Professional Documents
Culture Documents
The use of near-infrared Raman spectroscopy to interrogate epithelial tissue biochemistry and hence
distinguish between normal and abnormal tissues was investigated. Six different epithelial tissues from
the larynx, tonsil, oesophagus, stomach, bladder and prostate were measured. Spectral diagnostic models
were constructed using multivariate statistical analysis of the spectra to classify samples of epithelial
cancers and pre-cancers. Tissues were selected for clinical significance and to include those which develop
into carcinoma from squamous, transitional or columnar epithelial cells. Rigorous histopathological
protocols were followed and mixed pathology tissue samples were discarded from the study. Principal
component fed linear discriminant models demonstrated excellent group separation, when tested by cross-
validation. Larynx samples, with squamous epithelial tissue, were separated into three distinct groups
with sensitivities ranging from 86 to 90% and specificities from 87 to 95%. Bladder specimens, containing
transitional epithelial tissue, were separated into five distinct groups with sensitivities of between 78
and 98% and specificities between 96 and 99%. Oesophagus tissue can contain both squamous and
columnar cell carcinomas. A three group model discriminated the columnar cell pathological groups with
sensitivities of 84–97% and specificities of 93–99%, and an eight group model combining both columnar
and squamous tissues in the oesophagus was able to discriminate pathologies with sensitivities of 73–100%
and specificities of 92–100%. It is likely that any overlap between pathology group predictions will have
been due to a combination of the difficulty in histologically distinguishing between pre-cancerous states
and the fact that there is no biochemical boundary from one pathological group to the next, i.e. there is
believed to be a continuum of progression from the normal to the diseased state. Copyright 2002 John
Wiley & Sons, Ltd.
present. For example, long periods of exposure to gastro- value. The potential of Raman spectroscopy to distinguish
oesophageal reflux can lead to a change in the lining of the between normal and malignant tissue has been demonstrated
oesophagus, whereby the normal squamous epithelial lining by a number of workers.16 – 20
is replaced by protective columnar epithelial cells. This is The objective of this study was to evaluate the potential
often termed Barrett’s oesophagus and it is associated with an for near-infrared (NIR) Raman spectroscopy, a highly specific
increased risk of developing oesophageal adenocarcinoma.5,6 optical analysis technique, to classify or discriminate between
Barrett’s dysplasia is the most frequently used marker of the pathological state of different epithelial tissues and thus
increased cancer risk in Barrett’s oesophagus. It is more demonstrate a future capability for in vitro classification
commonly seen in areas of intestinal metaplasia (IM), one and in vivo detection of early epithelial malignancies. The
of three types of Barrett’s oesophagus, and it is unlikely larynx, oesophagus and bladder are the main subjects of
that oesophageal adenocarcinoma occurs except in patients this work. Carcinomas in each organ develop from three
with IM.7,8 The potential benefits of removing an oesophagus distinctly different types of epithelial cells. This has allowed
exhibiting dysplasia must be weighed against the relatively the evaluation of Raman spectral analysis for detection
high mortality associated with oesophagectomy (5–15%)9 and classification of disease in squamous, transitional and
and the poor outcome in patients who present with invasive columnar epithelial cells. To date only the oesophagus has
adenocarcinoma of the oesophagus (23% survival at 1 year been examined in any detail with Raman spectroscopy,
and 7% survival at 5 years in England and Wales between although only selected pathologies were compared.17,21
1986 and 1990).1 Although survival rates for laryngeal and This study followed extremely rigorous sampling pro-
bladder cancers are better (55–65% at 5 years),1 it is delayed cedures to reduce the errors caused by discrepancies in
diagnosis leading to systemic metastatic invasion that is the histopathology, sample misorientation and by measuring
main reason for poor survival from epithelial cancers.10 samples with mixed pathology. In addition, the full spec-
The ‘gold standard’ for the detection of malignan- trum of malignant disease found in each organ was studied,
cies and pre-malignancies is excisional biopsy followed by in order to remove any bias likely to be found in studies
histopathological analysis. This technique relies upon sec- comparing normal with cancerous tissue, i.e. the extreme
tioning tissue less than one cell thick (<10 µm), staining with pathologies, but not the stages in between.
haematoxylin and eosin (H&E) and viewing under a con-
ventional light microscope. The analysis depends mainly on
the subjective recognition of tissue morphology and architec- EXPERIMENTAL
tural patterns. There are significant difficulties in obtaining
an accurate diagnosis using the ‘gold standard.’ First, the Tissue samples
selection of biopsies for the detection of invisible microscopic Tissue specimens were collected for this study during routine
surface lesions must rely upon a blind targeting protocol. This endoscopic or surgical procedures. Informed written consent
can lead to a high probability of missing abnormal tissue and was obtained from each patient and the study was approved
large numbers of normal samples will be generated. A study by the relevant Local Research Ethics Committees (UK). The
in the oesophagus following 16 patients undergoing surveil- tissue samples were removed from the biopsy forceps and
lance for Barrett’s oesophagus demonstrated that even a orientated on acetate paper to enable a depth cross-section
rigorous sampling protocol will be likely to miss abnormal through the tissue to be seen. The mounted sample was then
lesions previously found in the organ.11 Second, in the anal- placed in a 2 ml cryovial (BDH) and dropped into liquid
ysis of pre-cancerous lesions there are often high levels of nitrogen. Histological tissue cross-sections were cut from
discrepancy between pathologists, owing to the subjective each of the samples using a freezing-microtome and stained
nature of histopathological analysis. Studies grading dys- with H&E. The use of the acetate backing paper made it
plasia in ulcerative colitis and colonic adenomas produced possible to adhere the sample to the microtome without
an overall agreement between any two pathologists ranging the use of a cutting agent and thus minimize chemical
from 42 to 65%.12,13 Similar studies demonstrated agreement contamination to the tissue sample. Stained sections were
of between 58 and 61% for oesophageal dysplasia14 and 54% analysed by between one and three registry histopathologists
for laryngeal dysplasia.15 Furthermore, the removal of tis- according to organ-specific architectural and morphological
sue samples can have a damaging effect on the function of grading systems.22 Only homogeneous samples with clearly
the organ, especially in sensitive organs such as the larynx, defined pathologies were used in this investigation and those
where permanent loss of voice may result. with mixed or indeterminate pathologies were discarded.
Histopathological examination of biopsy samples relies In addition only oesophagus samples with a unanimous
upon the subjective assessment of tissue architecture, which agreement from all three pathologists were used for this
is likely to exhibit abnormal changes at a later stage than sub- study. Both of these precautions were implemented to reduce
cellular biochemical changes. Evidently, the development errors from using misclassified and mixed pathology tissue
of a technique, permitting the objective, non-invasive samples. The retained samples (biopsy blocks) were stored
qualitative biochemical analysis of tissue would be of great at 85 ° C until spectroscopic studies were carried out. Prior
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
566 N. Stone et al.
to performing spectral measurements, the specimens were the higher line density grating, to scan the spectrum across
passively warmed to room temperature. the CCD using the ‘extended scan mode.’ The lower density
A total of 41 laryngeal biopsy specimens were collected grating allowed measurement of the full fingerprint region
from 35 patients undergoing routine microlaryngoscopy. of the tissue spectrum without the need to move the grating.
The larynx sections were classified by a histopathologist as However, a compromise in spectral resolution was made,
either normal squamous epithelium; squamous dysplasia but this has been shown not to affect the quality of the data.
or squamous cell carcinoma. Twenty-five samples from 20 High-quality NIRPLAN optics were selected to minimize
patients were retained for analysis with Raman spectroscopy, their spectral contribution in the fingerprint region, reducing
including 14 histologically normal samples, five exhibiting the requirement for optics signal subtraction.
dysplasia and six with squamous cell carcinoma. Samples were orientated so that the incident laser beam
Over 150 oesophageal jumbo biopsy samples were har- illuminated the epithelial surface, with a similar geometry to
vested from patients undergoing endoscopic surveillance that which would be used in vivo. Tissue samples were held
for dysplasia in Barrett’s oesophagus. Following section- in position on the microscope stage between two calcium
ing and histopathological analysis, only 89 homogeneous fluoride slides. A ð 80 ultra-long working distance lens
samples from 44 patients remained. A blind consensus of was used to focus the laser beam (power at the sample
histopathological opinion was achieved on 50 of these. The of 32 š 1.1 mW) to a spot size of 2–3 µm on the tissue surface
samples with consensus, retained for study with Raman and collect the scattered photons in non-confocal mode. At
spectroscopy, exhibited eight different pathological states least five spectra were acquired from each larynx sample
[normal squamous epithelium, cardiac Barrett’s (CB), fundic (mean number D 7.96, range 5–12) and a minimum of 10
Barrett’s (FB), intestinal metaplasia (IM), high-grade dys- spectra were acquired from each oesophagus sample (mean
plasia (HGD), adenocarcinoma, squamous dysplasia and number D 14, range 10–20). The scattered Raman signal
squamous cell carcinoma (SCC)]. Of all the samples in this was integrated for 30 s and measured over a spectral range
study there was no consensus of agreement on samples of 400–1800 cm1 with respect to the excitation frequency.
exhibiting low-grade dysplasia (LGD) and therefore this Bladder spectral measurements were performed in the same
group was omitted from the consensus prediction model manner, except that a ð20 objective was used with an
described in this paper. illumination power of ¾85 mW. At least 15 spectra were
Bladder samples exhibiting normal, carcinoma in situ acquired from each sample and spectra were integrated for
(CIS), low-grade carcinoma (LGC), moderate-grade carci- 10 s. The changes in spectrometer configuration between
noma (MGC) and high-grade carcinoma (HGC) were har- laryngeal to oesophageal to bladder spectral datasets were
vested during routine cystoscopy. Twelve homogeneous caused by ongoing optimization processes with the aim of
samples from 12 patients were retained for analysis. maximizing the signal and minimizing integration times.
The Raman spectrometer was calibrated for wavenumber
Other samples shift using a neon lamp standard and Renishaw WiRE
Homogeneous normal samples of stomach, prostate and software. Tissue spectra were corrected for the energy-
tonsil were also studied to demonstrate the facility of dependent response of the system by multiplying the
NIR-Raman spectroscopy to provide repeatable analysis of measured spectrum by the energy transfer function, which
tissue biochemistry in other epithelial tissues. Additionally, was measured using a tungsten-filament lamp with a
confounding factors or substances likely to contaminate spectral output calibrated by the UK National Physical
epithelial tissue spectra were collected and studied. These Laboratory. No further manipulation of the tissue spectra
included mucus, bile, blood, a coffee and spittle mixture and was performed prior to analysis. Intra-sample repeatability
endoscope lubricant. was evaluated and the mean spectra for each tissue pathology
were obtained. Any correlation between Raman spectral
Raman spectroscopy variations and histopathology was assessed using empirical
Raman scattering measurements were performed in vitro and multivariate data analysis techniques.
with a customized Renishaw System 1000 Raman micro-
spectrometer. A Spectra-Physics argon ion (2017) pumped
Ti:sapphire laser (3900S) was tuned to provide an excita- RESULTS AND DISCUSSION
tion wavelength of 830 nm and a custom set of edge filters
(Renishaw) were used to reject elastic scattered light of the The use of 830 nm excitation radiation facilitated the
laser wavelength. A single dispersion grating was used to acquisition of high-quality, low-fluorescence spectra in short
separate the collected light into its spectral components: time-scales. Little background fluorescence is present in the
1200 lines mm1 for laryngeal tissue measurements and spectra and the background slope is almost exclusively
300 lines mm1 for oesophagus and bladder tissue mea- caused by the diminishing CCD sensitivity at longer
surements. To allow analysis of the full fingerprint region wavelength. Signal-to-noise ratios between 10 and 20 were
(400–1800 cm1 ) of the tissue spectra it was necessary, with achieved (depending on the signal strength of the mode)
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
Classification of epithelial pre-cancers and cancers
0.65 5
Normal Squamous Mucosa
Squamous Dysplasia
4
0.6
Squamous Cell Carcinoma
3
0.55
Intensity / arbitr. units
2
0.5
1
LD 2
0.45
0
0.4
-1
0.35
-2
0.3 -3
Normal Larynx
Squamous Dysplasia
Squamous Cell Carcinoma
0.25 -4
400 600 800 1000 1200 1400 1600 -6 -4 -2 0 2
-1
Wavenumber / cm LD 1
Plate 1. Mean Raman spectra from each pathological group Plate 2. Plot of linear discriminant function weights for each
in the laryngeal tissue. Spectra have been normalized to the larynx spectrum, when tested against the optimized model using
intensity of the 1446 cm1 peak. a cross-validation process.
0.6
0.45
Intensity / arbitr. units
0.4
0.35 0
LD 2
0.3
-2
0.25 Normal Oesophagus
Cardiac Barrett's
Fundic Barrett's
0.2 Intestinal Metaplasia
Low Grade Dysplasia -4
High Grade Dysplasia
0.15 Adenocarcinoma
Squamous Dysplasia
Squamous Cell Carcinoma
0.1 -6
400 600 800 1000 1200 1400 1600 1800
-5 -3 -1 1 3 5 7
-1
Wavenumber / cm LD 1
Plate 3. Mean Raman spectra from all pathologies present in Plate 4. Plot of linear discriminant function weights of the
the oesophageal tissue specimens studied. Spectra have been three-group oesophagus model for each spectrum, when tested
normalized to the intensity of the 1446 cm1 peak. against the model using a cross-validation process.
6
Normal Bladder
0.9
Carcinoma in situ
0.7
LD 2
0.6
-2
Plate 5. Mean Raman spectra from all bladder specimens Plate 6. Plot of linear discriminant function weights for
studied, grouped by pathology. Spectra have been normalized each spectrum, when tested against the model using a
to the intensity of the 1446 cm1 peak. cross-validation process.
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33
N. Stone et al.
LD 3
LD 2
0
0
-2
-2
-4 -4
-6 -6
-8 -8
-6 -1 4 9 -5 0 5 10 15
LD 1 LD 1
4 5
2 3
0 1
LD 5
LD 4
-2 -1
-4 -3
-6 -5
-8 -7
-5 0 5 10 15 -6 -1 4 9 14 19
LD 1 LD 1
6 6
4 4
2 2
LD 6
LD 7
0 0
-2 -2
-4 -4
-6 -6
-5 0 5 10 15 -5 0 5 10 15
LD 1 LD 1
Plate 7. Plot of linear discriminant function weights of the eight-group oesophagus model for each spectrum, when tested against
the model using a cross-validation process.
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33
Classification of epithelial pre-cancers and cancers 567
in epithelial spectra when using the ð80 objective, the 300 mucosa, squamous dysplasia and squamous cell carci-
lines mm1 grating and 30 s of signal integration. noma, were calculated, normalized to the peak intensity
Plots of mean, energy sensitivity corrected, Raman at 1446 cm1 and plotted in Plate 1. Visual inspection of this
spectra from the normal epithelial tissues studied (larynx, plot shows that the differences observed in the spectra from
oesophagus, bladder, tonsil, stomach and prostate) are specimens at different stages of cancer development are very
shown in Fig. 1 to enable similarities and differences to subtle. At first glance the spectra appear almost identical.
be visualized. The Raman spectra of the normal epithelial However, a closer study of the data reveals miniscule changes
tissues are very similar, but subtle variations exist. The most in peak heights from one group to the next. Difference spectra
significant of these can be seen in the wavenumber ranges indicate that the magnitude of the spectral variance between
800–980 and 1200–1400 cm1 . pathology groups is between 1 and 2% of the Raman spec-
The spectral repeatability over 100 spectra from 10 tral signal. Therefore, rather than empirically select peaks
normal oesophageal specimens from different patients was or peak ratios for significance of correlation between inten-
evaluated. The spectral shape was consistent and the sity and pathology group, all major peak intensities were
wavenumber positions of the peaks vary about the mean measured for each spectrum in the data set and their ratios
positions by š1.12 cm1 . However, there was a significant calculated. These peak intensities and peak intensity ratios
baseline intensity variation from one sample to the next. The were then tested for significance of group separation using
mean spectrum from 1200 oesophagus spectra, covering all analysis of variance (ANOVA). The inter-sample variation
pathologies exhibited in the measured specimens, is shown of individual peak intensities was large enough to obscure
in Fig. 2. The most significant vibrational modes have had the inter-group differences. However, the use of peak ratios
their mean wavenumber positions labelled. An attempt to compensated for the overall baseline intensity changes from
assign these peaks tentatively has been made in Table 1. one sample to the next. The most significant peak ratio (F
value 33.46, p value −0.0001) for separation of the three
groups was shown to be I754 /I780 . Mean intensity ratios of
Larynx 0.953 š 0.036, 0.942 š 0.035 and 0.906 š 0.041 were observed
Laryngeal spectra were combined into three separate groups, for normal, dysplastic and cancerous tissue, respectively.
each defined by the tissue pathology of the measured spec- Although on average this peak ratio does vary significantly
imen. The mean spectra for each group, normal squamous with pathology group, the variation between spectra were
0.7 0.6
0.55 0.55
0.6
0.5 0.5
0.5
0.45
0.45
0.4 0.4
0.4 0.3 0.35
0.35 0.2 0.3
0.25
600 800 100012001400 1600 600 800 100012001400 1600 600 800 100012001400 1600
0.9 1.1
0.85 0.2
1
0.8
0.75 0.9
0.7 0.15
0.65 0.8
0.6
0.7
0.55 0.1
600 800 100012001400 1600 600 800 100012001400 1600 600 800 100012001400 1600
Wavenumber / cm-1
Figure 1. Mean energy sensitivity corrected Raman spectra from selected normal epithelial tissues.
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
568 N. Stone et al.
Mean Raman spectrum from oesophagus tissue (all pathologies) at 830 (n = 1200)
0.8
1446
0.7
1001
1313
1335
1259
935
0.6
1083
1123
1655
853
1031
Intensity / arbitr. units
1209
828
0.5
524
1155
755
490
719
1170
781
643
621
669
1616
0.4
1548
1579
0.3
0.2
0.1
400 600 800 1000 1200 1400 1600 1800
Wavenumber / cm-1
Figure 2. Mean Raman spectrum from all samples used in the oesophagus study. Wavenumber positions of major peaks are
labelled.
large enough to make this method of discrimination poor for significance of variance between groups were entered into
spectral correlation with pathology. the LDA model. To separate three groups of data maximally,
It is likely that the extremely complex changes from two LD functions are required; for greater numbers of groups
normal to cancerous tissue will be accompanied by many (n), n 1 linear discriminant functions are required. All the
molecular variations. Therefore, spectral analysis techniques linear discriminant ‘diagnostic models’ were tested using
involving all the data in the spectra will be more likely cross-validation or leave-one-out procedures, whereby each
to describe the subtle but complex spectral variations. spectrum is held back in turn, the model is calculated and
Principal components (PCs) were used to describe the spectra the withheld spectrum is projected on to the model. This
as a linear combination of weights and loading vectors. technique does not provide as good a test as projecting
Each loading vector or principal component described ever completely new data on to the model, but with relatively
decreasing variances in the spectra from the mean of the small numbers of patients and samples it will provide an
dataset. PCA was performed to reduce the number of unbiased estimation of the model performance.
variables in the analysis. The weights of the principal Plate 2 shows a two-dimensional scatter plot of linear dis-
components were used as the variables for entry into a criminant function weights for the optimized larynx model.
linear discriminant analysis (LDA) model, which maximized The resulting prediction performances for each group against
the variance between groups and minimized the variance the ‘gold standard’ histopathology are shown in Table 2. Out-
within groups. The multivariate analysis routines were come measures were calculated to allow evaluation of the
performed using the PLS-toolbox (Eigenvector) on the model in terms of sensitivity and specificity of in effect three
MATLAB platform. Spectral datasets were pre-processed tests for the larynx model: either normal versus other groups,
using Savitsky–Golay filters and mean-centering, following dysplasia versus other groups or cancer versus other groups.
correction for system energy sensitivity (see Experimental). These are also displayed in Table 2. From scatter plots of
Prior to calculation of the linear discriminant functions to the linear discriminant function weights it could be seen
separate the groups maximally, ANOVA was performed on that group clustering was achieved successfully, although
the PC weights to identify the diagnostically significant some groups appeared to cluster into two sub-groups. This
components (PCs) of the spectra. Six PCs with >99.9% is likely to be due to the evolving spectrometer configuration
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
Classification of epithelial pre-cancers and cancers 569
Peak
position/cm1 Major assignments
490 Glycogen
524 S–S disulfide stretch in proteins
621 C–C twisting mode of phenylalanine
643 C–C twisting mode of tyrosine
669 C–S stretching mode of cystine
719 C–N (membrane phospholipid head)/nucleotide peak
755 Symmetric breathing of tryptophan
781 Cytosine/uracil ring breathing (nucleotide)
828 Out of plane ring breathing tyrosine/O–P–O stretch DNA
853 Ring breathing mode of tyrosine and C–C stretch of proline ring
920 C–C stretch of proline ring/glucose/lactic acid
935 C–C stretching mode of proline and valine and protein backbone (˛-helix
conformation)/glycogen
1001 Symmetric ring breathing mode of phenylalanine
1031 C–H in-plane bending mode of phenylalanine
1083 C–N stretching mode of proteins (and lipid mode to lesser degree)
1123 C–C stretching mode of lipids/protein C–N stretch
1155 C–C (and C–N) stretching of proteins (also carotenoids)
1170 C–H in-plane bending mode of tyrosine
1209 Tryptophan and phenylalanine (C–C6 H5 ) mode
1240–1265 Amide III (C–N stretching mode of proteins, indicating mainly ˛-helix conformation)
1313 CH3 CH2 twisting mode of collagen/lipids
1335 CH3 CH2 wagging mode of collagen and polynucleotide chain (DNA–purine bases)
1446 CH2 bending mode of proteins and lipids
1548 Tryptophan
1579 Pyrimidine ring (nucleic acids) and heme protein
1603 C C in-plane bending mode of phenylalanine and tyrosine
1616 C C stretching mode of tyrosine and tryptophan
1655 Amide I (C O stretching mode of proteins, ˛-helix conformation)/C C lipid stretch
Sensitivity/% 89 90 86
Specificity/% 87 93 95
a
Each case was classified by the functions derived from all cases other than that
case; 88.4% of cross-validated grouped cases are correctly classified using this
method.
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
570 N. Stone et al.
that was employed during measurements of the laryngeal with 93.2 and 88.6% of cases correctly predicted by the three-
spectral dataset. Raman spectra were acquired with differ- and eight-group models, respectively.
ent resolutions, signal-to-noise ratios, calibration protocols The greatest cross-over between groups occurred
and laser stabilities. The use of a smoothing function to pre- between the intestinal metaplasia, high-grade dysplasia and
process the data was shown to reduce the effect of these adenocarcinoma groups. It is these that cause pathologists
variations in the model. the greatest difficulty in classification. The cross-over is likely
to be due in part to both the difficulty in pathological
Oesophagus discrimination between these groups and the fact that
Oesophageal spectra were measured from homogeneous abnormal development of the epithelial tissue follows a
samples of eight different pathological groups. The mean continuum with many biochemical similarities.
normalized spectra from each pathological group are shown Utilization of tentative peak assignments for significant
in Plate 3 for visual comparison.
positive and negative peaks in the difference spectra (not
The spectra were combined into clinically significant
shown) enabled some sense to be made of the biochemical
groups for construction of a multivariate linear discrim-
changes that accompany the progress towards adenocar-
inant model in the same way as that described for the
cinoma in oesophageal tissue. Dysplastic and cancerous
larynx. However, following significant enhancement in laser
tissues have been shown (tentatively) to exhibit greater DNA,
and spectrometer wavenumber repeatability afforded by
hydroxyapatite, urea/phenylalanine, ˛-helix and unordered
improvements in temperature stabilization, it was found to
proteins, but lower glycogen, carbohydrates, protein disul-
be unnecessary to smooth/pre-process the oesophageal and
fide bonding and carotenoids, than normal tissue. This
bladder data prior to calculation of principle components.
tends to corroborate known biochemical changes leading
Both eight-group and three-group linear discriminant mod-
els were constructed and tested. The three-group model, to tumour develoment such as increased energy consump-
relevant for biopsy targeting applications, used spectra tion from cell division, increased nuclear/cytoplasmic ratio,
measured from normal, Barrett’s oesophagus and Barrett’s protein conformation changes and tissue calcification. The
neoplasia. This required two linear discriminant functions histopathological groupings have been developed by pathol-
for optimum separation of the groups. The cross-validated ogists utilizing architectural and morphological differences
LD function weights are plotted in Plate 4. However, the to recognize from collective experience when a tissue has
construction of an eight-group discriminant model for clas- undergone a change likely to be clinically significant. How-
sification necessitated the use of seven LD functions and a ever, cells are undergoing a continuum of change from
seven-dimensional decision space. In this case it is difficult to Barrett’s mucosa to dysplasia to adenocarcinoma, and there-
visualize group separations using 2D scatter plots. However, fore full separation, either histologically or spectroscopically,
an attempt was made to include an informative selection will be difficult.
to allow visualization of some grouping within two dimen- Ten principal components (1, 2, 3, 4, 5, 6, 7, 8, 10 and
sions (Plate 7). The resulting prediction performances and 12) were shown to exhibit >99.9% significance of discrim-
outcome measures for the three-group model are presented ination between the eight pathological groups designated
in Table 3 and for the eight-group model in Table 4. Excellent from a consensus of opinion of three pathologists. The most
separation of samples into pathological groups was achieved significant PC (6) described only 0.05% of the total variance
Sensitivity/% 97 84 94
Specificity/% 99 98 93
a
Each case was classified by the functions derived from all cases other than that
case; 93.2% of cross-validated grouped cases are correctly classified using this
method.
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
Classification of epithelial pre-cancers and cancers 571
Table 4. Cross-validation classification results obtained from the eight-group linear discriminant model calculated from
clinically significant principle components of the oesophagus spectraa
Normal 14 0 0 1 0 15
CIS 1 33 4 2 0 40
LGC 0 2 53 0 0 55
MGC 1 3 0 36 6 46
HGC 0 0 1 0 39 40
Sensitivity/% 93 83 96 78 98
Specificity/% 99 97 96 98 96
a
Each case was classified by the functions derived from all cases other than that case;
89.3% of cross-validated grouped cases are correctly classified using this method.
from the mean, indicating that subtle modifications in bio- spectra for each group are shown in Plate 5. Subtle dif-
chemistry precede and accompany significant pathological ferences can be seen. These have been exploited with
changes to the tissue. the use of a principal component fed linear discriminant
Although most principal components are made up of a model (as described above) to separate maximally the
complex combination of tissue components, PC3 can be seen five pathology groups exhibited by the specimens mea-
to represent glycogen, which is known to be the main energy sured. Four LD functions were calculated and the weights
source found in human cells. Normal cells will have this in of the first two are plotted in Plate 6 to demonstrate
abundance whereas in abnormal cells it will be in deficit. group clustering. The cross-validation prediction perfor-
This appears to be demonstrated in both difference spectra mance and the outcome measures are outlined in Table 5.
and principal component analyses. Excellent group separation is achieved, although there is
some overlap between carcinoma in situ and the low-
Bladder grade cancers and between the moderate- and high-grade
Raman spectra of bladder specimens were measured and cancers. This may be due in part to the biochemical contin-
grouped according to histopathology opinion. The mean uum between the increasing grades of malignancy and the
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
572 N. Stone et al.
difficulty in distinguishing fully between these groupings and coffee combined, blood and endoscope lubricant, all
with histopathology. measured under the same conditions as the tissue spectra, are
shown for comparison of signal strength with an oesophagus
Potential confounding factors spectrum. The tissue spectra measured to date (without
Figure 3 displays Raman spectra of potential contaminants washing) were not unduly affected by these contaminants,
to tissue spectra, that may be confounding factors in spectral although the strengths of the signals indicate that they should
diagnostic model prediction. Plots of bile, mucus, spittle not be ignored from analysis altogether.
9 0.08
8 0.06
7 0.04
6 0.02
5 0
400 600 800 1000 1200 1400 1600 1800 400 600 800 1000 12001400 1600 1800
Mucus Spittle
0.11 0.5
Intensity / arbitr. units
0.1
0.45
0.09
0.4
0.08
0.35
0.07
0.06 0.3
400 600 800 1000 1200 1400 1600 1800 400 600 800 1000 12001400 1600 1800
Wavenumber / cm -1 Wavenumber / cm -1
0.45 0.4
0.35
0.4
0.3
0.35
0.25
0.3
0.2
0.25 0.15
0.2 0.1
400 600 800 1000 12001400 1600 1800 400 600 800 1000 12001400 1600 1800
0.8
0.16
0.7
0.14 0.6
0.5
0.12 0.4
0.3
0.1
0.2
0.08 0.1
400 600 800 1000 12001400 1600 1800 400 600 800 1000 12001400 1600 1800
Wavenumber / cm -1 Wavenumber / cm -1
Figure 3. Raman spectra of possible contaminants of in vivo and in vitro tissue Raman spectra, measured in 30 s and corrected for
system energy sensitivity. An oesophagus spectrum is included for comparison.
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573
Classification of epithelial pre-cancers and cancers 573
Copyright 2002 John Wiley & Sons, Ltd. J. Raman Spectrosc. 2002; 33: 564–573