of
fibrous plaques. These fibrous plague spectra are
quite similar in both relative and absolute band
intensities to the spectra of normal aorta. The
most significant differences are that the C-H
bending band, peaking near 1448 cm~ on average, is
shifted by 2 cm'1 to a slightly lower frequency.
This may be the result of a small increase in the
lipid content of these plaques relative to normal
aorta. In addition, the band near 1340 cm'',
attributed to elastin in normal aorta, is decreased
relative to amide III at 1265 cm's. The putative
explanation is that the collagenous intima is
thickened in these specimens, so that the spectral
contribution from the elastic media is reduced
relative to that of normal aorta.
The NIR FT Raman spectra of other fibrous
plaque specimens appeared similar to atheromatous
plaques' spectra (Figure 2b). These are
substantially different than either normal aorta, or
adipose tissue. In these cases, the intense C-H
bending band occurs at 1440 cm'', characteristic of
lipid material. This band is roughly twice as
intense as the C-H bending band in normal aorta.
The complete absence of a band at 1746 cm~ indicates
that this lipid is not triglyceride. In fact, this
lipid appears to be predominantly cholesterols, as
identified by the sharp, characteristic band at 700
cm'' and comparison to the cholesterol spectrum shown
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-23-
in Figure 4c. Again, this is not surprising, since
cholesterols accumulate in high concentrations in
' atherosclerotic lesions. Several of the bands
between 1000 and 500 cm's are assignable to
vibrational modes of the sterol rings. These
include the bands at 959, 882, 844, 805, 700, 605,
and 546 cm''. In addition, the 1666 cm'' band is
attributed in part to the C=C stretching vibration
of the steroid nucleus.
The presence of fatty acid chains in the
atheromatous plaque spectra is evidenced by bands at
1300/1262 cm'1 and 1130/1088 cm'1, due to C-H bending
and C-C stretching vibrations, respectively. These
bands may contain contributions from cholesterol as
well. The relative intensities of the fatty acid
band at 1300 cm'1 and the sterol ring bands suggest a
mixture of free cholesterol and cholesterol-fatty
acid esters. Moreover, the relative intensities of
the 1130 cm'1 C-C stretching and the 700 cm'1 sterol
bands indicate that most of the fatty acid chains
are in the gauche conformation, consistent with the
predominance of unsaturated fatty acid chains in the
cholesterol esters in these plaques. It is
particularly noteworthy in the atheromatous plaques
that the cholesterol deposits are detected from
material below a thick fibrous cap indicating the
ability of the NIR FT Raman method to probe
materials several hundred microns below the tissue
surface.
In addition to the cholesterol and cholesterol
ester bands, the NIR FT Raman spectra of some of the
fibrous plaques contained two unique bands, at 1519
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92100420
-24-
and 1157 cm''. The intensities of these bands are
highly correlated, which suggests that they are due
to a single component. These bands, which have been
previously observed in visibly-excited Raman spectra
5 of atherosclerotic plaques, are assigned to
carotenoids. The amount of carotenoid in these
plaques is probably much smaller than the amounts of
cholesterols or proteins, but may be strongly pre-
resonance enhanced (14). The carotenoid bands are
to observed only in this subset of fibrous plaques.
In an advanced plaque, calcium may begin to
accumulate leading to the formation of calcium
hydroxyapatite crystals in the tissue as shown by
the spectra of Figures 5 and 6. The FT-Raman
15 spectrum of a calcified plaque with a thick (several
hundred microns) fibrous connective tissue cap
overlying a calcified deposit is shown in Figure 5a.
The spectrum clearly indicates bands due to the
protein in the fibrous cap, amide I and III at 1665
20 and 1255 c~i', respectively. However, additional
bands are observed between 1250 and 1350 cml and
around 1100 cm's, as well as a strikingly sharp band
at 961 cm'I. The latter is readily assigned to the
symmetric phosphate stretching vibration associated
25 with phosphate groups in the calcium hydroxyapatite
deposits, while the band around 1100 cail is an
asymmetric phosphate stretch. These assignments are
confirmed by excising the solid "rock" from a
different calcified plaque, and obtaining its
3o spectrum as shown in Figure 5b. A strong Raman
signal from the phosphate stretching vibration in ,
solid calcifications in advanced atherosclerotic
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-25-
plaques can also be observed utilizing standard
visible Raman instrumentation. The ability to
detect the calcifications several hundred microns
below the tissue surface when using near IR FT-Raman
spectroscopy, however, is a diagnostic measurement
which can be utilized to guide treatment.
A measurement was attempted to determine if the
calcification might be detected when the tissue was
irradiated from the adventitial side. The resulting
FT-Raman spectrum is shown in Figure 5c. No
evidence of the strong phosphate vibration is
apparent. In contrast, sharp vibrational bands at
1745, 1656, 1444, 1303, 1267 and 1082 cm's are
observed which are mainly associated with the lipid
material that constitutes the majority of the
adventitia.
The NIR FT Raman spectrum of calcified plaque,
containing a subsurface calcified deposit and an
overlying soft fibrous cap, exhibits an intense,
sharp, new band at 960 cm'' (Figure 6a). This band,
specific to calcified tissue, is assigned to the
symmetric stretching vibration of phosphate groups
(15), which are present in high concentrations in
the solid calcium salts. The weaker phosphate
antisymmetric stretch is also present at 1072 cm''.
A symmetric stretching vibration of carbonate groups
may also contribute to this latter band. The
phosphate vibrations are easily observed from
subsurface deposits in the calcified plaques: the
~ 30 960 cm'' band can be observed from deposits up to 1.5
mm beneath a soft tissue cap with the current
~ signal-to-noise level (See below). The calcified
CA 02497575 1992-O1-17
. ~r-~---
WO 92/15008 PGT/US92/00420
-26-
plaque also displays protein vibrations from the .
fibrous tissue cap. These include amide I at 1664
cml and amide III near 1257 cm'1. The C-H bending
band at 1447 cm'i suggests a mixture of protein and
lipid, and the weak band at 699 cm'' is likely due to
cholesterol that is either in the fibrous cap, the
calcified deposit, or both.
The NIR FT Raman spectra of exposed
calcifications (Figure 6b) display a range of
features. In all cases, the major bands are due to
the calcium salts. These include the 960 cml
phosphate and 1072 cml phosphate/carbonate bands as
well as the band at 587 cm', which is assigned to
another phosphate vibrational mode. On the other
hand, several differences are apparent in the weaker
bands, which are presumably due to soft tissue
components which are embedded in the calcification.
In some cases (not shown), the C-H bending band
occurs at 1450 cml, and the band at 1663 cm'1 is
similar in shape to amide I for some of the
calcifications, indicative of protein vibrational
modes. In other calcified plaques, such as that in
Figure 5b, the C-H bending band occurs at 1440 cm'I,
and the 1667 cm's band, which is much sharper, is
more like due to C=C stretching vibrations. In this
plaque, the bands are due to lipid, in particular
cholesterols, as evidenced by the 700 cm'' and 1300
cm' bands.
In our histological examinations of aorta, two
distinct types of exposed calcifications have been
noted. In one type, the fibrous tissue cap is
calcified. In the other, the necrotic core of an
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-27-
atheromatous plaque is calcified, and the calcified
deposit is exposed by ulceration of the soft tissue
fibrous cap. A positive explanation for the two
' spectral types of exposed calcifications is that the
specimens which exhibit protein bands are of the
former histologic type, while the specimens which
exhibit lipid bands are of the latter type.
The present methods provide an IR FT-Raman
technique for differentiating various stages of
atherosclerosis in human aorta. They demonstrate
that molecular level information is available using
these methods. This information is useful for
following the pathogenesis of the disease and in
guiding the treatment of different lesions. The
near IR FT-Raman method, with its relatively deep
penetration depth, is able to obtain spectroscopic
signals from below the tissue surface, yielding
details about the atheromatous necrotic tissue and
sub-surface calcifications. These signals can be
utilized with an optical fiber based imaging system
to determine the content and composition of
different atherosclerotic plaques in vivo.
With the observation that several of the
biochemical species important in the atherosclerotic
process, including cholesterol and calcium
hydroxyapatite, can be easily detected below the
tissue surface, we wished to determine the depth
limit of detection using the NIR FT Raman technique.
In order to address this question, ten 200 um
sections of aortic media were cut and placed one at
a time over a large calcified deposit (6x6x3 mm),
and the FT Raman spectra of the 960 cm'i band
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-28-
monitored as a function of depth below the surface. _
As indicated by the plot of FT Roman intensity
versus depth shown in Figure 7, the signal from the
calcified deposit was detectable until the deposit
was greater than 1.6 mm below the irradiated
surface. Even slightly deeper depths could be
probed if the focus of the collection optics was
moved into the tissue.
The two dimensional resolution of the NIR FT
Roman signal for material below the tissue surface
was then tested by glacing 1 mm of aortic media
above another calcified deposit, and moving the
tissue transversely in two dimensions through the
laser beam and collection lens. The FT Roman signal
i5 was observed to drop-off rapidly as the beam and
collection optics moved from the calcified deposit.
The detected FT Roman signal closely followed the
geometry of the calcified deposit below the surface,
despite the significant scattering of the overlying
20 layer of tissue. This result suggest that the Roman
scattered light may be utilized for imaging objects
below the tissue surface with minimal image blurring
due to elastic scattering in the tissue.
A second spectroscopic method is also used to
25 obtain molecular vibration information, attenuated
total reflective (ATR) of infrared light.
Human aorta was chosen as an example to
illustrate the diagnosis of atherosclerotic artery
tissue. As in the samples obtained for the Roman
3o spectral measurements human aorta samples were
obtained for ATR measurements at the time of post
mortem examination. Sample storage and preparation -
procedures are identical to those set forth for the
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-29-
Raman spectral measurements. These reflectance
measurements can be used by themselves to provide
diagnostic data in conjunction with either the Raman
spectroscopic measurements described above or with
fluorescence measurements, or with both types of
measurements to enhance diagnosis for specific
applications.
The medial layers of a normal arteries and the
necrotic cores of atheromatous plaques were exposed
by blunt dissection and spectroscopically examined.
ATR spectra were also collected from several
purified tissue components including collagen,
elastin, and cholesterol to assist in analysis of
the spectra.
Mid-infrared ATR spectra were measured from
4000 to 700 cm's with a commercially available FT-IR
spectrometer and a horizontal ATR accessory. The
sampling area was purged with dry nitrogen gas to
control background absorption from atmospheric water
vapor and COi. Spectra were collected at 4 cm's
resolution with 50 scans. The artery specimens,
kept physiologically moist with isotonic saline
(buffered at pH 7.4), were placed in contact with
the ATR element (ZnSe crystal 45 ends). A 5 gram
weight placed on the tissue sample ensured uniform
sample contact with the ATR element. An ATR
spectrum of the saline solution with absorbance
similar to that of the artery samples was also
. obtained and used for subtraction of spectral
components due to water. Collagen (Calbiochem:
. type I, bovine achilles tendon) and elastin (Sigma:
bovine neck ligament) were prepared as saline
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-30-
slurries. Cholesterol (Sigma) was prepared as a dry
film on the ATR element by evaporation of a benzene
solution. These elements can be clearly identified
in the resulting spectra.
The ATR sampling crystal is a rod of high
refractive index material which acts as a waveguide
for the infrared sampling beam. This waveguide can
be in the form of a needle that is adapted for
penetration into the tissue to be diagnosed.
l0 Alternatively, the probe will have a geometry
suitable for contacting the surface of exposed
tissue sites or for contacting internal locations
with a catheter.
The devices shown in Figures 16A and 16B
illustrate preferred embodiments of the invention
adapted for ATR diagnostic measurements within the
human body. In Figure 16A a single-ended probe 100
is shown where one or more optical fibers 102 both
the incident light to, and the transmitted
(reflected) light from, the ATR element 104. A 100$
infrared reflector 106 such as gold is placed at the
distal surface 108 of the ATR element 104 functions
to return the transmitted light back through the
same fiber as well as to provide double pass
sampling. The ATR element 104 can be a separate
component optically fastened to the optical fibers
102, or alternatively, it can be constructed from
the end of the optical fiber by removing the
cladding material. Sampling is provided by placing
the ATR element in contact with the tissue 110 of
interest. Radiation is transmitted 112 and
collected 114 in a radial direction from element
104. The probe can either be inserted through a
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92100420
-31-
standard endoscope or catheter to sample a hollow
organ, or, if made with sufficiently thin optical
fiber, it can be directly inserted directly into a
solid organ as in the case of needle biopsy. In
this particular embodiment the distal tip 108 is in
the form of a needle. The cone or needle
configuration on the end of the catheter can be long
or shallow.
A double-ended probe is illustrated in Figure
16B. Incident IR beam from FT-IR is transmitted
through IR optical fiber 122 to ATR element 128
positioned at the distal end of catheter body 120.
The ATR element is placed in contact with tissue 126
surface to be sampled. Transmitted light is
conducted through a second IR optical fiber 124 back
to an IR detector. The ATR element may be a
separate component optically fastened to the two
optical fibers 122, 124, or it may be simply a
region of a single optical fiber in which the fiber
cladding material has been removed. The entire
apparatus can be inserted through a standard
endoscope or outer catheter.
For methods of measuring excised samples, the
specimen to be sampled is placed in optical contact
with the surface of the waveguide or ATR element.
The evanescent wave which extends outside of the
waveguide surface is absorbed by the sample in
proportion to its absorption coefficient. The
penetration depth of the evanescent wave into the
' 30 sample depends on the wavelength of the infrared
radiation and the refractive indices of the
' waveguide and the sample; for a ZnSe-water
interface, this depth is roughly 1 ~cm from 1800 to
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-32-
700 cm'1. The 1/e penetration depth of the
evanescent wave into the sample is given by
~/2~r (nZ sin28-nW2)'~, where a is wavelength, a is angle
of incidence and nZ and nW are the refractive
indices of ZnSe and water respectively.
Consequently, only tissue that is in good optical
contact with the ATR element will be sampled. In
addition, individual components in the sample can
exhibit different affinities for the waveguide
to material (ZnSe in this case), which can influence
the relative concentrations of the components at the
waveguide surface. Despite relatively high
concentrations in the bulk tissue, components with
poor optical contact can be difficult to measure in
the ATR spectrum.
Figure 8 shows FT-IR ATR spectra of (a) normal
aorta (intimal side) and (b) buffered saline. A
comparison of these spectra shows that a majority of
the IR absorption of normal intima can be attributed
to water, which comprises roughly 80% of the tissue
by weight. The large, broad bands peaking at 3300
cml and 1636 cm'' axe due to the O-H stretching and
H-O-H bending vibrations, respectively, of water,
and the weak band at 2120 cm1 is due to a water
combination vibration. The 3300 cm's and 1636 cm'1
bands also include contributions from the N-H
stretching and amide I vibrations. The relatively
flat absorption between 1500 and 900 cm'' and the
rising absorption below 900 cm'1 is also due
primarily to water; however, in the intima, a number
of very weak bands due to other tissue components
are also present in this region.
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-33-
Most biomolecules give rise to IR absorption
bands between 1800 and 700 cm'', which is known as
the "fingerprint region" or primary absorption
region. The dominant absorption of tissue water in
this region obscures the absorption bands from other
tissue components. To observe the IR bands from
these components, one must eliminate the water
interference. With the ATR method, spectral
deconvolution or subtraction of the water component
is particularly easy. By using the 2120 cm'' band,
which is due solely to water, as an internal
intensity standard the spectrum of buffered saline
(Figure 8b) can be accurately and reliably
subtracted from the spectrum of aorta intima (Figure
8a), yielding a water-subtracted spectrum of intima
(Figure 9a).
In the water-subtracted spectrum, the
previously weak bands are easily observed. Band
assignments, based on the spectra of the major
tissue components are listed in Table I.
Most of the vibrational bands observed in the
spectrum of normal intima (Figure 9a) can be divided
into two broad categories: lipid-associated bands
and protein-associated bands. All of the strong
bands in normal intima are associated with one of
these moieties (see Table I). This can be seen as a
simple consequence of the large concentrations of
these two materials. Aside from water, a large
fraction of tissue can be divided into one of these
two groups. Moveover, both protein and lipid
components contain repeating molecular units which
are common to all members of the group. For
- CA 02497575 1992-O1-17
WO 91/15008 PGT/US92/00420
-33.1-
Table I. Preliminary assignments of IR absorption peaks in
the ATR spectra of normal aorta intima.
v Preliminary vibrational Associated Tissue
~(+icm'tlAssi~ent Coy ne is
2923(x) C-H stretch Lipid, Protein, Others
2853(x) C H stretch Lipid, Protein, Others
1744(x) C=O (ester) stretch Lipid
1651(x) Amide I Protein
1635(sh) Amide I, H-O-H bend Protein, Water
1548(x) Amide II Protein
1465(s) CH2 bend Lipid
1457 (s) CHZ bend, CH3 anti- Lipid
symmetric deformation
1454(x) CH bend, CH3 anti- Protein, others
symmetric deformation
1417(w) CHz bend adaacent to C=O Lipid
1401(m) COO' symmetric stretch, Protein, others
amide C-N stretch
1378 (w) CH3 symmetric Lipid
deformation
1244(m) Amide III, POZ anti- Protein, others
symmetric stretch
1239 (m) CHZ wag, POz anti- Lipid
symmetric stretch
1159 (s) CH= wag, C-O-C Lipid
antisymmetric stretch
1117(w) C-C stretch, O-C-O Lipid
stretch
1096(w) Lipid
1083(w) POZ symmetric stretch Protein, others
1030(w) Lipid
965(w) C=CH deformation (traps) Lipid
722 (m) CHZ rock Lipid
SUBSTtTVTE SHEET
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-33.2-
w
to
O
ro b
..
M .~. .O ~ x ,y -~i t0,a .C
N
~ d u ~ : - m x -- v
~
y ~ . . ~ o r ~ o d
,. a
a
p 01 i~ 01 ~ O a ~, H ~ 'C3
U 'O U "~ .~ "
p UJ 1.1 1-1~ ~ b 'CS ~ Z) H ~ i1
i~ ~-I ~ N ~ N rl
d ~ ~ '"~ y
,- ~ O.
~
d -~ .4 N ' N ~ ~ U ~ ~ N
a 0 ~ ~'' d R U Q " a
tl
~
V IA O U .,iV ~ ~.l ~. 'db V
f0 ~.1 ,
..~i w-i
"
a U V a V U x ~7 v.7
~~., ~
UW
OJ
ro
o ~
w
- o
ro
ty er e1 N
U
s
v
s~
, O
N
0
O
V
~D ~0 V' N rl
N H
H H ~"~ H ri
W U
.Q
X O ~ N O
3
01 ~ 0
ro D
W ~ N
) ~ e ri e rt
0 ~i -i ~
H
0l
N
d
~ O
O O ~ 1
a, 0
N .tZ 1p In sf c'f N e-1 v-1
~ O
~ w ~ ...~..
~
.
' ro
o~
o ~ ~
o
w ~ w o ~ ~r o vo 00
~ ~ ~ ~
.
ro
w
H ~ t0 N.
i1
W
D er N
x ~ .~ .a
ro
E
SUBSTITUTE SHEET
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-33.3-
11
m
a~ a o cr
N
d b 41 p ~ ~ ~~ p N
fpm ~y V N.-.
"ar ~' a ~ ~ ~ +~ ~ .G c
o.~ o ~n ~n o
~
~
.' , ' ~ ~~~ ~v ~~~
c >,'~~
~~'
m ~i m~ d N~ o .c~ R
~
~ ~ y~ ~
aU G U " V >
r d
tr
H
N N 3
~
1
0
DO~0
'
!~ (t CD CJ
X O ~ 1 O 01 d' u~
U 00 0o a~
to tn
W
U
+I
b
-. _ -
b~ ~
~
.
d N vi R
w
' c1 ~-a
0 ~
~
f
N
V
G
5, 3 ~
m i.n
C' 0 t~D N ~ ~
O t30 t~0 x
~ f,0"~ 01 00 0D
r~i CO n t00
~
~~a
U
b
m
~!' ~ I~NNt"'f0~010 N
~
~
r
~ O 1~1C0d'OOOef' t
RS 01
f
CD O O tWP
I
f
W O
~
.~
-ri
'~"
_-_ _ -.
w
~
~
Ki a
C
..-1~
~
V ~ U
i
r
N'
G
+~
~u
~
,., 3 tr
N
y
N ~ ~ Li
w-rl
W
rN-1O O ~
C
Rf
la
'~
H ~'
~'
a
~
~a
.
et tcteTiT! ITF S~IEE~'
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-34-
protein, the polypeptide backbone of repeating amide
groups is the dominant element. In lipids, the
' repeating hydrocarbon chain is the defining quality.
The end result is that these molecular units are
present in very large concentrations, and their
vibrational bands tend to dominate the spectrum.
Note that this does not imply that no further level
of detail is derivable from the IR spectrum of
tissue. For example, the frequencies of the amide
l0 group vibrations are sensitive to protein
configuration and conformation. Therefore, shifts
in protein makeup might be expected to produce
observable changes in the amide bands.
The water-subtracted spectrum of sub-
adventitial fat shown in Figure 9b, more clearly
illustrates the division of bands into lipid and
non-lipid categories. This fat can be considered as
the model of the lipid component. Protein
contributions, as judged from the intensities of the
amide I and II bands, are negligible for the
purposes of this model. All of the bands observed
in the fat spectrum can be attributed to the lipid
component. These include the strong bands at 1744
cm'1 ( C=O stretch) , 14 65 cm'1 ( C-H bend) , 1161 cm'1 ( CHT
wag, C-O-C stretch), as well as the weaker bands at
1378 cm'', 1239 cm's, 1118 cm'1, 1099 cm's, 966 cm'i, and
722 cm'' .
The bands observed in the water-subtracted
spectrum of intima constitute less than 30% of the
total absorption, the rest being due to water. Any
conclusions regarding these relatively weak bands
depends critically upon the accuracy of the water
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-35-
substraction. The accuracy of this subtraction can
be judged from the reproducibility of spectra .
obtained sequentially from the same sample. Two
consecutive water-subtracted spectra collected 10
minutes apart from a sample of normal aorta (intimal
side) are shown in Figure l0a (solid and dashed
curves). Most of the IR bands exhibit a substantial
increase in absorbance with time. This trend
continues for consecutive spectra collected more
than an hour after the placement of the sample on
the ATR element. However, not all of the bands
change by the same fraction, so that the relative
intensities differ between consecutive spectra. For
instance, in Figure 10a, the C=O band at 1744 cmi is
relatively constant, while the amide bands at
1650 cm'1 and 1547 cm'i increase by 50% in the later
spectrum. Although these changes might seem to
indicate that the water subtraction is inaccurate,
the changes with time are systematic and
predictable, which suggests that the optical contact
between the sample and the ATR element is changing
regularly With time.
In fact, the constancy of the 1744 cml C~
band, which is due solely to lipid, and the
increases in the amide bands, which are due to
protein, indicate that the lipid contributions to
the IR absorption remain unchanged while the non-
lipid contributions increase between consecutive
scans. This is confirmed by subtracting the
spectrum of lipid (Figure 9b) from the water-
subtracted spectra of aorta intima (Figure 10a),
using the 1744 coil band for normalization. The
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420 '
-36-
resulting lipid-subtracted spectra of aorta intima
are shown, normalized to peak absorbance, in Figure
10b. As can be seen, the relative peak absorbencies
and spectral bandshapes in the lipid- subtracted
spectra reproduce quite well, reflecting the
accuracy of both the water and the lipid-
subtraction procedures.
The constancy of the lipid bands and the
variation of the non-lipid bands between successive
scans may seem somewhat puzzling. An explanation of
this apparent anomaly can be inferred from a water-
subtracted spectrum of saline solution which is
rinsed off the surface of the tissue (Figure 9c).
This spectrum, aside from the weak amide I and II
bands, matches quite closely with that of
adventitial fat. The lipid component observed in
the tissue appears to be due to free lipid particles
that have equilibrated with the tissue surface
water, forming a thin water-lipid film on the tissue
surface which is in full optical contact with the
ATR element immediately after the tissue specimen is
placed upon the crystal. The tissue components
beneath this film presumably achieve better optical
contact with the ATR crystal as the sample settles.
As a result, the content of lipid in a spectrum of
aorta intima or media may be influenced by the
presence of sub-adventitial fat in the specimen, and
the relative lipid-protein absorbencies are accurate
to 50% at best with the present experimental design.
For the reason, all of the remaining spectra shown
are both water and lipid subtracted.
With the lipid bands removed, assessment of the
non-lipid bands in the spectrum of normal intima
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-37-
(Figure lOb} is greatly simplified. The major bands
in the spectrum may be assigned to protein backbone
vibrations. These include the bands at 1648 cm''
(amide I), 1549 cm'1 (amide II), 1455 cm'1 (C-H bend),
1401 cm'' (amide C-N stretch} , and 1244 cm'1 (amide
III}. The frequency of the amide I peak (1648 cm's),
which is sensitive to protein secondary structure,
may indicate contributions from a-helix, disordered,
and collagen helix conformations. This band also
exhibits a shoulder near 1634 cm'', which may be due
to the ~-sheet regions of proteins or water. The
protein C-H bending band at 1455 cm'i is distinct
from the corresponding vibration in lipid, which
occurs as a double-peaked band at 1465/1457 crn~.
Note that all of these bands may include
contributions from other moieties. For instance,
the symmetric stretch of carboxylate groups and the
antisymmetric stretch of phosphate groups may also
contribute, respectively, to the 1401 cm'1 and 1244
cm'i bands. This correlation of components is
summarized in Table I above.
A typical spectrum of the medial layer of
nonaal aorta is shown in Figure 11a. A comparison
of this spectrum to that of normal intima (Figure
lOb) fails to reveal any significant differences.
This result is somewhat surprising, considering that
normal aorta intima and media have significantly
different compositions. Typical spectra of an
atherosclerotic plaque and a non-ulcerated -
atheromatous plaque are shown in Figures llb and
llc, respectively. For these plaques, only the
intact fibrous cap at the intimal surface is probed
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-38-
' due to the short penetration depth (1 ~cm) of the
beam. Any necrotic, atheromatous material beneath
' this fibrous cap is not sampled. Even so, the
fibrous caps of these plaques are known to be
compositionally different than normal intima and one
might expect these differences to be reflected in
the IR ATR spectrum. However, as in the case of
media, no consistent differences are observed in the
spectra of these plaques (Figures llb and ilc) and
normal intima (Figure lOb). This issue will
expanded upon in the discussion below.
On the other hand, substantial differences are
obvious in the spectrum of the necrotic,
atheromatous core of an atheromatous plaque (Figure
12a) as compared with the corresponding spectra of
normal intima (Figure lOb) as well as those of
intact atherosclerotic (Figure lib) and atheromatous
(Figure lic) plaques. In this case, the necrotic
core was presumably exposed in vivo as disease
progressed by ulceration of the overlying intimal
fibrous tissue cap. (The spectrum of necrotic core
exposed by dissecting away the fibrous cap of a non-
ulcerated atheromatous plaque is similar.) A new
band appears at 1050 cm'', with a secondary peak at
1023 cm''. In addition, the necrotic core spectrum
exhibits an increase and frequency shift in the 1466
cm-' band as compared with the 1455 cm' protein band
in normal intima as well as a set of unique bands
near 1382 cm''. These characteristic bands are found
in the spectra of all the exposed necrotic core
. samples and in none of the other samples (see
below).
,~.._ CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-39-
The source of these unique bands in the
necrotic core spectra may be cholesterol, which is
known to accumulate in large amounts in atheromatous '
cores. An ATR spectrum of cholesterol (dry film) is
shown in Figure i2b. The three major bands unique
to the necrotic core, near 1463 cm', 1382 cm'', and
1050 em', match closely in position and relative
intensities with the three main cholesterol bands at
1466 cm'', 137? cm'', and 1056 cm''. Each of the main
l0 cholesterol bands has a secondary peak, which also
appear to be present in the necrotic core bands.
These secondary peaks occur at 1445/1436 cm'', and
1023 cm'' in the cholesterol spectrum and at 1441 cm'
', 1367 cm', and 1023 cm's in the necrotic core
spectrum. In addition, several of the weak bands in
the necrotic core spectrum, including the peaks at
1334 cm'', 1109 cm'', 954 cm'', and 79? cm'', are
associated with the weaker cholesterol bands near
these frequencies. Other components in the necrotic
core may also contribute to some of these distinct
bands.
The consistency of the spectral differences
which are attributed to cholesterol between the
necrotic core specimens and the normal,
atherosclerotic, arid non-ulcerated atheromatous
specimens are illustrated in the scatter plot in
Figure 13. This plot depicts the integrated
intensities (areas) of the 1050 cm'' cholesterol band
ratioed to the total protein content, as measured by .
the area of the amide II band at 1548 cm''. The 1050
cm'' band was integrated from 1075 to 1000 cm'' and _
baseline subtracted using these endpoints, and the
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-40-
amide II band was integrated from 1593 to 1485 cm'1
with a similar baseline subtraction. This ratio is
a measure as the relative cholesterol contribution
to the spectrum, and is proportional to the relative
cholesterol concentration of the sample with the
assumption that the area of the 1050 cm's band is due
solely to cholesterol. As can be seen in Figure 13,
this ratio is higher for all the exposed necrotic
core specimens than for all the other specimens.
l0 The consistent results of this sample analysis,
which is possible because of the separation and
molecular identification of the bands, indicates the
potential of IR spectroscopy for tissue
characterization.
Investigations of human arteries and
atherosclerosis by mid-IR spectroscopy have been
limited to date. It has been reported that ATR
spectra have been recorded from partially dried
human artery, among other tissues. In comparing a
normal aorta from an infant to an atherosclerotic
plaque in an adult, they observed increases in
several bands in the atherosclerotic aorta. Most of
these bands were associated with lipids and
lipoproteins. IR spectroscopy has been employed to
determine the chemical composition of calcified
atherosclerotic deposits. A more detailed IR study
of atherosclerotic aorta involves recorded IR
transmission spectra from thin layers sectioned at
different depths into the arterial wall. Results
showed increased absorption near 1739 cm's in the
fatty (atheromatous) regions of plaque, which was
' attributed to absorption by cholesterol esters in
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-41-
the plaque. IR spectra from the fibrous tissue cap
at the surface of the plaques were similar to normal
intima.
Ons of the main difficulties in measuring mid-
infrared spectra of intact human tissue is the
intense water absorption, which dominates and
obscures the absorption of other tissue components
of interest. In most of the studies cited above,
the water absorption was not eliminated, limiting
ZO the quality and amount of information available from
the spectra. With the ATR sampling method, this
water interference is easily removed (see Figure 9).
The ATR method is also naturally amenable to
sampling with fiber optic probes in yivQ. Water
i5 interference in fiber optic probe ATR spectra of
aqueous protein solutions has been accurately
eliminated with a water subtraction procedure
similar to the one employed in the present study.
While the ATR method is well suited to inin vivo
20 sampling and to accurate subtraction of the water
signal, spectra collected with the ATR method are
not equivalent to IR absorption spectra, but depend
on properties of the ATR material and the sample in
addition to the sample absorption coefficient. For
25 instance, the penetration depth of the evanescent
sampling wave depends on the refractive indices of
the ATR material and the sample. However, the
refractive indices of both ZnSe and human tissue are
expected to vary slowly with frequency between 1800
30 and 700 cm'' and such variations will at most affect
the relative intensities of bands at different
frequencies. All of the structure observed in the
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-42-
tissue spectra is attributed to absorption bands in
the tissue.
The component absorptions observed in an ATR
spectrum also depends upon the optical contact of
the sample and ATR element. The small penetration
depth of the evanescent wave into the tissue sample
implies that only a 5 ,um thick layer, and preferably
about 1 micron, of material at the surface is
observed. This is referred to as the near surface
region of the tissue for the purposes of this
application. The tissue deeper than 5 microns from
the surface is defined as the sub-surface region.
This thin, sampled near-surface layer may differ in
composition with the bulk sample. For example, a
film of free water may be present on the surface of
wet tissue, with different levels of some molecular
species of the tissue relative to their
concentrations in the bulk tissue. In addition, the
varied affinities for the ATR material of different
moieties in the tissue may play an important role in
the intensities of the observed hands.
These considerations may explain the lack of
substantial differences among the ATR spectra of
normal intima, plaque fibrous cap, and media. For
instance, normal aorta intima is composed of roughly
25% collagen (dry weight) and 20% elastin, while
aorta media has 20% collagen and 50% elastin. The
ATR spectra of purified collagen and purified
elastin (not shown) differ substantially. In
particular, amide I/II occur at 1657/1556 cm'1 in
hydrated collagen (type I) and 1553/1543 cm-' in
hydrated elastin (spectra not shown).
CA 02497575 1992-O1-17
wo 9insoos rcrivs92rooa2o
-43-
One might expect these differences to be
reflected in the intima and media ATR spectra. A
possible explanation of why this is not the case is
that the thin layer in optical contact wit the ATR
element is compositionally different from the bulk
tissue, and collagen and elastin make only a minor
contribution to the IR ATR bands of this layer.
Such an effect may also explain the lack of
significant differences among the plaque fibrous cap
intima and normal intima ATR spectra. In ATR
elements made of other substances with different
biochemical affinities, the spectral differences
among these tissues can be substantially enhanced
depending on the tissue type.
The results of the present investigation
demonstrate that high quality water-subtracted
spectra can be readily obtained from human tissue
with the ATR technique. Similar results have been
obtained in other mammalian tissues. Accurate
removal of the water interference is crucial to
isolating the relatively weak tissue absorption
bands of lipid, protein, as well as other tissue
components. It is worth noting that the observation
of these relatively weak bands via spectral
subtraction depends entirely upon quality of the
tissue and saline spectra. For instance, the
absorbance of the normal intima specimen (Figure 8a)
between 1500 and 900 cm'1 is approximately 0.06. In
the water-subtracted spectrum (Figure 9a), the peak
absorbencies range from 0.018 (30%) for the
strongest bands to 0.003 (5%) for the weakest ones.
The detection of a 0.003 absorbance peak in a
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-44-
subtracted spectrum with a 0.06 absorbance
' background requires a signal-to-noise ratio of 700
or better in the 100% baseline. Such a signal-to
' noise is easily achieved with an FT-spectrometer.
The high linearity, baseline stability, and
wavelength precision of the FT-spectrometer are also
obviously critical for accurate background
subtraction.
While water subtraction is relatively easy and
accurate with ATR, it may be substantially more
difficult with other clinically applicable sampling
techniques such as diffuse reflectance or
photoacoustic sampling. These alternative sampling
techniques are clinically useful, however, because
of their longer tissue penetration depths
(approximately 10~m). As an alternative to water
subtraction, one can exploit the properties of the
spectral lineshape of water to eliminate the water
signal by other computational methods.
Specifically, the spectral lineshape of water varies
rather slowly with frequency over much of the region
of interest, especially between 1500 and 700 cm'i.
Therefore, any method which filters this slower
variation and spares the sharper features of the
non-water bands can separate the water and non-water
components.
One such method is second derivative
spectroscopy. Differentiation of a spectrum is
typically used to narrow absorption bands and
resolve overlapping peaks. Differentiation also
tends to deemphasize broad bands relative to sharper
ones. In IR spectra of artery, the broad,
- CA 02497575 1992-O1-17
WO 9Z/I5008 PGT/US92/00420
-45-
featureless absorption of water can be nearly '
eliminated in favor of the sharper non-Water bands
by computing the second derivative of the spectra. '
This is clearly demonstrated in Figure 14, which
shows the second derivative of a spectrum of normal
aorta intima (Figure 14a), along with the water-
subtracted spectrum of the same specimen (Figure
14b). Essentially only the 1633 cmi water band is
left, partially obscuring the amide I band.
Elsewhere in this spectrum, the water contribution
is minimal. All of the bands identified in the
water-subtracted spectrum are easily observed in the
second derivative spectrum.
In addition to elimination of water
interference, several of the unresolved double peaks
and shoulders in the water-subtracted spectrum
appear as distinct peaks in the second derivative
spectrum. For example, the amide II band in normal
intima (Figure 14b) has a very weak shoulder near
1518 cm'', and the C-H bending region near 1468 cm'1
appears to include two overlapping peaks. In the
second derivative spectrum (Figure 14a), the 1518
cm~t band is clearly visible, and the C-H region
exhibits two separate peaks at 1469 and 1456 cm's.
Moreover, by sharpening the bands, the second
derivative spectrum allows a more precise
determination of peak frequencies, so that
relatively small frequency shifts are observed.
Such frequency shifts can be of importance in
detecting and characterising subtle molecular
alterations involved in certain tissue conditions.
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-46-
The observation of individual, resolved bands
in the artery IR ATR spectra is of considerable
interest, since separation of bands is the first
step determining the composition of a sample from
its spectrum. Once a band has been isolated, its
integrated intensity is proportional to the
concentration of the moiety responsible for that
band. In particular, since the amide I and II bands
are due entirely to protein, these bands can be used
to isolate the overall protein content in the
spectrum. The sharp, well resolved 1744 cm'' C=O
ester band appears to be due to solely to lipid, and
the integrated intensity of this band should be
proportional to the relative lipid content are
technique should largely eliminate the inaccuracies.
Finally, it should be remembered that the relative
water content of the tissue sample is automatically
computed from the 2120 cm'1 band in the water
subtraction algorithm. However, as noted earlier,
the composition of tissue as determined from an ATR
spectrum may not be precisely identical to the
composition of the bulk tissue.
The tissue composition can also be determined
from overlapping bands by first deconvolving the
bands of interest into their individual components.
This is especially easy if one component has an
additional, isolated band elsewhere in the spectrum.
An example is the 1465 cm~ C-H bending region, which
is due to different tissue components with distinct
spectral features in this region. In the normal
intima spectrum (Figure 9a), this band is attributed
to a combination of lipid and protein components.
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-47-
Since the lipid component also exhibits the isolated
1744 cm'i band, this band can be used to subtract the
lipid C-H bending component and isolate the protein
C-H bending component at 1455 cml (Figure lOb),
effectively deconvolving this band. Note that this
deconvolution depends on having a reliable spectrum
of one of the individual components, which, in this
example, is the lipid spectrum in Figure 9b.
The detection of distinct bands attributed to
cholesterol in necrotic core may provide a useful
means of determining cholesterol concentrations in
vivo. Both the 1050 cm'1 and 1382 cm'I cholesterol
bands appear to be fairly isolated in the necrotic
core spectrum after lipid-subtraction (Figure 12).
Tf these two bands are due to a single component,
namely cholesterol, the ratio of their integrated
intensities should be a constant for all the
samples. The baseline-subtracted area of the 1050
cm'I band, A(1050), is plotted versus that of the
1382 cm'I band, A(1382), for all the samples,
normalized to the protein content in Figure 15. As
can be seen in the plot, there is a roughly linear
relationship between A(1050) and A(1382). A linear
least squares fit to this data yields the line shown
in the Blot, with a high regression coefficient of
r=0.967. The slope of this line 2.8, while the
ratio A(1050)/A(1382j for the pure cholesterol ATR
spectrum is 2.3. The reasonable agreement between
these two numbers provides additional evidence for
the assignment of these bands to cholesterol.
Moreover, it indicates that the relative spectral
content of cholesterol is reasonably approximated by
CA 02497575 1992-O1-17
WO 92/15008 PC'T/US92/00420
-48-
the integrated intensities of either of these bands.
Figure 15 also shows that the ATR spectra of all the
specimens other than exposed necrotic core exhibit
almost no intensity in both the 1050 and 1382 cm'I
bands, in contrast to the necrotic specimens, all of
which have significant bands at both frequencies.
The present systems and methods demonstrate
that infrared spectra of moist, bulk tissues can be
reliably obtained with the ATR technique. Although
water is the dominant absorber throughout much of
the mid-infrared region, the high quality spectra
acquired with the FT-IR ATR technique allow for
accurate subtraction of the water signal.
Elimination of the water interference is critical
for identifying and assigning the absorption bands
of other tissue species. The isolation and
designation of these relatively sharp bands provides
a means of analyzing spectroscopically the
composition of arterial tissue non-destructively.
There methods are also applicable to the study and
diagnosis of other tissues and tissue conditions,
such as neoplasia.
The observation of both lipids and cholesterol
in the spectra of necrotic atheromatous core samples
is particularly exciting, because lipids and
cholesterol are thought to play major roles in the
pathogenesis of atherosclerosis. The spectral
observation of these components, cholesterol in
particular, provides a reliable means of detecting
and characterizing advanced atheromatous plaques in
which ulceration of the fibrous cap has occurred (as
demonstrated in Figures 13 and 15). Intimal
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-49-
accumulations of lipid and cholesterol occur early '
in the atherogenic process. Therefore, the mid-IR
ATR technique can also be useful in detecting and
studying the early fatty streak lesion.
_Spectroqraph_(CCD System for NI,~R,~.man Spectra
NIR Raman spectroscopy using a single stage
spectrograph and a charge coupled device (CCD)
detector offers superior sensitivity over the Nd:YAG
excited FT-Raman system of Figures lA and 1C. By
shifting the wavelength of the laser excitation from
1064 nm to the 800-900 nm region, a CCD can be used
to detect the Raman scattered signals while still
avoiding fluorescence excitation in most molecules.
The system can operate usefully in the range of 750
nm to 1050 nm. Although the fluorescence emission
from tissue is significantly higher with 810 nm than
with 1064 nm excitation, the Raman signals are
readily observed. This is because the dominant
noise source in the spectrograph/CCD system is shot
noise associated with the fluorescence emission,
which is 2-3 orders of magnitude smaller than the
dark current noise of the InGaAs detector, which is
the dominant noise source in the FT-Raman system.
Figure 17 shows the Iaser diagnosis and
treatment system of Figure lA modified to use the
spectrograph/CCD system of this invention. The
diagnostic subsystem 201' includes a single stage
spectrograph 310 and charge-coupled device (CCD)
detector 322 for collecting near-infrared (NIR)
Raman spectra from intact human arterial tissue.
With 810 nm laser light excitation, preferably
pulsed, the fluorescence emission from human artery
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-50-
tissue is sufficiently weak to observe Raman bands
more rapidly with the spectrograph/CCD system than
. with the 1064 nm excited FT-Raman system of Figures
lA and 1C. A method for removing the broadband
emission from the spectra by computing the
difference of two emission spectra collected at
slightly different excitation frequencies was used
to enhance observation of the Raman bands. This
method relies on the stability, linearity, and low
noise characteristics of the CCD detector. The
results indicate that high quality NIR Raman spectra
can be collected in under 1 second with the
spectrograph/CCD system and an optical fiber probe,
as compared to 30 minutes with the FT-Raman system
at similar laser power levels, further improving the
use of the technique for in vivo clinical
applications.
A preferred embodiment of a spectrograph/CCD
system 300 employed for the collection of near
infrared (NIR) Raman spectral data from excised
tissue samples using a spectrograph and a charge
coupled device (CCD) array is illustrated in Figure
18. NIR Raman spectra were measured from 100 - 2000
cm'1 below the laser excitation frequency with a
single stage imaging spectrograph 310 (Acton Model
ARC275, 0.25 m, F/3.8) and a CCD array 312
(Princeton Instruments EEV Model 88130).
System 300 can use a NIR 810 nm Nd:YAG pumped
pulsed dye laser 314 operating at l0 Hz for
irradiating a sample 46. Alternatively, a CW or
diode laser source can also be employed. Laser 314
generated a laser beam 316 which is directed by
CA 02497575 1992-O1-17
:.
WO 92/15008 PCT/US92/00420
-51-
mirror 318 through focusing optics 320 to impinge on
sample 46 mounted behind a transparent window 321.
The laser beam was focused on the sample at a 70°
angle of incidence, yielding a spot size of 0.7 x 2
mm on the tissue surface. The average incident
power at the sample was maintained at 20 mW to avoid
excessive peak intensities during an individual
pulse. The spectral signals were observed to be
linear over a range of average incident powers from
l0 2 to as mw.
A portion of the scattered light 322 emitted by
sample 46 was collected by collecting optics 324 at
a 90° angle relative to the incident laser beam.
Collecting optics 324 collimates and F/matches the
collected light far the spectrograph 310. Prior to
entering the entrance slit of the spectrograph 310,
the collected light was passed through a series of
Schott glass filters 326 which attenuated the
elastically scattered component of the collected
light. The combined effect of the Schott glass
filters provided an optical density of 7 at 8i0 nm,
a transmission of 20% at 850 nm (580 cm'' from 810
nm), and a transmission of 85% above 900 nm (1200
cm'I ) .
The spectrograph 310 utilized a 200 ~Sm slit
width and a 600 groove/mm grating blazed~at 1 ~m and
could be scanned to provide spectral coverage over
different wavelength regions. The 200 ~Cm slit width
provided a resolution of roughly 15 cm'1.
The CCD array 312 consisted of 298 (column) by
1152 (tow) pixel elements having a total active area
of 6.7 mm x 26 mm, with the short axis parallel to
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-52-
the slit. The CCD array was cooled to -110°C to
eliminate dark current. Each row of pixels was
binned to reduce readout noise. Commercially
available CCD detectors offer extremely low detector
noise and usable quantum efficiencies out to 1050 nm
and provide substantial advantages over InGaAs and
other NIR detectors. These advantages outweigh the
lower throughput of the grating spectrograph,
provided that broadband fluorescence interference is
not too great with the shorter excitation
wavelengths.
Excised human aorta samples 46 obtained at the
time of post-mortem examination were rinsed with
isotonic saline solution (buffered at pH 7.4), snap-
frozen in liquid nitrogen, and stored at -85°C.
Prior to spectroscopic examination, samples were
passively warmed to room temperature and were kept
moist with the saline solution. Normal and
atherosclerotic areas of tissue were identified by
gross inspection, separated, and sliced inta roughly
8 x 8 mm pieces.
The tissue samples 46 Were placed in a suprasil
quartz cuvette with a small amount of isotonic
saline to keep the specimens moist, and with one
surface in contact with the transparent window 321
and irradiated by the laser 314.
Raman spectra were typically measured between
loo cml and 2000 cm'' below the laser excitation
frequency. Each spectrum was background subtracted
to remove the DC offset of the A/D converter of the
CCD controller. In addition, hot pixels due to high
energy radiation events were removed from the
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-53-
recorded spectrum by applying a median filter having
a 7 pixel Wide window as to each spectrum. Roman
frequencies were calibrated with the spectra of
benzene and barium sulfate powder and are accurate
to ~5 cm-1. The spectra were not corrected for the
wavelength dependent response of the filters,
spectrograph, and CCD. For each spectrum shown in
the following Figures, Roman signals were
accumulated for 5 minutes. Substantially shorter
collection times can also be used as described
herein.
Figure 19A shows the Roman spectra of a normal
aorta sample excited with 810 nm laser light and
collected with the spectrograph/CCD system 300. In
this case, the broadband background emission, which
is presumably due to tissue fluorescence, is roughly
five times more intense than the strongest Roman
bands at 1650, 1451, 1330, and 1253 cm'1. In
contrast, the 1064 nm FT-Roman study of normal human
aorta shown in Figure 2a exhibited Roman signals
with the peak intensities of the strongest bands,
amide I at 1650 cm'1 and C-H bend at 1451 cm-~, being
roughly three times larger than the broadband
background emission. However, this background
emission in the spectrograph/CCD system is
relatively weak with respect to the Roman signals
(i.e., on the order of the Roman signals} and
therefore the shat noise associated with detecting
this background emission is substantially smaller .
than the Roman signals, allowing the Roman bands to
be made distinct after the background emission
signals are removed through filtering or
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-54-
subtraction. The shot noise is typically random
noise exhibiting a Poisson distribution and is
associated with the detector and/or the background
emission itself.
In contrast, with visible excitations, the
fluorescence background emission from the arterial
pathology tissue types described is 3 to 4 orders of
magnitude larger than the Raman signals, and the
shot noise associated with this stronger background
l0 emission completely obscures the Raman bands even
after the background emissions are removed.
However, certain other types of tissue, e.g., colon
and bladder, do not exhibit such high level
fluorescence reactions at visible excitation
frequencies, and therefore can probably operate with
visible excitations.
The signal-to-noise ratio of the spectrum of
normal aorta collected with the spectrograph/CCD
system 300 with 20 mW incident power and 5 minutes
collection time (Figure 19A) is similar to that
obtained with the FT-Raman system of Figure 1C with
500 mW incident power and 35 minute collection time.
Since the observed spectral signal-to-noise ratios
are similar, we estimate that the noise level
observed with the CCD detector 312 of Figure 18 is
roughly 3400 times less than that observed with the
InGaAs detector 42 of Figure 1C. For the InGaAs
detector, the major noise source is the shot noise
of the dark current, while with the CCD detector the
dominant noise source is the shot noise of the
_ broadband tissue emission, as the dark current and
readout electrons of the CCD are much smaller than
this emission.
CA 02497575 1992-O1-17
~r0 92/15008 PCT/US92/00420
-55-
This simple analysis has several important
implications. First, since the maior noise source
encountered with the spectrograph/CCD system is shot
noise from broadband emission by the tissue sample,
the spectral signal-to-noise ratio is proportional
to the square root of the product of incident
intensity and the collection time.
The FT-Raman and spectrograph/CCD systems can
be compared as follows. For the FT-Raman system,
the incident intensity is 640 mW/mm2. The quantum
efficiency of the InGaAs detector at 1200 nm is 0.7,
and the FT-spectrometer throughput is 1.1 mm=sr, and
the transmission efficiency of the FT-spectrometer
and filters is roughly 0.062. For the
spectrograph/CCD system, the incident intensity is
14 mW/mmZ. The CCD quantum efficiency is 0.15 at
900 nm, the spectrograph throughput is 0.043 mm2sr,
and the transmission efficiency of the spectrograph
and filters is 0.24. Combining these factors and
taking into account the v4 dependence of the Raman
cross-sections, the signal level measured by the FT-
Raman spectrum is estimated to be 3400 times greater
than that of the spectragraph/CCD spectrum.
Therefore, if the laser intensity is increased
to the level employed in the FT-Raman experiments,
the collection time could be reduced by a factor of
40, to 8 seconds, with no change in the spectral
signal-ta-noise ratio. Second, the noise level can
be further reduced by using longer excitation
wavelengths which minimize the tissue fluorescence
emission. However, such reductions in fluorescence
emission must be balanced against the decreasing
CA 02497575 1992-O1-17
WO 9Z/15008 PGT/US92/00420
-56-
quantum efficiency of the CCD at longer wavelengths,
and the optimum excitation wavelength also depends
on the fluorescence excitation profile of the
tissue. For tissue types that exhibit little
fluorescence emission at visible wavelengths, such
as colon and bladder tissue, the CCD can be operated
at visible or near visible wavelengths to take
advantage of increased quantum efficiency of the CCD
at these wavelengths. Finally, the throughput of a
500 ~cm core, 0.2 numerical aperture fused silica
optical fiber is 0.03 mmzsr, which is roughly the
same as that of the spectrograph/CCD system. This
means that the present lens collection system can be
replaced with an optical fiber probe, as is required
for in vivo operation, with no additional loss in
signal.
Figure 19A shows that although the shot noise
due to the broadband tissue emission is relatively
small, the sloping broadband fluorescence emission
still obscures the sharper Raman signals and
complicates determination of peak frequencies and
identification of weak bands. Furthermore, given
the complexity of human tissue, it is likely that
this broadband emission will be significant
throughout the useful range of the CCD. Any
quantitative analysis of the Raman bands in Figure
19A requires that this broadband emission be first
removed from the spectrum. The standard methods of
removing fluorescence emission from Raman spectra
utilize mathematical filters, which rely upon the
. fluorescence emission being relatively featureless.
In an alternative method the excitation frequency is
,~..~,. CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-57-
varied over a narrow range (10 - 30 cm'1). The Roman '
band positions vary directly with the excitation
frequency, while the fluorescence emission remains
fairly constant With such small changes in
excitation frequency, allowing it to be efficiently
subtracted cut. In contrast with mathematical
filters, this operation requires no assumptions
about the emission lineshape.
To implement this method, the Roman spectrum of
id the normal aorta specimen is recorded with
excitation wavelengths of 810 nm (Figure 19A) and
812 nm. The Roman bands shift with the excitation
frequency by 30 cm'i, while the fluorescence emission
remains fairly constant. By subtracting these two
spectra, the broadband emission is greatly reduced,
and the Roman bands are more readily observed
(Figure 19B). This operation is mathematically
analogous to taking the derivative of the Roman
spectrum, so that the original Roman spectrum can be
2o recovered by integrating the difference spectrum, as
shown in Figure 19C. The fluorescence background is
greatly reduced in Figure 19C as compared with
Figure 19A, allowing easier identification of the
Roman bands and their peak frequencies. The
integration also smooths the Roman spectrum over a
bandwidth similar to the excitation frequency shift
and causes some linewidth broadening, as is evident
from Figure 19C. Note that the accuracy of this
method depends upon the high linearity and stability
of the CCD array.
The NIR Roman spectrum of an atherosclerotic
plaque with a calcified deposit exposed at the
CA 02497575 1992-O1-17
WO 92/15008 PCT/US92/00420
-58-
surface collected with the spectrograph/CCD system
is shown in Figure 20A. In this case, the broadband
emission is nearly io times greater than that
observed in~normal aorta (Figure 19A), resulting in
increased noise. However, the intense phosphate
stretching vibration at 960 cm'i, due to the
calcified salts, is readily identified. This band
is sufficiently intense to be observed in real time
and Was used in aligning the collection optics.
to Some weaker bands may also be identified, such as
the phosphate/carbonate band at 1070 cm's, although
these are obscured by the large fluorescence
background. By subtracting out this fluorescence
(Figure 20B), as above, these bands are much more
easily distinguished. The Raman spectrum obtained
by integrating the difference spectrum is shown in
Figure 20C. The broadband emission is reduced by a
factor of 50 relative to Raman bands, and several
weaker bands are readily observed. This spectrum is
remarkably similar to that of Figure 5a which was
observed with the FT-Raman system and 1064 nm
excitation.
As another example of the sensitivity of the
spectrograph/CCD system 300, the Raman spectrum of
adventitial adipose tissue is shown in Figure 21,
which can be compared to the FT-Raman spectrum shown
in Figure 5c. The broadband emission is similar to
that of normal aorta, While-the Raman bands, due
mainly to triglycerides in the tissue, are very
3o strong, resulting in an excellent spectral signal-
to-noise ratio.
CA 02497575 1992-O1-17
WO 91/I5008 PCr/US92/00420
-59-
Thus, the spectrograph/CCD system with 81o nm
excitation offers a faster alternative to FT-Raman
with 1064 nm excitation and which has greater
sensitivity. Even fn complex mixtures such as human
5 tissue, the level of background emission observed
with 810 nm excitation is low enough to observe the
Raman signals. This fluorescence emission daes not
excessively degrade the signal-to-noise ratio. By
subtracting two spectra collected at slightly
iQ different excitation wavelengths, and then
integrating the difference spectrum, this broadband
emission is rejected, yielding high quality Raman
spectra. Deconvolution techniques can also be used
to selectively remove, or reduce, Raman,
15 fluorescence, or noise light components.
Improvements such as using a CW laser to increase
the incident intensity and a back-thinned CCD having
better red response allows Raman spectra to be
collected from intact human tissue in under 1
20 second. Longer excitation wavelengths may reduce
the background emission further. Implementation of
the spectrograph/CCD system With a high power diode
laser and an optical fiber probe will grovide a
compact, mobile system for rapidly acquiring NIR
25 Raman spectra remotely from human tissues and will
provide a powerful tool for in v~vo clinical
applications.
CA 02497575 1992-O1-17
WO 92/15008 PGT/US92/00420
-60-
~,g~.valents
Those skilled in the art will recognize, or be
able to ascertain using routine experimentation,
many equivalents to the specific embodiments of the
invention described herein. These and all other
equivalents are intended to be encompassed by the
following claims.
