CA2190374C - Optical method and apparatus for the diagnosis of cervical precancers using raman and fluorescence spectroscopies - Google Patents

Abstract

A method and apparatus for detecting tissue abnormality, particularly precancerous cervical tissue, through fluorescence or Raman spectroscopy, or a combination of fluorescence and Raman spectroscopy. In vivo fluorescence measurements were followed by in vitro NIR
Raman measurements on human cervical biopsies. Fluorescence spectra collected at 337, 380 and 460 nm excitation were used to develop a diagnostic method to differentiate between normal and dysplastic tissues. Using a fluorescence diagnostic method, a sensitivity and specificity of 80 % and 67 % were observed for differentiating squamous intraepithelial lesions (SILs) from all other tissues. In accordance with another aspect of the invention, using Raman scattering peaks observed at selected wavenumbers, SILs were separated from other tissues with a sensitivity and specificity of 88 % and 100 %. In addition, inflammation and metaplasia samples are correctly separated from the SILs.

Description

W0 96/_8084 P~l/~l.. , ? 11 i ` `' ~ i ~ 1 - 2190374 DESCRIPTION
OPTICAD MET~IOD AND APPARATUS FQF~ THE DT~r~NOSIS
OF rFl~vIcAL PR~r~ ~c~S USING R1~'~ ,l~NT) F~IJORESCENCE SPECTROSCOP~
.--BA~ tKUUN~J OF T~IE lNv~ .~LU~
The invention relates to optical methods and 10 apparatus used for the diagnosis of cer~lical precancers.
Cervical cancer is the second most common malignancy in women worldwide r exceeded only by breast cancer and in the United ~States, it is the third most common neoplasm 15 of the female genital tract. 15, o00 new cases of invasive cervical cancer and 55, 000 cases of carcinoma in situ (CIS) were reported in the U.S. in 1994. In 1994, an estimated 4, 600 deaths occurred in the United States alone from cervical cancer. Xowever, in recent years, 20 the inciaence Qf pre-invasive squamous carcinoma of the cervix has risen dramatically, especially among young women . Women under the age of 35 years account f or up to 24 . 5% of patients with invasive cervical cancer, and the incidence is continuing to increase for women in this age 25 group. It has been estimated that the mortality of cervical cancer~ may rise by 20% in the next decade unless further i,,,u~uv~ ts are made in detection techniques The mortality associated with cervical cancer can be 3 o reduced if this disease is detected at the early stages of development or at the pre-cancerous state (cervical intraepithelial neoplasia (CIN~ ) . ~ Pap smear is used to screen for CIN and cervical cancer in the general female population. This technique has a false-negative error 35 rate of 15-40%: An abnormal pap smear is followed by colposcopic examination, biopsy and~histologic confirmation of the clinical diagnosis. Colposcu~y Wo 96l28084 PCrlUS96/02644

2 ~ 374 requires extensive training and its accuracy for diagnosis is variable and limited even in expert hands~
A diagnostic method that could improve the perfo~mance of colpoæcopy in the hands of leæs~ experienced 5 practitioners, eliminate the ~eed for multiple biopsies and allow more effective wide scale diagnosis cQuld potentially reduce~ the mortality àssociated with cervical cancer.
Recently, fIuoreæcence, in~rared absorption and Raman spectroæcopies have been proposed for cancer and precancer diagnosis. Many groups have æucceææfully demonætrated their uæe in various organ~systems. Auto and dye induced fluoreæcence have shown promise in 15 recognizing atherQsclerosis and various types of cancers and precancers. Many groups have demonstrated that autofluorescence may be used for dif~erentiation of normal and abnormal tisæues in the human brea6t and lung, bronchus and gastrointestinal tract. Fluorescence 20 spectroscopic techniqueæ have alæo been investigated fo~
improved detection of cervical dysplasia.
An automated diagnostic ~ethod with improved diagnostic capability CQUld allow ~aster, more :ef~fective 25 patient management and potentially further reduce mortality .
S~Ia~Y OF T~R INVENTIo~
The present invention demonætrates that fluQrescence and Raman spectrQsCopy are promlsing techniques for the clinical diagnosis of cervicaI precancer. ~ =
Studies were~conducted ln vitro to establish a strategy for clinisal in vivo diagnosis,- and indlcated, that 337, -380 and ~60 nm (~ 10 nm) are optimal excitation wavelengths for the identification of cervical precancer.

W0 96/28084 P~
~ 3 _ == ~ 9 0 3 7 4 In vivo fluorescence spectra collected at 337 nm from 92 patients were used to develop æpectroscopic methods to dif ferentiate normaI from abnormal tissues . Using empirical parameters such as peak intensity and slope of 5 the spectra, abnormal and normal tissues were differentiated with a sensitivity and speciiicity of 859 and 75~6. Using multivariate statistical methods at 337 nm excitation, normaI and squamous intraepithelial lesions (SI~s - lesions with dysplasia and human lO papilloma virus IHPV) ) were differentiated with a sensitivity of 919~ and specificity of 8296. At 380 nm excitation, can be differentiated from columnar normal tissues and frdm tissues with inflammation with a sensitivity of 7796 and a specificity of 7296. At 460 nm 15 excitation, high grade SILs (moderate to severe aysplasia carcinoma) and low grade SI~s (mild dysplasia, HPV) were differentiated with a sensitivity and specificity of 73~6 and 8596. As used herein the calculations of sensitivity and spe~ificity are presented in detail in Appendix I.
The present invention also contemplates the use of Raman spectroscopy for the diagnosis ~f disease in tissue Raman scattering signals are weak compared to fluorescence. However, Raman spectroscopy provides 25 molecular specific information and can be applied towards tissue diagnosis. The present invention exploits the capabilities of near infrared (NIR~ Raman s~ectroscopy and fluorescence spectroscopy to di~i:erentiate normal, metaplastic and inflammatory tissues from SI~s. Further, 30 the ability of these techniques to separate high grade dysplastic lesions from low grade Iesions is also exploited.
The invention also contemplates the use of 35 fluorescence spectroscopy in combinatipn with Raman spectroscopy for the diagnosis of disease in tissue.

W096128084 r~ )., 6~n~'11 ~t,~',i`;tt''. 4_ 219~374 More particularly, the present invention - -contemplates methoas and apparatus for the optisal diagnosis of cervical precancers. Specifically, one embodiment of the method of the present invention detects 5 tissue abnormality in a tissue sample by illuminating a tissue sample with a first electrotnagnetic radiation wavelength selectea to cauæe the tissue sample to produce a fluorescence lntensity spectrum indicative of ~tissue~
abnormality. Then, a first fluorescence intensity 10 spectrum emitted from the tissue sample as a result of_ illumination with the first wavelength is detected. The tissue sample is then illuminated with a second electro~agnetic radiation wavelength selected to cause the tissue sample to produce a fluorescence i~tensity 15 spectrum indicative of tissue abnormality, and a second fluorescence: intensity spectrum emitted fr~m the tissue sample as a result of illuminatlon with the second wavelength is detected. Finally, a probability that the tissue sample is normal or abnormal is calculated ~rom 20 the f irst fluorescence intensity spectrum, and a degree of abnormality of the cervical tissue sample is--calculated from tXe second fluorescence intensity spectrum .
The invention further contemplates that each of the calculations include principal component analysis of the first and second spéctra, relatlve to a plurality of preprocessed spectra obtained frQm tissue samples of known diagnosis. ~The invention also contemplates normalizing the first and second spectra, relative to a maximum intensity within the spectra, and mean-scaling the first and second spectra as a function of a mean intensity of the first and second spectra.
Another emboaiment of the invention includes detecting tissue abnormality in a diagnostic tissue sample by illuminating the tissue sample with an wo96n8084 r~ . ?'1 5 _ ? l q o 3 7 4 illumination wavelength of electromagnetic radiation selected to cause the tissue sample to emit a Raman spectrum comprising a plurality of wavele~gths shifted from the illumination wavelength. A plurality of peak 5 intensities of the Raman spectrum at wavelength shifts selected for their ability to distinguish normal tissue from abnormal tissue are detected, and each of the plurality of ~t~tf~tl peak intensities :at the wavelength shifts are compared with intensities of a Raman spectrum 10 from known normal tissue at corresponding wavelength shifts. Abnormality of the tissue sample is assessed as a function of the comparison. This embodiment also contemplates calculating a ratIo between selected~
intensities o~ the Raman spectrum, and detecting 15 Ahn~rr~l ity of the tissue sample as a function of the rat io .
The invention also contemplates calculating a second ratio between another two ot the plurality of peak 20 intensities, an=d detecting a degree of tissue abnormality as a function of the second ratio_ ~
In yet another embodiment of the method of the present invention, calculation o~ one or more ratios 25 between Raman spectrum intensities is used alone to detect tissue abnormality without comparing individual intensities with those of known normal tissue.
The present invention also contemplates the 30 combination o~ Raman and fluoresce~ce spectroscopy to detect tissue abnormality The apparatus of the present invention includes a controllable; l l llmi nAtion device fsr emitting a plurality 35 o~ tr~rAgnetiC radiation wavelengths selected to cause a tissue sample to produce a fluorescence intensity spectrum indicative of tissue abnormality, an optical WO 96/28084 PCrlUS96102644 ~1 90374 system for applying the plurality of radiation wavelengths to a tissue sample, a fluorescence intensity spectrum detecting device i-or detecting an intensity of fluorescence spectra emitted by the sample as a result of 5 illumination by the plurality of electromagnetic radiation wavelengths, a data processor, connected to ~he ,"
detecting device, for analyzing detected fIuorescence ~~
spectra to calculate a probability that the sample is abnormal A Raman spectroscopy apparatus in accordance with-the present invention includes an illumination~-device for generating at least one i l l llrni nAtion ~zavelength of electromagnetic radiation selected to cause a tissue 15 sample: to emit a Raman spectrum comprising a plurality of wavelengths shifted from the; ~ min~tion wavelength, a Raman spectrum detector for detecting a plurality of peak intensities of the Raman spectrum at selected wavelength shifts, and a programmed computer connected to the Raman 20 spectrum detector, programmed to compare each of the plurality of detected peak int~nRi t; ~q with corresponding peak intensities of a Raman spectrum from known norma tissue, to detect tissue abnormality.
These and other features-and advantages of -the present invention will become apparent to those-of ordinary skill in thls art with reference to the appended drawings and following detailed descrlption.

3 o P~T~ I?ES~'~TPTION OF ~ DRA~INGS
FIG. l is a block diagram of an exemplary fluorescence spectroscopy diagnostic apparatus, in accordance with the present invention.
: . :

W096/28084 r~
' G ~tt~ ; 7 _ ; ~ 1 9 0 3 7 4 FIG. 2 is a block diagram an exemplary Raman spectroscopy diagnostic apparatus, in accordance with the present invention.
FIG. 3A, FIG. 3B and FIG. 3C are flowcharts of exemplary fluorescence spectroscopy diagnostic methods, in accordance with the present inventione FIG. 4A, FIG. 4B, FIG. 4C and FIG. 4D are flowcharts of an exemplary Raman spectroscopy diagnostic method, in accordance with the present invention.
FIG. 5 and FIG. 6 are graphs depicting the performance of the fluorescence diagnostic method of the present invention, with 337 nm excitation.
FIG. 7A, FIG. 7B and FIG. 8 are graphs illustrating the performance of the fluorescence spectrum diagnostic method of the present invention at 380 nm excitation.
FIG. 9A, FIG. 9B and FIG. 10 are graphs illustrating the performance of the fluorescence spectrum diagnostic method of the present invention to distinguish s~uamous normal tissue from SIL at 460 nm excitation.
FIG. llA, FIG. llB and FIG. 12 are graphs illustrating the performance of the fluorescence spectrum diagnostic method of the present invention, to distinguish low grade SIL from high grade SIL at 460 nm excitation.
FIG . 13 is ~ graph depicting measured r~ m; nG
Raman spectra with high and low signal to noise ratios.
", FIG. 14 i5 a graph depicting the elimination of fluorescence from a Raman spectrum, using a polynomial fit .

4 ~ ~ ~
8 - 2 t 9 0 3 7 4 FIG. 15 is a graph comparing low and high 6ignal to noise ratio rhodamine Raman spectra.
FIG. 16 is a graph depicting a typ~cal pair of Raman

5 spectra obtai~ed frDm a patien~ with dysplasia.
FIG. 17 is a histogram of the Raman band at 1325 cm~
1 illustrating patient to patient uariation.
lD FIG. 18A and FIG. 18B are graphs illustrat~ng ~he diagnostic capability of the Raman spectroscopy diagnostic method of the present invention.
FIG. l~ is a~other graph depicting the diagnostic 15 capability of the Raman spectroscopy diagnostic method of the present invention.
FIG. 20 and FIG. 21 are graphs ilIustrating the diagnostic capability of the Raman spectroscopy 20 diagnostic method of the present invention.
FIG. 22 is a graph of a hypothetical distribution of test values.
DETATT T~'n DESCRIPTION .
I. Introduction To illustrate the aduantages of the present 30 invention fluorescence spectra were collected i~ vlvo :at colposcopy from 2D patients. Fluorescence spectra were measured from three to fDur colposcopically normal and three to four colposcopically abnormal sites as identified by the practitioner ~using techniques known in 35 the art. Specifically, in cervical tissue, nonacetowhite epithelium is considered normal, whereas acetowhite ~
epithelium and the presence of vascular atypias (such as Wo 96128084 PCrluS96/02644 ~ l ~ ` 9 2 ~ 90374 punctuat ion, ~ mosai cism, and atypi cal vessel 8 ) are considered abnormal. One normal and abnormal site from those inveGtigated were biopsied from each patient.
These biopsies were snap frozen in li~luid nitrogen and 5 stored in an ultra low temperature freezer at -85C until Raman measurements were made.
II. Diagnoatic Apparatu~
10 1. ~luQ~esçençe SPectro~coPY Diaqno3tic APParatu~
Fluorescence spectra were recorded with a spectroscopic system incorporating a pulsed nitrogen pumped dye las=er, an optical~ fiber probe and an optical 15 multi-channel analyzer at colposcopy. The laser characteristics for the study were: 337, 3go and 460 nm wavelengths, transmitted pulse energy of 50 uJ, a pulse duration of 5 ns and a repetition rate of 30 Hz. The probe includes 2 excitation fibers,~ one for each 20 wavelength and 5 collection fibers. Rhodamine 6G (8 mg/ml) was used as a standard to calibrate for day to day variations in the detector throughput. The spectra were background subtracted and normali2ed to the peak intensity of rhodamine. The spectra were also calibrated 25 for the wavelength dependence of the system.
FIG. 1 is an exemplary spectroscopic system for collecting and analyzing fluorescence spectra from cerYical tissue, in accordance with the present 30 invention, incQrporating a pulsed nitrogen pumped dye laser 100;- an optical fiber probe 101 and an optical multi-channel analyzer 103 utilized to record fluorescence spectra from the intact cerYix at colposcopy. The probe 101 comprises a central fiber 104 35 surrounded by a circular array of six fibers. All seven fibers have the same characteristics (0.22 NA, 200 micron core diameterl. Two of the peripheral fibers, 106 and Wo 96/28084 P~~ .'n7' 1 ~

107, deliYer excitatlon light to the tissue sur~ace;
fiber 106 delivers excitation light from the nltrogen laser and f iber 107 delivers light iro~ the dye module ~overlap of the~illumination area viewed by both optical fibers 106, 107 is greater than 859~). The purpose of the rf~ n~n~ five ~ibers (104 and 108-111) is to collect the emitted fluorescence from the tissue sur~ace directly illuminated by each excitation-fibers 10~, 107. A quartz shield 112 is placed at the tip of the probe 101 to provide a subst~nt;~lly fixed distance between the fibers and the tissue surface, 60 fluoresce~ce intensity can be reported in calibrated units.
Excitation light at 337 r~ excitation was focused into the proximal end of excitation fiber 106 to produce a 1 mm diameter spot at the outer face of the shield 112.
Excitation light from the dye module 113, couplea into excitation fiber 107 was produced by using appropriate~
fluorescence dyes, in this example, BBQ (lE-03M in 7 parts toluene and 3 parts ethanol) was used to generate light at 380 nm excitation, and Coumarin 460 -(lE-02 M in ethanol) was used to generate light at 460 nm excitation.
The average transmitted pulse energy at 337, 380 and 460 nm excitation were 20, 12 and 25 mJ, respective1y. The laser characteristics for this embodiment are: a 5 ns pulse duration and a repetition rate of 30 ~z, however other characteri~tics would also be acceptable.
Excitation fluences should remain low enough so that cervical tissue is-not vaporized and so that significant photo-bleaching does not occur.~- In arterial tissue, for example, signi:f icant photo-bleaching occurs ~above excitation fluences of 80 mJ/mm.
The proximal ends of the collection fibers 104, lOa-111 are arranged in a circular array and imaged at the entrance slit of a polychromator 114 (Jarrell Ash, Monospec 18) couplea to an intensified 1024-aiode array Wo 96/28084 ~ ~ "~
- 2 1 9 ~ 3 7 4 116 controlled by a multi-channel analyzer 117 (Princeton Instruments, OMA) . 370, 400 and 470 nm long pass filter6 were used to block scattered excitation light at 337, 380 and 4~0 nm excitation respectively. A 205 ns collection gate, synchronized-to the leadi~g eage of the laser pulse using a pulser 118 (Princeton Instruments, PG200), effectively eliminated the effect6 of the colposcope~s white light illumination during fluorescence mea,ju~ nts . Data acquisition and analysis were controlled by computer 119 in accordance with the f luorescence diagnostic method described below in more detail with reference to the flowcharts of EIG. 3A, FIG.
3s and FIG. 3 2 . R s~ectro~co~Y Diaqno~tic APParatu~
FIG. 2 is an exemplary apparatus for the collection of near-IR Raman spectra in accordance with the present inventio~l Near- IR Raman measurements were made in vi tro using the system shown in FIG. 2, however the system of FIG. 2 could be readily adapted for use in vivo, for example by increasing laser power and by using a probe structure similar to that of probe 101 shown in FIG. lA
and FIG. lB. ~ 4D m 1~1 GaAIAs diode laser 200 (Diolite 80D, Li O~ix, CA) excites the samples near 789 nm through a 200 micron glass fiber 201. The biopsies measuring about 2 x 1 x 1 mm are placed moist in a quartz cuvette 202 with the epithelium towards the face of the cuvette 202 and ~he be~am. The excitation beam is incident at an angle of; appro~cimately 75 degr=ees to avoid specular reflection and is focused to a spot size of 200 ~m at the tissue surface. A bandpass (BP) filter 203 with a transmission of 85~6 at 789 nm is used to clean up the output of the laser 200. The laser power at the sample is maintained at approximately 25 mW. The scattered Raman signal is collected at an angle of 90 from the excitation beam and imaged on the entrance slit of the Wo 96l28084 PCTIUS96/02644 4 ~ ~el~ 12 _ 2~9n374 detection system 204, however other angles would al~io be acceptable. A holographic notch (~IN) filter 206 (HSNF
789, Kaiger Optical Systems, MI) with an optical densi~y >6 at 789 nm is used to attemlate the elastic scattering.
5 The detection system 204 includes an imaging spectrograph 207 (500IS, Chromex Inc. NM) and a liquid nitrogen cooled ~, CCD camera 208 with associated ~camera controller 209 (LN-1152E, Princeton Instruments, NJ) . The spectr~graph 207 was used with a 3D0 gr/mm grating blazed at 500~nm which 10 yielded a spectral resolution of 10 cm~l with an entrarce slit 0~ 100 ,um.
Detection system 204 is controlled by computer 211 which is programmed in accordance with the ~aman 15 spectroscopy diagnostic method described below in detail with refere~ce to the ~lowcharts of FIG. 4A, FIG. 4B, FIG. 4C and FIG. 4D.
Raman spectr~ were measured over a range of 500 -__ 20 2000 cm~l with respect to the excitation frequency and each sample spectrum was integrated for 15 minutes, however, shorter i~tegration times would also be acceptable ~ombined with higher laser intensity. Each background subtracted spectrum was co~rected f or~
25 wavelength dependent response of the spectrDgraph~207, camera 208, grating and filters 203 and 206. The system was calibrated for day to day through"out variations using naphthalene, rhodamine ~G and c~rbon tetrachlorlde. The Raman shift was found to be accurate to 7 cm 1 and the 30 intensity was found to be constant within 12~6 of~ the ~..
mean .
The systems of FIG 1 and FIG. 2 are e:~emplary embodiments and should not be considered to limit the 35 invention as claimed. It will be understoDd that apparatus other than that depicted in FIG. 1 and~FIG. 2 may be used without departing from the~ scope of the invention.

Wo~6/28084 P~~ ?''' t~ 13_ 2190374 III. Diagnostic Methods 1. Develo~ment of Diacrnostic MethQds A. Two-Stage ~ethod Dev.o7~ at 337 nm Excitatio~
The parameters f or stage 1 of the two - stage method:
relative peak intensity (peak intensity of each sample divided by the average peak intensity of corresponding 10 normal (squamous) samples from the same patient) and a linear apprQximation of slope of the spectrum from 420-440 nm were calculated from the fluorescence spectrum of each sample in the calibration set. The relative peak i~tensity accounts for the inter-patient variation of 15 normal tissue fluorescence intensity. A two-dimensional scattergram of the two diagnostic parameters was plotted for all the samples in the calibration set. A linear decision line was developed to minimize misclassification (non diseased vs. diseased). Similarly, the parameters 20 for stage 2 of the method: slope of the spectrum from 440-460 nm of each diseased sample and average slope from 420-4~0 nm of spectra of corresponding normal (squamous) samples were ~ calculatea from the calibration set . A
scattergram of the these two diagnostic parameters was 25 plotted for all diseased samples. Again, a linear decision line was developed to minimize misclassification (low grade ~IL `vs . ~igh grade :SIL) _~=The optimized method was implemented on spectra of each sample in the prediction set. The optimal decision lines developed 30 from the data in the calibration set were compared to that developed in the initial clinical study for both stages of the method. The two-stage fluorescence diagnostic method is disclosed in more detail in application Serial No 08/060,432, filed May 12, 1993, 35 assigned to the same assignee as the present invention.
The ~ l nqlir~ n~ this prior application is expressly incorporated herein by reference.

WO 96128081 I ~
4~J~ 14 _ 21 90374 B. Multi-Va~iate Stati3tic~1 Metllod Dev~i:~ t The five primary steps involved in the multivariate statistical method are l) preprocessing of spectral data 5 from each patient to account for inter-patient variation, 2) partitioning of the preproce5sed spectral data~ from .
all patients into calibration and predIction sets, 3) dimension reduction of the preprocessed spectra in the~
calibration set using principal componen~ analysis, 4 ) lO selection of the diagnostically mo5t useful principal components using a two-sided unpaired t-test and ~) ~
development of a~ optimal classification scheme based on Bayes theorem using the diagnostically useful principal component scores of the calibration set as inputs These 15 five individual steps of the multivariate statistical method are presented below in more detail.
1) Preproce~sing: The objective of preprocessing is to calibrate tissue spectra for i~ter=patient variation 20 which might=obscure differences in the spectra of different tissue types. Four methods of preprocessing were ~nvo~ced on the spectral data: 1) normalization 2) mean scaling 3 ) a combination cf nArrn~ t; on and mean scaling and 4) median scaling.
Spectra were normalizea by dividing the fluoresce~lce intensity at eac~emissio~ wavelength by the maximum fluorescence intensity of that sample. Normalizing a fluorescence spec~rum removes a}:solute intensity 30 information; methods developed ~from normalized fluorescence spectra rely on differenccs in spectral line shape information: fo~ diagnosis~. If the contribution ~f the absolute inte~zsity information is not significant, two advantages are reali2ed ~y utilizing normalized 35 spectra: -1~ it is no longer necessary to calibrate ~or inter-patient variation of normal tissue fluorescence ::
intensity as in the two-stage method, and 2) W0 96/28084 r~
2 1 9 ~ 3 7 4 identification of a colposcopically normal reference site in each patient prior to spectroscopic analysis is no longer needed.
Mean scaling was performed by calculating the mean spectrum for a patient (using all spectra obtained from cervical sites in that patient) and subtracting it from each spectrum in that patient. Mean-scaling can be performed on both unnormalized (original) and normalized spectra. Mean-scaling does not require colposcopy to identify a reference normal site in each patient prior to spectroscopic analysi6. However, unlike normalization, mean-scaling displays the differences in the fluorescence spectrum from a particular site with respect to the average spectrum from that patient. Therefore this method can enhance differences in fluorescence spectra between tissue categories most effectively when spectra are acquired f rom approximately equal numbers of non diseased and diseased sites from each patient.
Median scaling is performed by calculating the median spectrum for a patient (using all spectra obtained from cervical sites in that patient) and subtracting it ~rom each spectrum in that patient. ~ike mean scaling, median scaling can be performed on both unnormalized (original) and normalized spectra, =and median scaling does not require colposcopy to identi~y a reference normal site in each patient prior to spectroscopic analysis. However, unlike mean scalir~g, median scaling does not require the acquisition of spectra from equal numbers of non diseased and diseased sites from each patient .
2) Calibration and Prediction Data Sets: The preproc~ssed spéctral data were randomly assigned into either a calibration or prediction set. The multivariate statistical method was developed and optlmized using the WO96/28084 I~~ ~?~1~
~ ` f ~ - 16 - 2 1 9 0 3 7 4 calibration set. It was then tested prospectively on the predictioll data set.
3) Principal C ~ t AnalysiO Principal component 5 analysis (PCP.) is a linear model which transforms the original Yariables of a fluorescence emission spectrum into a smaller set of linear combinations of the original variables ca~ led principal components that account for~= =
most of the variance of the original data set. ~Principal 10 ~ '`nt analysis is described in Dillon W.R., Goldstein M., Multivariate Analysis: Methods and Applications, John Wiley and Sons, 19-84, pp. 23-52, the disclosure~of which is expressly incorporated herein by refere~ce. While=PCA
may not provide direct insight to the morphologic and 15 biochemical basis of tissue spectra, it provides a novel approach of condensing all the spectral informat~ion into a few manageable components, with minimal information loss. Furthermore, each principal component can be easily related to the origi~al emission spectrum, thus 20 providing insight into diagnostically useful emission =
variables . -- ~
Prior to PCA, a data matri-x is created where each ro~ of the matrix rnnti:lin}~ the preprocessed fluorescence 25 spectrum of a sample and each column contains the pre-processed fluQrescence intensity at each errission wavelength. The data matrix D (r x c), consisting of r rows (corresponding to r total samples from all patients in the training set) and c columns (cQr~espondiny to 30 i tensity at c emission wavelengths) can be written as:
.

WO 96/28084 ~ PCrlUS96/02644 f ~ 2 ~ ~ ~ 3 7 4 ' Dll Dl2 DlC
D2l D22 . . D2c D = (1) \ Drl Dr2 Drc ~
-The f irst step in PC~ is to calculate the covariance matrixr ~ Eirst, each column of the preprocessed data matrix D i8 mean-scaled. The mean-scaled preprocessed 5 data matrix, Dm is then multiplied by its transpose and each element o~ the resulting square matrix is divided by (r-1), where r is the total number :of samples. The equation for calculating Z is defined as:
Z = 11 (Dm/ Dm) (2) The square cov~riance matrix, Z (c x c) is decomposed into its respective eigenvalues and eigenvectors.
Because of experimental error, the total number of eigenvalues will always equal the ~total number of columns 15 (c) in the data matrix D assuming that c ~ r. The goal is to select n ~ c eigenvalues that can describe most of the variance of the original data matrix to within experimental error. The variance, V accounted for by the first n eigenvalues can be calculated as follows:

/ n V = ~LOO j 1 (3) ~i ~ j=l ) t~ t ~ 18 ~ 2 1 9 o 3 7 4~CI;~S96102~44 The criterion used in this anaIysis was to retain the f irst n eigenvalues and corresponding eigenvectors that account for 99 % of: the variance i~ the original data set .

Next, the princlpal componellt score matrix:~can be calculated according to the following ectuation:
R = D C ~ 4 ) 10 where, D (r x c) is the preprocRssed data matri~ and C (c x n~ is a matrix whose columns contain the n eigenvectors which correspond to the first n eigenvalues. Each row of the score matrix R tr x c) corresponds to the principal component scores of a sample a~d each column corresponds 1~ to a principal component. The principal components are mutually orthogon~[l to Rach other.
Finally, thR component loading is calculated for each principal component. The component loading 20 represents the correlation between the principal rrrnrrnf~nt and the variables of t~e original fluorescence emission spectrum. The component loading can be calculated as shown below:
CL~ ~ (5 2~ :
where, C~ij represents-~the correIat'lon between the ith~
variable (preprocessed intensity at Ith emission wavelength) and the j th principal component . Ci~ is the ith component of = the j th eigenvector, A] is the j th 30 eigenvalue-and Sii is the varia ce:of the ith variable.

Wo g6128084 ~ 19 - PCr/US96/02644 Principal component analysis was performed on each type of preprQcessed data matrix, described above.
Eigenvalues accounting for 99~ of the variance in the original preprocessed data set were retained The 5 corresponding eigenvectors were then multiplied by the original data matrix to obtain the principal component score matrix R.
4) Student's T-Test: Average values of principal 10 component scores were calculated for each histo-pathologic tissue category for each principal component obtained from the preprocessed data matrix. A two-sided unpaired student' 8 t-test was employed to determine the diagnostic contribution of each principal component.
15 Such a test is disclosed in Devore J.L., Probability and Statistics for Engineering and the Sciences, Brooks/Cole, 1992, and in Walpole R.E., Myers R.H., Probability and Statistics for Engineers and Scientists, Macmillan Publishing Co ., 1978 , Chapter 7 , the disclosures of which 20 are expressly incorporated hereirl by reference. The hypothesis that the means of the principal component scores of two=tissue categories are different were tested for 1) normal squamous epithelia a~d SILs, 2~ columnar normal epithelia and SILs and 3) inflammation and SILs.
25 The t-test was extended a step further to determine if there are any statistically signifïcant differences between the means of the principal component scores of high grade SI=Ls and low grade SILs, Principal components for which the hypothesis stated abQve were true below the 30 o . 05 level of significance were retained for further analysis .
.

5) Logistic Discrimination: Logistic discriminant analysis is a statistical technique that can be used to 35 develop diagnostic methods based on posterior:
probabilities, overcoming the drawback of the binary decision scheme employed in the two-stage method. This w0 96128084 r~
~Q~ ~ 21~0374 statistical classification method is based on sayes theorem and can be used to calculate the posterior probability that an unknown sample belongs to each of the possible tissue categories identified. Logistic 5 discrimination ~is discussed in Albert A., Harris E.K., IYultivariate Interpretation ol~ t~linical Laboratory Data, Marcel Dekker, 1987, the disclosure -of which is expressly incorporated herein by reference. Classifying the unknown sample into the tissue category for which its 10 posteriDr~probability is highest results in a classif ication scheme that mini~Lizes the rate of misclassification~
For two diagnostic categQrles, G1 and G2, t~e 15 posterior probability of being a member of G1, given measurement x, according to Bayes theorem is:
1I P (x¦G1) P (G1) C (2 ¦1) +P (x¦G2) P (G2) C (1¦2) where P (x¦Gi) is the conditional ~oint probability that a 20 tissue sample of type i will havë principal component score x, and P (Gi) is the prior -probability of finding tissue type i in the sample population. C(j ¦i) is the cost of misclassifying a sample into group ~ when the actual membership is group i.
The prior probability P (Gi) is an estimate of the likelihood that a sample of type i belongs to a particular group when no infDrmation about it is :
available. If the sample is considered representative of 30 the population, the observed proportions~ Df ~cases in each group can serve a~ estimates of_the prior probabilitiesJ
In a clinica~ setting, either historical incidence figures aFprop~iate fbr the pat1ent population can be used to generate prior probabilities, or the W0 96128084 ~ . PCTIUS96102644 practitioner' s colposcopic assessment of the likelihood of precancer can be used to estimate prior probabilities.
The conditional probabilities can be developed from 5 the probability distributions of the n principal component scores for each tissue type, i. The probability distributions can be modeled using the gamma function, which is characterized by two parameters, alpha and beta, which are related to the mean and standard 10 deviation of the data set. The Gamma function is typically used to model skewed distributions and is def ined below: -f (x;a,~ xa-1e ~ (7l The gamma function can be used to calculate the conditional probability that a sample from tissue type i, will exhibit the principal component score, x. If more than one principal component is needed to describe a sample population, then the conditional joint probability 20 is simply the product of the conditional probabilities of each principal component (assuming that each principal component is an independent variable) for that sample populat ion .
25 C. ~ultivari~te Analysis of Tissue Fluorescence Spectr~
- 1) SI~s vs. Normal S~auamous Tissue at 337 ~ excitation A summary of the f luorescence diagnostic method 30 developed an~` tested in a previous group of 92 patients (476 sites) is presented here.: The spectral data were preprocessed by normalizing each spectrum to a peak intensity of one, followed by mean-scaling. Mean scaling WO ~)6128084 ~ , 0 ~ 22 ~ ~ ~ 9 03 / ~ E ~'11 is performed by calculating the mean spectrum ~or a patient (using all spectra obtained from cervical sites in that patient) and subtractin~ it from each spectrum in that patient. Next, principal component analysis ~PCA) 5is used to transform the original variables of each preprocessed fluorescence emission spectrum into a "
smaller set of linear combinations called principal components that account for 99% of the variance :of the.
original data set. Only the diagnostically useful 10 principal components are retained for further analysis.
Posterior probabilities for each tissue type are determined for all samples in the data set using calculated prior and conditional joint probabilities.
The prior probability is calcuLated as the percentage of 15 each tissue type in the data. The conditional probability was calculated from the gamma function which modeled the probability distributions of the retained principal components scores for each tissue category. ~
The entire data set was split ln two groups: calibration 20 and prediction data set such that their prior probabilities were approximately equal. The method is optimized-using the calibration qet and then implemented on the prediction set to estimate its performance in an unbiased manner. The methods using PC~ and ~ayes theorem 25 were developed using the caIibration set=conslsting of previously collected spectra from 46 patients (239 sites~ . These methods were ~then applied to the prediction set (previously collected spectra from another 46 patientsi 237 sltes) and the current data set of 36_ 30 samples More specifically, at 337~nm excitation, fluoresce~ce spectLa were acqui=red from a total of 476 sites in 92 patients. The data were randomly assigned to 35 either a calibration set or prediction set with the condition that both sets contain roughly equal number of WO 96/_8084 ~ P~ l/~. 6.' ? ~ ~
2 1 9 ~ 3 7 4 samples i-roTn each histo-pathologic category, as ~hown in Table l Table 1. (a) Hi~to-path~l 9~; c cla~sification of samples in the training and the validation ~3et ~T~n;ned at 337 nm excitation and (b) hi~tological ~ if;~ation of cervical samples ~e. L~ ~,scopically interrogated in vivo from 40 patient~ at 380 nm excitation and 24 patients in 460 nm excitation.
(a) ~Iistology Training Set Validation Set Squamoua Normal 127 126 Columnar Normal 25 25 Tnfl, tion 16 16 Low Grade SIL 40 40 Eligh Grade SIL 31 30 (b) Eistology 380 nm excitation 460 nm excitation (40 patientl3) (24 patient~) Squamou~ Normal 82 76 25.. ~ Normal 20 24 Inf lammation 10 11 Low Grade SIL 28 14 /Iigh Grade SIL 15 22 W096/28084 r~~ n7'~1 S 2~9~374 The ra~dom aeeignment ensured that not all spectra from a single patient were ~-nnt~lnefl 111 the same data:
set. The purpose of the calibratio~ set is to develop and optimize the method and the purpose of the prediction 5 set is to prospectively test its accuracy in an unbiased manner. ~The two-stage method and the multivariate statistical methoa were optimized using the calibration set. The performance of these~methods were then tested prospectively on the prediction set.
Principal component analysis of mean-scaled normalized spectra at 337 nm e*citation frQm the calibration data set resulted in 3 principal components accounting for 9~~ of the total variance. ~nly, the 15 first two principal components obtained ~rom the preprQcessed data matrix containing mean-scaled normalized spectra demonstrate . the statisticalIy most significant differences ~P ~ Q.05) between normal squamous tissues and SILs (PCl. P < lE-25; ~C2: P <
20 0 . 006~ . ~The two-tail P values of the scores of the tbird principal component were nQt statistically significant (P
~ 0.2). TherefQre, the rest of the analysis was performed using these two principal components. All of the principal ~ nn~nts are ~nrl~ 1 in Appendix II
For excitation at 337 nm, the prior probability was determined by calculating the percentage of each tissue type in the calibration set: 6596 normal squamous tissues and 35% SILs. More generally, ~prior probabilities should 30 be selected tQ describe the patient population under ~~
study; the values used here are appropriate as they describe the prediction set as ~well .
Posterior pr-obabilities of belonging to ea~h tifiszue 35 type (normal squamous or SIL) were calculated iQr all samples in the calibration set, using the known prior: :
probabilities and the conditional probabilities r ~ 2l ~a374 calculated from the gamma function. A cost of misclassification of SILs equal to 0 . 5 was assumed.
FIG. 5 illustrates the posterior probability of belonging to the SII, category. The posterior probability is 5 plotted for all samples in the calibration set. This plot indicates that 75~ of the high grade SILs have a posterior probability greater than 0.75 and almost 90~ of high grade SILs have a posterior probability greater than 0 . 6 . While 85% of low grade SILs have a posterior probability greater than 0 . 5, only 60~6 of low grade SILs have a posterior probability greater than 0 . 75 . More than 80~6 of normal squamous epithelia have a posterior probability less than 0.25. Note that evaluation of normal ~ rnn~7^ epithelia and samples with inflammation 15 using this method results in classifyinq them as S:[Ls.
FIG. 6 shows the percentage of normal squamous tissues and SII,s correctly classified versus cost of misclassification of SILs for the data from the 20 calibration set. An increase in the SIL
misclassif ication cost results in an increase in the proportion of correctly classified SILs and a decrease in the proportion of correctly classified normal squamous tissues. Note, that varying the cost from .4 to .6 25 alters the cIassification accuracy of both SILs and normal tissues by less than 15% indicating that a small change in the cost does not significantly alter the performance of the method. An optimal cost of misclassification would be 0.6-0.7 as this correctly 30 classifies :~lmost 95~ of SILs and 80% of normal squamous epithelia, for the prior probabilities used and is not sensitivity to small changes in prior probability.
The method was implemented on mean-scaled spectra of 35 the prediction set, to obtain an unbiased estimate of its accuracy. The two eigenvectors obtained from the calibration set were multiplied by the prediction matrix Wo 96/28084 PClrUS96/02644 ~ ~ r~ r (~ - 2 6 - 2 ~ 9 ~ 3 7 4 to obtain the new principal component score matrix.
Using the same prior probabilities, a cost of misclassification of SILs equal to 0.5, and conditional joint probabilities calculated from the gamma function,~ -all developed from the calibration r~et, Bayes rule was used to calculate the posterior probabilities for all samples in the prediction set.
Confusion matrices in Tables 2 (a) and 2 (b) show the spectroscopic classification using this method for the calibration set and the prediction set, respectively. A
comparison o~ the sample classi~ication between the prediction and calibration sets indicates that the method per~orms within 7% on an unknown data set of appro~cimately er~ual prior probability.
Table 2. Results of multivariate statistical method applied to the entire fluorescence emission spectra of squamous normal tissues and SILs at 337 nm excitation in (a) calibration set and (b) prediction set.
(a) 2~Classi~ication Squamous Low Grade }~igh Grade Normal SI~
Squamous Normal 83% 15% 10%
S~17% 85% 90%

Wo 96l28084 PCrlUS96/02644 t~ - 27 _ 219~374 (b) Classification Squamous Low Grade Eigh Grade Normal SIL SIL
Squamous Normal 81% 2296 6%
5SIL 19% 78% 94%
The utility of another parameter called the component loadings was explored for. reducing the number 10 of emission variables required to achieve classification with minimal decrease in predictive ability. Portions of the emission spectrum most highly correlated (cDrrelation , 0 . 9 or ~ 0 . 9) with the component loadings were sel~cted and the reduced data matrix was used tc~ regenerate and 15 evaluate the method. Using intensity at 2 emission wavelengths, the method was developed in an identical manner as was done with the entire emission spectrum. It was optimized using the calibration set and implemented on the prediction set. A comparison of the sample 20 classification based on the method using the entire emission spectrum to that using intçnsity at 2 emission wavelengths indicates that the latter method performs equally well in~ classifying normal squamous epithelia and low grade SII,s. The performance of the latter method is 25 6% lower for classifying high grade ~ SILs .
2) SI~s vs . Normal Columnar Epithelia and Tnfl ~ tion at 380 nm Excitation Principal components obtained from the preprocessed data matrix cor~taining mean-scaled normalized spectra at 380 nm excitation could be used to different~iate SI~s from non diseased tissues (normal columnar epithelia and inf lammation~ . The principal components are included in 35 Ap~endix II. Furthermoret a two-sided unpaired t-test Wo 96/280g4 PCT/~S96/02644 r~ 28 - 2~90374 indicated that only principal component 2 (PC2) and principal component 5 (PC5) demonstrated the statistically most significant~differences ~p s 0.05) between SILs and non :~iseased tissues (normal columnar 5 epithelia and inflammation). The p values of the r~--;n;n~ principal component scores were not -statistically significant (p ~ 0.13) Therefore, the rest of the analysis wa6 performed using these three principal ,~ lon~ntS which account collectively for 321i of the 10 variation in the original data set.
FIG. 7A and FIG. 7B illustrate the measured probability distribution and the best fit of the normai probability density function to PC2 and PC5 of non 15 diseased tissues and SILs, respectiveIy. ~ There is reasonable agreement between the measured and calculated probability distribution, for each case. The prior probability was determined by calculating the percentage of each tissue type in the data set: 4196 non diseased 20 tissues and 59~6 SILs. Posterior probabilities of belonging to each tissue type were calculated for all samples in the data set, using=the known=prior =~
probabilities and the conditional joint probabilities calculated from the normal probability density functio~.
FIG. 8 illustrates the retrospe~tive performance~ of the diagnostic method on the same~ data set used tD optimize it. The posterior probability of being classified into the SIL category is plotted for= all samples evaluated.
The results æhown are for a cost of misclassification of SILs equal to 5896. FIG: 8 indicates that 7896 of SILs ~ ~
have a posterior probability greater tha~ 0.~5, 789~ of normal columnar tissues have a posterior probability less than 0.5 and 60~ of samples with ;nfli tion have a posterior:probability less than 0.5. Note that, there are only 10 samples with ;nfli~rrn~tion in this study.

WO 96/28084 . r~ r r~J/~
~ 29 - ~219~)37Dr Tables 3 (a) and ~b) compare ~a) the retrospective performance of the diagnoætic met~od on the data set used to optimize it to ~b) a prospective estimate of the method' s performance using croææ-validation. Table 3 ~a) 5 indicates that for a cost of misclassification of 5096, 749~ o-f high grade SIJ~s, 78~ of low grade SILs, 78~r of normal ~ mn~r samples and 609~ of samples with inflammatiQn are correctly classified. The unbiased estimate of the method's performance ~Table 3 ~b) ) 10 indicates ~that there is no change in the percentage of correctly classlfied SI~s and approximately only a 10~
decrease in the proportion of correctly classified normal columnar samples.

W0 96/28084 ;~ 'Q1' 11 3 ~ -Table 3. (a) A retro~pective ~d (b) prospective e~timate of the multivariate ~tati~tical method' 8 ~ n~ e u~in_ mean-scaled nn~ ed r~pectra at 380 nm excitation to differentiate SIL~ from non disea~ed ti~ue~
(normal columnar epithelia and infl: tion).
(a) lO " Normal I '' LoW Grade High Grade Columnar SIL SIL
Non diseasHd 78% 60% 21% 26%
SIL22% 40% 79% 74%
(b) 15 = ~
Cia~ i " Normal ' " Low Grade High Grade Columnar SIL SIL
Non diseased 65% 30% 22% 26%
SIL35% 70% 78% 74%

3) Squamoua Normal Tissue v~. SIL~ at 460 nm Excitation ~ rincipal components obtained from the preprocessed data matrix cnnt~inin~ mean-scaled normalized spectra at 25 460 nm excitation could be used to differentiate SIL from normal s 3uamous tissue . These principal components are included in Appendix II. Only principal components l and 2 demonstrated the statisticaliy most significant differences (p s O . 05) between SILs and no~rmal squamous 30 tissues. The p values of the remaining principal W0 96/28084 r~ t i'; " ~ 31 - ~ 2i 9~37~
component scores, were not statistically significant (p ~
0 . 06) . Therefore, the rest of the analysis was performed using these two principal components which account collectively for 759~ of the variation in the original 5 data set.
FIG. 9A and FIG. 9B illustrate the measured probability distribution and the best fit of the normal probability density function to PC1 and PC2 of normal 10 squamous tissues and SILs, reæpectively. There is reasonabl~ agreement between the measured and calculated probability distribution, for each case. The prior probabilities were determined to be: 6796 normal squamous tissues and 33~; SILs. Next, posterior probabilities of 15 belonging~ to each tissue type were calculated for all sample6 in the data set. FIG. 10 illustrates the retrospective performance of the diagnostic method on the same data set used to optimize it. The posterior probability of being classified into the SIL category is 20 plotted for all samples evaluated. The results shown are for a cost of misclassification of SI~s equal to 5596 FIG~ 10 inaicates that 9296 of SILs have a posterior probability greater than 0.5, and 769~ of normal squamous tissues have a posterior proba~ility less than o . 5 .
A prospective estimate of the method~ s performance was obtained using cross-validation. Table 4 (a) and (b) compares (a) the retrospective performance of the method on the data set used to optimize it to (b) the 30 prospectiYe estimate of the method' s performance using cross-validation Table 4 (a) indicates that for a cost of misclassification of SILs equal to 55~6, 929~ of high grade SILs, 9096 of low grade SILs, and 76% of normal squamous samples are correctly classif ied. The unbiased 35 estimate of the method's performance (Table 4 (b) ) indicates that there is no change in the percentage of correctly classified high grade SILs ~r nprmal squamous _ _ _ _ _ _ ~ _ _ _ _ . . . ... . .

W096l28084 P~ v,E' '~'11 32- 2f90374 tissue; there is a 59~i decrease in the proportion of corre tly classifiea low grade=SILs.
Table 4. (a) A retro~pective and (b) prospective estimate of the multivariate ~tati~tical method' ~ performance u~ing mean-~caled nn~-l; 7ed ~pectra at 460 nm excitation to differentiate SILs from normal ~quamou~
tissue~
(a) r~ 3;fi-~tion Normal Low Grade ~igh Grade Squamous SIL SIL
Normal S q~ 7 6 ~s 7 9~ 9 ~6 15SIL 249s 939~ 9l96 (b) Cla~aification Normal Low Grade lligh Grade SquamouE SIL SIL
20Normal Squamous 75g~ 14% 9~
SIL 259~ 86% 9l96 4) Low Grade SIL~ v~. IIigh Grade SIL~ at 460 nm 25 Excitation Principal components obtained from the preprocessed data matrix containing normalized spectra at 46Q nm excitation could be used to differentiate high grade gILs 30 from low grade SILs. These prrncipal components are included in Appen~ix II. Principal cnn~pnn~nt 4 (PC4) and principal component 7 (PC7) demonstrated the = ~
_ W0 96/28084 r~
;~ 33 _ 21 ~037 statistically most significant differences (p c 0.05) between high grade SILs and low grade SILs. The p values of the L` in;ng principal component scores were not statistically significant (p ~ 0 . 09) . Therefore, the 5 rest of the analysis was performed using these two principal components which account collectively for 8~6 of the variation in the original data set.
FIG. 11~ and FIG . - llB illustrate the measured 10 probability distribution and the best f it of the normal probability density function of PC4 and ~C7 for normal squamous tissues and SILs, respectively. There is reasonable agreement between the measured and calculated probability distribution, for each case. The prior 15 probability was determined to be: 39~ low grade SILs and 61~ high grade SILs. ~ Posterior prol~abilities of belonging to each tissue type were salculated FIG. 12 illustrates the retrospective performance of the diagnostic method on the same data .,et used to optimize 20 it. The posterior probability of being classified into the SIL category is plotted for all samples evaluated.
The results shown are for a cost of~misclassification of SILs equal to 65'6. FIG. 12 indicates; that 82g6 of high grade SILs have a posterior probability greater than o . 5, 25 and 789; of low grade SILs have a posterior probability less than 0 . 5 .
A prospective estimate of the method~s performance was obtained using cross-validation. Table 5 (a) and (b) 30 compares (a) the retrospective performance of the method on the data set used~ to optimize it to (b) the unbiased estimate of the method' s performance using cross-validation. Table 5 Sa) indicates that for a cost o~
misclassification of 659~ 8296 of high grade SILs and 7896 35 of low grade SILs are correctly classified. The unbiased estimate of the method's performance ~Table 5 (b) ) indicates that there i~ a 5~ decrease in the percentage Wo 96l28084 PcrluS96/02644 ~r~ f ~ 34 _ ~ ~ 90374 of correctly classified high grade SI:Ls and low~ grade ~ ~ "
SILs.
Table 5. (a) A reL..,J~e_l.ive and (b) prospective estimate of the multivariate statistical method' ~ peL r~ e using mean-scaled nnrr~ ed spectra at 460 ~m excitation to differentiate high grade from low grade SILs .
(a) if1~tion how Grade SIL High Grade SIL
Low Grade S IL 7 9 % l 8 ~
l 5 ~igh Grade S IL 2 l % 8 2 96 (b) t~lA~E~ifinAtion ~ow Grade SIL }Iigh Grade SIL
ZOLow Grade SIL 72g6 27%
Eigh Grade S IL 2 l 96 7 7 %
FIG. 3}~, FIG.- 3B and FIG. ~C are flowcharts of the above-described ~luorescence sp-ec~roscopy aiagnostic methods. In practice, the flowcharts .of FIG. ~A, FIG. 3 and FIG. 3~ are eoded into appropriaté ~orm~ ana are loaded into the program memory of computer ll~ [FIG. l) 3 0 which then controls the apparatus of FIG . l to ca~se the performance of the diagnostic method of the present invention .

WO 96/28084 P~lll : A'"1 1 ! t~ 3 5 - 2 1 9 ~ 3 7 4 Referring first to. FIG. 3A, control begin in block 300 where fluorescence spectra are obtained from the patient at 337, 380 and 460 nm excitation. Control then passes to block 3 01 where the probability of the tissue 5 sample under co~sideration being SIL is calculated from the spectra obtained from the patient at 337 or 460 nm.
This method is shown in more detail with reference to FIG. 3B.
Control then passes to decision block 302 where the probability of SIL calculated in block 301 is compared against a threshold of 0 . 5 . If the probability is not greater than 0.5, control passes to block 303 where the tissue sample is diagnosed normal, and the routine is 15 ended. On the other hand, if the probability calculated in block 301 is greater than 0.5, control passes to block 304 where the probability of the tLssue -nnt; ;ning SIL is calculated basea upon the emission spectra obtained from excitation at 380 nm. This method is identical to the 20 method used to. calculate probability of SIL from fluorescence spectra due to 33'7 or 460 hm, and is also presented below in more detail with reference to FIG. 3B.
Control then passes to decisio~ block 306 where the 25 probability of SIL calculated in block 304 is compared against a threshold of 0.5. If the probability calculated in block 3 04 is not greater than 0 . 5, control passes to block 3 07 where normal tissue is diagnosed and the routine is ended. Otherwise, if decision block 306 30 determines that the probability calculated in block 304 is greater than 0.5, control passes to block 308 where the probability o:E high grade SIL is~c~lculated from the fluorescence emission spectra obtained from a 460 nm excitation. This method is discusscd below in greater 35 detail with re~erence to FIG. 3C. ~

Wo 96/28084 PCrrUS96102644 9 o3~ 4 ContrQl then pas~es to decision block 309 where the probability of high grade SIL calculated in block 308 is compared with a threshold of 0.5. Ii- the probability calculated in block 308 is not greater than 0.5, low 5 grade SIL is diagnosed: ~block 311), otherwise high grade SIL is diagnosed (block 312).
Referring now to FIG. 3B, the conditioning of the~
fluorescence spectra by blocks 301 and 304 is presented 10 in more detail. It should be noted that while the processing of block 301 and 304 is identical, block 301 operates on spectra obtained frDm a 337 or g60 nm excitation, whereas block 304 operates on spectra obtain from a 380 nm excitation. In either case, control begins 15 in block 315 where the fluorescence spectra data matrix, D, i5 constructed, each row of which corresponds: to a sample fluorescence spectrum taken from the patient. ~
Control then passes to block 3L6 where the mean intensity at each emission wavelength of the detected fluorescence 20 spectra is calculated. : Then, in block 317, each spectrum of the data matri~ is normalized relative to a maximum of each spectrum. Then, in block 318, each spectrum of the data matrix is mean scaled relative the mean calculated in block 316. The output of block 318 is a preprocesse~
25 data matrix, comprisi~g preprocessed spectra= for~ the patient under examination.
Control then passes to block 319 where principal component analysis is conducted, as discussed above, with 30 reference to equations 2, 3, 4 and 5.: During principai component analysis, the coYariance matrix Z (equation =
(2) ), is calculated using a ~L~ L~ SSed data matrix, -the rows of which comprise normalized, mean scaled spectra obtained from all patients, including the patient 35 presently~ under consideration. ~ The result of block 319 is applied to block 3Z1 where a two-sided Student's T-test is conducted, which results in selection of only wo s6nsos4 PCrmss6/02644 _ 37 - ~ }374 diagnostic principal components. Control then passes to block 322 where logistic discrimination is conducted, which was discussed above with reference to equations 6 and 7.

The quantity calculated by block 322 is the posterior:probability of the sample belonging to the SIL
category (block 323).
Referring now to FIG. 3C, presented are the details of the determination of the probability of high grade SIL
from excitation at 460 nm (block 308, FIG. 3A) . Control begins in blo,ck 324 where the fluorescence spectra data matrix, D, is constructed, each row of which corresponds 15 to a sample fluorescence spectrum taken from the patient.
Control then passe6 to block 326 where each spectrum of the data matrix is normalized relative to a maximum of each spectrum. The output of block 326 is a preprocessed data matrix, comprising preprocessea spectra for the 20 patient under examination. It should be noted that, in contrast to the preprocessing performed in the SIL
probability calculating routine of FIG_ 3B, there is no mean scaling performed when calculating the probability of high grade SIL. ~
Control then passes to block 327 `where principal component analysis is conducted, as discussed above, with ref erence to equations 2, 3, 4 and 5 . ~uring principal component analysi:s, the covariance ~matrix Z (equation 30 (2) ), i5 calculated using a preprocessed data matrix, the rows of which comprise noemalized, mean scaled spectra obtained from all patients, including the patient presently under consideration. The result of block 327 is applied to block 328 where a two-sided Student's T-35 test is conducted, which results in selection of onlydiagnostic principal components. Control then passes to block 32~ where logistic discrimination is conducted, _ _ _ _ _ .. _ . . .. . .. _ Wo 96/28084 PCT/US96102644 y(~ ?~93374 ~ ~ - 38 -which was discussed above with,~eference to equations 6 and 7.
The quantity calculated by block 329 is the 5 posterior~ probability of the sample belo~ging to the hïgh grade SIL category ~block 331~
2. Raman SPectro~coPY Diaano~tic Method ~ _ To lllustrate the ef ficacy of the present inventio~, twenty colposcopically normal and twenty colposcopical~y abnormal samples were studied.; Two sample pairs were -:
discarded due to experimental errors. Histologically, there were 19 normal, 2 metaplasia, 4 inflammation, 2 HPV
1~ and 9 dysplasia sam~les (2 mild and ~ mo~3erate to severe dysplasias). For the purposes=of this study the samples are classified as follows: normal, metaplasia, inflammation, low grade SIL a~d high grade~~IL. Two types of ~l;ff~r~n~;~tion are of interest clirlically~
20 SILs from alI other tissues and ~2) high grade SILs from low grade SIhs. The diagnostic methods developed using fluorescence and Raman spectroscopy are targeted towaras achieving optimal sensitivity in this differentiation.
Near infrared spectra of cervical tissues ~ere =~
obtained using the system shown in F=IG 2 These spectra are distorted by ~oise and autofluorescence ana are ~
preferably processed to yield the tissue vlbrational =
spectrum. Rhodamine 6~ powder packed in a quart~ cuvette was used ~fc;I calib~ation purposes since it has well ~=
documented Raman and fluorescent properties A rhodamine spectrum with high sïgnal to noise ratio was obtained using a 20 second integration time and a rhodamine spectrum with low signal to noise ratio was obtained 3~ using 1 second integration FIG. 13: is a gra-ph showing measured_rhodamine spectra with high and low S/~ ratio~s respectively .

Wo 96~28084 P S96/02644 2 1 9 0 3 7 ~

The ob6erved noise in the spectra was established to be approximately gaussian. This implies that the use of simple filtering techniques would be effective in smoothing the curves. Using a moving average window on 5 median filter yields acceptable regults. However, optimal results are obtained when the spectrum is convolved with a gaussian whose full width half maximum equals the resolution of the system. This technique discards any signal with bandwidth less than the 10 resolution of the system. The filtered scattering signal is still distorted by the residual fluorescence. A
simple but accurate method to eliminate this fluorescence is to f it the spectrum containing both Raman and fluorescence information to a polynomial of high enough 15 order to capture the f luorescence line shape but not the higher frequency Raman signal. A 5th degree polynomial was used. The polynomial was then subtracted from the spectrum to yield the Raman signal alone. In FIG. 14, the efficiency of eliminating fluorescence from a Raman 20 signal using a polynomial fit is illustrated.
FIG. 15 is a graph showing that the processed low S/N rhodamine spectrum is similar to the high S/N
rhodamine spectrum and is not distorted by the filtering 25 process.= ~Referring to FIG. 15, in comparison, it can be ~een at that the initially noisy spectrum of the low S/N
rhodamine once processed show the same principle and secondary peaks at the spectrum of high S/N rhodamine.
This validates the signal processing techni~ues used and 30 indicates that the technique does not distort the resultant spectrum. Each tissue spectrum was thus processed. Peak intensities of relevant bands from these spect~:a were ~measured and used for ~iagnosis. 'I :
Typical processed spectra for :~pair of ~or~Lal and abnormal samples from the same patient are shown in FIG. 16 which is a graph of a typical pair of processed Wo 96l28084 PCr/Uss6/oz644 i f' ~ C 2 1 9 0 3 7 4 spectra from a patient with dysplasia showing the dif f erent peaks observed . Several peaks are observed at 626, 818, 978, 1070, 1175, 1246, 1325, 1454==and '1656 cm~
(~ 11 cm~l) . Several of the peaks observed have been cited in studies o~ gynecologic tissues by other groups such as Lui et al., "Fluorescence and Time-Resolved Light Scattering as Optical Diagnostic Tecb~iques to Separate Diseased and Normal Biomedical Media", J Photochem Photobiol B: Biol, 16, 187-209, 1992, on colon tissues, and in IR absorption studies on cervical cells by Wong et al., "Infrared Spectroscopy of Human Cervical Cells:
Evidence of Extensive Structural Changes during ~ - =
Carcinogenesis", Proc Natl ~cad Sci USA, 88, 10988-10992, 1991. FIG. 17 is a graph of the intensity ~of the band at 1325 cm~l for all biopsies to illustrate the patient to patient variation in :the intensities of the Raman bands.
The intensity of the various Raman bands show a significant patient to patient variability. In FIG. 17, the samples are plotted as pairs from each patient. To account for this patient to patient variability, each peak in a spectrum was normalized to the corresponding peak of the colposcopic and histologic normal sample from the same patient. Thus all colposcopic normal samples~
that are histologically normal have a peak intensity of one. Normalized and unnormalized spectra were analyzed for diagnostic information.
Each of the bands observed colltains some diagnostic information and can differentiate between tissue types with varying accuracy. Clinically, the separati=on of ~ ~
SILs from all otE~er tlssues. and high grade SILs from low grade SILs ¦s of i=rit~erest. Because of the patient to patient variability more signif~lcant differelltiation was obtained using paired analysis._ The bands at 626, lQ70 and 1656 cm~1 can each differentiate SILs from all other -issues. At all three bands, the intensity of the normal WO 96/28084 ~ " t "`?' I ~

is greater than the intensity of the SIL This is illustrated in FIG. 18 and FIG. 19.
FIG. 18A and FIG. 18B are yraphs showing diagnostic capability of normalized peak intensity of Raman bands a~ 626 cm~l, and 1070 cm~l, respectively. The band at 626 cm~l which is due to ring deformations differ,~nt;~
SILs from all other tis6ues with a sensitivity and specificity of 91~ and 92~ (FIG. 18~) . One SIL sample ~focal ~PV) is misclassified. The metaplasia samples are incorrectly classified as SILs at this band. However, using the intensity at the C-O stretchlng and bending vibrational band of about lQ70 cm~l for a similar claseification, all metaplasia and inflammation samples are correctly classified as non-SIBs (FIG. 18B) . Of the two samples incorrectly classif ied, it was determined that one has focal dysplasia and the other is the same sample with focal HPV that was misclassified at 626 cm~l.
Only one normal sample is misclassified as SIB. A
sensit:ivity and speciiicity of 829~ and 9696 is achieved.
This band has been attributed to glycogen and cellular lipids/phosphate by Wong et al., "Infrared Spectroscopy of Human Cervical Cells: Evidence of Extensive Structural Changes during Carcinogenesis, " Proc Natl Acac~
sci USA, 88, 10988-10992, 1991. --FIG. 19 is a graph showing the di~gnostic capability of the band at 1656 cm~1. Decision line (1) separates SILs from all other tissues Decision line (2) séparates 30 high grade from~low grade SILs. The normalized peak intensity at 1656 cm~l can differentiate SILs from other tissues using line (1) as the decisi-on line with a sensitivity and specificity of 9196 and 889~. The focal dysplasia sample incorrectly classif~ied at lQ70 cm~l is 35 again misclassified. The metaplastic samples are again classif ied as SILs . The advantage of using this peak is that it can also differentiate between high grade and low _ _ .. _ _ _ _ _ .... . .

Wo 96/28084 PCrNS96/02644 j C ~ 42 _ - 2 ~ 9 037 4 ~
grade SILs. :~sing line (2~ as a decision line, this peak can separate high and low grade: SILs with a sensitivity of 86%. The metaplasia samples misclassified as SILs are separated fr4m the high grade samples. Only one~normal sample is misclassified. In the cervix, this peak has been associated with cellular proteins from the nuclei of the epithelial cells. These proteins have been suggested to be primarily collagen and elastin by Lui et al., "Fluorescence and Time-Resolved Light Scattering as Optical Diagnostic Technigues to Separate Diseased and Normal Biomedical Media, " J Photochem Photobiol B: Biol, . 16, la7-209, 1992.
of the other features Qbserved in the Raman spectra of cervical tissues, the band at 818 cm-l is associated with ring ' breathing' and is at=tributed to blood. The intensity of this band is greater in dysplasia samples relative to respect to normal samples. The peak at 978 cm~l is assQciated with phosphorylated proteins and nucleic acids. This band differentiates SILs from other tissues with a sensitivity and specificity of 82% and 80%. The band at 1175 cm~l can separate normal from dysplasia samples with a sensitivity of 88% The decrease ~in intensity of this band with dysplasia has been reported by Wong et al., "Infrared Spectroscopy of=~
Human Cervical Cells: Evidence of~Extensive StructuraI
Changes during Carcinogenesis", Proc Natl Acad gci USA, 88, 10988-lQ992, 1991, as well_ This band around 1175=
cm~1 has been associated with C-O-stretching and in cervical cells, this band consists of three overlapping lines at 1153, 1161 and 1172 cm~l. A similar trend is~
also observed~ in cervical tissue samples. These bands have been attributed to C-O s~:~etching of cell proteins such as tyrosine and carbohydrates. The Raman line at 1246 cm~1 is assigned to the strething vibrations of C-N
(amide III) . The line at 1325 cm~l is due to ring vibrations and is assQciated with tryptophan by Lui et WO Y6/28084 ~ ~; PCINS96102644 r ~ 2 t 9 0 3 7 4 al., "Fluorescenee and Time-Resolved Light Scattering as Optical Diagnostic Technigues to Separate Diseased and Normal E~iomedical Media", J Photochem Photobiol B: Biol, 16, 187-209, 1992, and nucleic ~cids. ll,n increase in the 5 intensity of this peak in the SILs with respect to the other tlssues is observed. This has been associated with increased cellular nuclear cantent in the colon. The lines at 1401 and 1454 cm~1 are due to symmetric and asymmetric CH3 bending modes of proteins (methyl group).
The line at 1454 cm~1 differentiates high grade from low grade - SILs with a 91~ accuracy. These lines have been associated with elastin and collagen by Lui et al., ~ Fluorescence and Time-Resolved Light Scattering as Optical Diagnostic Techniques to Separate Diseased and Normal Biomedical Media", J Photochem Photobiol B: Biol, 16, 187-209, 1992.
Analyzing the diagnostic information from the tissue Raman spectra in a paired manner, SILs may be differentiated from all other tissues at several peaks with an average sensitivity of 88~ 696j and a specificity 92~ (i 796) . The best sensitivity is achieved at 91~6 with the bands at 626`and 1656 cm~l. The best specificity is achieved at 100~6 using a combination of the bands at 1070 and 626 cm~l. In differentiating SII,s from normals, the sensitivity and specificity of the Raman methods are greater than those of the fluorescence based methods for the 36 samples but ~are similar when compared to the fluorescence res~ilts from the larger sample study. Inflammation and metaplasia samples can be separated from the SILs using the Raman band at 1070 cm~
and at 1656 cm~l. Raman spectra are ~successfl~l in differentiating high grade SILs from low grade SILs with an average sensitivity of 86g6 (~:4~) . The sensitivity is improved when compared to fluorescence based diagnosis of the same 36 samples as well as the larger sample population. The invention also accommodates patient to . _ _ _ ~ , . . , .. . , , _ _ . .. . _ _ _ 4 4 _ 2 t ~ ~ 3 j ~11~ ? 1~
patient variability in the intensities of the Raman lines by use of paired analysis as pr.esented above. In addition, unpaired f~;ff,-r~ntiation may be done by using the peaks at 1325, 145~ and 1656 cm~1 with a comparable 5 sensitivity.
For unpaired differentiation, the ratio of ~:
intensities at 1656 and 1325 cm~l differentiate ~SILs from all other tissues wi~h a sensitivity and a specificity of 8296 and 80%, respectively (FIG. 20) . In addition, the ratio of the intensities at 1656 and 1454 cm~1 may be used in an unpaired manner to differentiate high grade SILs from low grade SlLs with a sensitivity and specificity of: 1009; ana 10096 (FIG. 21) .
Further, as mentioned above, each of these:~specifled peaks of the Raman spectrum contain some diagnostic information for tissue: dif Eerentla~ion.~~~ ~lultivariate techniques using principal component analysis and Baye's 20 theorem, similar to the conditioning of the fluorescence spectra described above, would use information from all of the peaks of ~the Raman spectrum, and ~ould thus improve the diagnostic performance cf the Raman signals The methods usinq Raman signals presented here have been 25 optimized for the 36 ~sample: data se3~ ana ar~e thus a bias estimate of =their performance.- A true estimate of the~
diagnostic f'~rAh; l; ty of Raman spectroscopy would require an unbiased assessment of the performance of the method which for the small number of samples could be Qbtained 30 using cross validation techniques, or Qther types of ~ =
validation technique:ii. ~ =
The present invention exploits several potential advantages of Raman spectroscopy over fluorescence. The 35 Raman diagnostic methods used in the lnvention reiter~[te the simplicity of Raman spectroscopy for diagnosis and W0 961280W , ~ r ~ ? ' ~ ' indicate the potential of improved diagnostic capability using this technique.
FIG. 4A, FIG. 4B, FIG. 4C and FIG 4D are flowcharts 5 of the above-described Raman spectroscopy diagnostic method. In practice, the flowcharts of FIG. 4A, FIG. 4B, FIG. 4C and FIG. 4D are coded into approp~riate form and are loaded into the program memory of computer 211 ~FIG. 2) which then controls the apparatus of FI~. 2 to 10 cause the performance of the Raman spectroscopy diagnostic method of the present invention.
Referrillg first to FIG 4A, after the method is started, the NIR Raman spectrum is acquired from the 15 cervical tissue sample of unknown diagnosis in step 4 0 0 .
Then, in step 401, the acquired spectrum is corrected as a function of the rhodamine calibration process. Then, in block 402, the spectrum is convolved with a gaussian G
having a full width half maximum of 11 wavenumbers, thus 20 providing a corrected noise spectrum R. In step 403, the broad band baseline of the noise corrected spectrum is f it to a polynomial L, and the polynomial is subtracted from the spectrum to give the Raman signal for the ~ample under considera~ion.

Control then passes to step 404 where the maximum intensities at 626, 818, 978, 1070, 1175, 1246, 1325, 1454 and 1656 wavenumbers (in units of cm~l) are noted.
Alsc in block 4D4, maximum intensities at five selected 30 wavenumbers are stored. These include:
P1 = i~tensity at 626 cm~
P2 = intensity at 1070 c~n~
P3 = intensity at 1325 cm~
3~ ==~4 = intensity at 1454 cm~
P~ = intensity at 1656 cm~

W096/28084 P~l/u.,,:'A~'11 46_ 2~a37~ -Control then passes to bloclc 405 where the stored intensities are analyzed in order to diagnose the tissue sample. This analysis is presented below in more detail with referencé to FIG. 4B, FIG. 4C and ~IG. 4D.-Referring to FIG. 4Br decision bloclc 406 determineswhether paired analysis is desired, and if so control ~
passes to block 407 where the paired diagnostic method is conducted. This is presented below in more detail with 10 reference to FIG. 4C ~ -Control then passes to deci~ion block 408 where itdetermined whether unpaired analysis i~ desired~ If so, control passes to block 409 where the unpaired diagnostic 15 method is conducted.
Referring to the paired d~agno6tic ~ethod, ~presented with ref erence to FIG .~ 4C, three parallel analyses may be conducted, one with re~pect to intensity P1, one with 20 respect to intensity P2, and one with respect to intensity P5. For intensity P1, control begins in block 411 where quantity N1 is set equal to the intensity at the selected wavenumber for a normal tissue sample of the patient under consideration. ~ontrol then passes to 25 block 412 where the ratio between measured intensity P1 and normal intensity N1 is calculated. In block 413, the ratio is compared with a threshold of 1. If the ratio ~is greater than or equal to 1, the diagnosis is non-SIL --~~step 414), whereas if the ratiQ is less than 1, the 30 diagnosis is SI~ (step 416). ~
A similar analysis is conducted with respect to intensities P2 and P5. Specifically, for intensity P2, control begins in block 417 where quantity N2 is set =
35 equal to the intensity at the sel~cted wavenumber for a normal tissue sample of the patient under consideration.
Control then passes to block 41~ where :the ratio between Wo 96/28084 PCrlUS96/02644 ~ 4 7 2 i 9 0 3 7 ~
measured intensity P2 and normal intensity N2 is calculated. In block 419, the ratio is compared with a threshold of 1. If the ratio i8 greater than or equal to l, the diagnosis is~non-SIL Sstep 421), whereas if the ratio is less than 1, the diagnosis is SIL ~step 422~.
For intensity P5, control begins in block 423 where quantity Ns is set equal to the lntensity at the selected wavenumber for~a normal tissue sample o the patient under consideration. Control then passes to b1ock 424 wher=e the ratio between measured intensity P5 and normal intensity N5 is calculated. In block 426, the ratio is compared with a threshold of 1. If the ratio is greater than or equal to 1, the diagnosis is non-SIL (step 427), whereas if the ratio is less than 1, the diagnosis is SIL
(step 428).
If SIL is concluded in step 428, control passes to decision block 429 where the ratio calculated in block 424 is compared against a threshold of 0 . 75 . If the ratio is greater than or equal to 0.75, then low grade SIL i~ diagnosed (step 431), whereas if- the ratio is less than 0.75, high grade SIL is diagnosed (step 432).
Unpaired analysis of the NIR Raman spectrum is presented in FIG. 4D Beginning in~ step 432, ratio r1 is calculated between intensity P5 and intensity P3, and ratio r2 is calculated between intensity P5 and intensity P4. Control then passes to decision block 434 where ratio ri i5 c~mpared against a threshold of 1.8. If ratio r1 is gre-ater than or equal to 1. 8, the tissue sample is diagnosed as non-SIL (s~e--p ~3~), whereas if ratio ri is less than 1. 8, the tissue is diagnosed as SIL
(step 4371. Control then passes to decision block 438 where ratio r2 i5 compared against the threshold of 2. 6 .
If ratio r2 is greater than or equal to 2 . 6, low grade WO 96l28084 P~~ .'^?~11 '~"~ 48 ~ 2 1 q (~374 SIL is diagnosed ~step 439), whereas ratio r2 i8 less --than 2.6, high grade SIL is=diagnosed (step 441).
It 6hould be noted that the various thresholds used 5 for the decisiQn blocks in FIG. 4C and FIG. 4D may be adjusted without departing from the scQpe of the invention. The thresholds presented were chosen as a -function of the training data, and other or more complete training data may result in different threshQlds.

3. t~orbiIIed Fluoresçence Anrl pAr`~n S~eÇI.L~ÇY~Y Method The present invention also contemplate6 a system that sequentially acquires fluorescence and NIR Raman 15 spectra in vivo through an Pptical probe, such as a fiber optic probe or other optical coupling system. The optical probe is selectively coupled to ultraviolet or visible sources of electromagnetic radiation to excite f luorescence, and then selectively coupled to NIR sources 20 to excite fluorescence free- Raman spectra;: The fluorescence spectra may be used to improve the analytical rejection of fluorescence from the Raman spectrum .
2~ The apparatus used for this purpose is a combination of the apparatus disclosed in FIG 1 and FIG. 2. A ~
dichroic mirror or swing-away mirror is ~se=d so that each electromagnetic radiation source is selectively~~coupled sequentially into the optical probe. Similarly, light 30 collected by the probe is $electively coupled to the appropriate detectors to sense=the fluorescence- spectra and Raman spectra.
In analyzing the spectra -for dia~nostic purposes,: it 3~ is presently contemplated that the above-described ability of ~luorescence to identify normal tissue, and low and high grade lesions, be:fs)llowed by the above-wos6~sos4 r~.",~ 6~
; ~. f ~ ~ ~ . 2 1 ~ 0 3 7 4 described use of NIR Raman spectra to identifyinflammation and metaplasia. Alternatively, information gathered about the tissue type, in accordance with the above-described fluorescence diagnosls, is used to 5 improve the Raman diagnostic capability. This is accomplished by using f luorescence spectra to calculate the posterior probability that a tissue is normal, low or high grade SI~. Then, this classification is used as the prior probability in a Bayesian method, based on the lO detected Raman spectra. In yet another embodiment, information gathered with NIR Raman spectroscopy is used to calculate the posterior probability that the tissue is inflamed or metaplastic. Then, this information is used as the prior probability in a Bayesian method, based on 15 the detected f luorescence spectrum .
While the present invention has been described with reference to several exemplary embodiments, it will be understood that modifications, additions and deletions 20 may be made to these embodiments without departing from the spirit and scope o~ the present invention.

.

2 1 ~ ~ 3 7 4 APPENDIX I: SPECIFICAI-LY AND :i~;N~il'l'l\/l'~
Summarized from. Albert A., Harris E.K.: Multivariate rn t erpre ta ti on of C1 ini cal La~oora tory Da ta, Marce 1 D ekker Inc., New York, pp. 75-82, rl387), the disclosure of which is expre~3sly incorporated herein by refererce. -Assuming a group of T samples which can be categorized as no~mal ~N samples) or' diseased (D
10 samples~. A diagnostic test, designed to determine whether the sample is normal or diseased, is'applied to each sample. ~ The results Qf the tests is the continuous variable x, which is then used to determine the sample type. FI~. 22 illustrates a hypothetical distribution of 15 test values for each sample type. A diaynostic method based on this test can easily be def ined by choosing a cutoff point, d, such that a sample with an observed value x<d is diagnosed as ~ormal and a sample with an observed value x2d is diagnosed as abnormal.
2~ : ~
Several quantitative measures have been defined to `evaluate' the performance of thi~ type of method. The f irst type evaluates the test itself ~ i . e . measures the ability of the test to separate the two populations, N
25 and D) . Sensitivity and specificity are two such measures. The second type is designed to a~id in the interpretation of a particular~test result li.e. deciding whether the individual test measurement has come from a-normal or diseased sample). Positive and negative 3~:) predictive value are two measures of this type.
To def ine these measures, ~some ~erminology and notation must be introduced. Referri~g to Table 6, a sample to be tested can be either normal or diseased; the 35 result of the test for each type of sample can be either negative or positive. True negative~ represent those ~
normal with a positive test result In these ca~es, the WO 961280R4 . PC~/US96~02644 51 - 2 ~ 9 ~) 3 7 4 diagnosis based on the rest re6ult is correct. False positives are those normal samples which have a positive test result and false negative6 are those diseased samples which have a negative test result. In these 5 cases, the diagnosis based on the test result is incorrect .
TA~3LE 6 ¦ Normal Diseased Total Samples Test Ncgative True Negatives False Negatives Negatives Ix < d) ITN~ ~FN) (Neg) Test Positive False Positives True Positives Positives (x 2 d~ (FP) (TP) (Pos) 15Total Samples N D T
With this terminology, Table 7 contains a definition of sensitivity and specificity, the two measures which 20 assess the performance o~ the diagnostic method.
Specificity is the proportion of ~ormal samples with a negative test result (proportion of normal samples diagnosed cDrrectly). Sensitivity is the proportion of diseased samples with a positive test result (Proportion 25 of diseased samples correctly diagnosed). FIG. 22 also contains a graphical representation of specif icity and sensitivity. Specificity represents the area under the normal sample distribution curve to the le~t of the cut off point while sensitivity represent the area under the 30 diseased sample distribution curve to the right oi the cut of f poin j .

WO 96n8084 PCIIUS96/02644 ~ r''`~ 52- 2~374 TAB~E 7 Test Measure Meaning Calculation Sper;f;~ity Proportion of normal Sp=TN/N
samples with negative test result 5Sensitivity Proportion of diseased Se=TP/D
samples with positive te6t result While sensitivity and specificity charactelize the performance of a particular method, another set of 10 statistics is required to interpret the laboratory test result for a given specimen. The positive and negative predictive value quantify the meaning of an individual~
test result (Table 8) . The positive predictive value is the probability that if the test result is positive, the 15 sample is diseased. The negative predictive value is the probability that if the test result is negative, the ~
sample is normal. Positive and negative predictive value are calculated from Baye's rule as outlined in Albert and ~arris. Table 8 cr~nt~;nC t~o equivalent ~ormulas ~or =
20 calculation positive and negative predictive value.
T~BLE: 8 Measure Meaning C~lGulation 1 Calculation 2 25 Positive The probability that, PV+-TPlPos PV+-DSel(DSe+N[l Sp)l PredictiYe if the test is Value positive, the sample is diseased Negative The probability that, PV -TNlNeg PV -NSpllNSp+D(l-Se)) Predictive if the test is 30 Value negative, the sample is normal W0 96/28084 ;, ' ~ PCINS96/~2644 APPENDIX II: PRINCIPAI. - - iJ~
337 nm excitation 460 nm excitation 380 nm excitation 460 nm excitation 5 El E2 El E2 E2 E5 E4 E7 0.12 0.11 0.147 -0.275 -0.615 0.532 0.69 0.10 0.17 0.12 -0.093 -0.319 0.464 0.151 0.09 0.07 0.22 0.12 0.074 0.360 0.378 0.1 -0.14 -0.17 0.25 0.11 0.056 0.345 0.317 0.308 0.23 0.07 0.27 0.1 0.027 0.314 -0.236 0.373 0.24 0.06 0.28 0.11 -0.004 -0.253 0.157 0.348 0.23 0.04 0.28 0.12 0.010 0.193 0.086 -0.236 -0.19 0.01 0.28 0.12 0.024 0.121 -0.04 0.161 0.15 0.00 0.28 0.1 1 0.029 0.048 -0.004 0.071 0.09 0.05 15 0.26 0.11 0.016 0.030 0.025 0.055 0.01 0.07 0.24 0.1 1 0.001 0.097 0.044 0.013 0.06 -0.07 0.22 0.11 -0.026 0.153 0.06 0.068 0.12 0.24 0.2 0.09 0.052 0.201 0.06 0.108 0.14 0.40 0.17 0.08 0.025 0.203 0.055 0.123 0.16 0.30 20 0.13 0.05 0.019 0.192 0.046 0.159 0.16 0.04 0.09 0.04 0.062 0.160 0.023 0.133 0.16 0.12 0.06 0.04 0.090 0.153 0.006 0.15 0.14 -0.18 0.02 0.05 0.091 0.153 0.014 0.089 0.14 0.14 0.01 0.05 0.088 0.164 0.026 0.075 0.16 0.24 25 -0.04 0.05 0.087 0.158 0.044 0.047 0.17 0.23 0.06 0.05 0.106 0.146 0.055 0.025 0.17 0.16 -0.08 0.07 0.145 0.092 0.063 0.018 0.11 0.12 0.09 0.09 0.189 0.020 0.071 -0.089 0.05 0.18 -0.1 0.11 0.218 -0.023 0.072 0.102 0.01 0.09 .. . . . . . . .. .. . .. . . .. . . . .. . . . . . ...

WO 96n8084 ~ 2 1 9 0 3 ~C$/U596/Q2644 APPENDIX II: PRINCIPAL ~ ~)N~ D (c~:)r~n~lecl) I
337 nm excitation 460 nm excitation 380 nm excitation 460 nm excitation ElE2 El E2 E2 E5 E4 E7 -0.11 0.13 0.240 0.054 0.078 0.104 0.02 -0.11 -0.1 1 0.1 5 0.249 0.060 -0.071 0.078 0.04 0.04 -0.12 0.17 0.242 -0.073 -0.071 0.091 0.03 0.06 0.12 0.18 0.238 0.075 -0.066 -0.087 0.02 0.08 5 0.12 0.2 0.240 0.064 0.062 -0.095 -0.03 0.15 0.11 0.2 0.230 0.063 -0.06 -0.08 -0.03 0.18 -0.1 0.21 0.221 0.061 0.057 0.067 -0.03 0.19 -0.09 0.22 0.21 1 -0.060 -0.048 0.086 -0.02 0.25 -0.08 0.22 0.204 -0.052 0.039 0.068 0.01 0.26 0.07 0.21 0.199 0.045 0.031 -0.039 0.00 0.17 -0.07 0.21 0.185 0.044 0.027 0.034 0.01 0.10 -0.07 0.2 0.181 0.045 -0.019 -0.028 0.01 0.03 0.06 0.2 0.176 -0.042 -0.019 0.032 0.00 0.02 0.06 0.19 0.170 0.037 0.015 -0.01 0.00 -0.01 0.06 0.18 0.167 0.035 0.008 0.039 0.01 0.12 -0.05 0.17 0.159 -0.030 -0.008 -0.037 0.03 0.13 -0.05 0.16 0.158 0.032 -0.01 -0.068 0.01 -0.21 -0.05 0.15 0.151 -0.027 -0.009 -0.085 0.01 0.00 -0.05 0.14 0.146 -0.027 -0.005 -0.095 o oo 0.03 2()0.05 0.13 0.137 0.019 0.01 0.069 0.01 0.03 0.05 0.12 0.128 0.015 0.007 0.084 0.01 0.03 -0.05 0.1 1 0.012 0.034 -0.05 0.1 0.012 0.036 0.04 0.11 .. .. . . . . .

r . ;~ :. 2~0374 APPENDIX II: PRINCIPAI ~ (continued) 337 nm excitation 460 nm excitatiDn 380 nm excitation 460 nm excitation El E2 El E2 E2 E5 E4 E7 -0.04 0.09 0.04 0.09 0.03 0.09 0.03 0.09 5 -0.03 0.08 0.03 0.08 0.03 0.08 0.02 0.09 0.02 0.12

Claims (37)

CLAIMS:

1. A method of detecting and quantifying tissue abnormality in a tissue sample, comprising:
illuminating said tissue sample with a first electromagnetic radiation wavelength selected to cause said tissue sample to produce a fluorescence intensity spectrum indicative of tissue abnormality;
detecting a first fluorescence intensity spectrum emitted from said tissue sample as a result of illumination with said first wavelength;
illuminating said tissue sample with a second electromagnetic radiation wavelength selected to cause said tissue sample to produce a fluorescence intensity spectrum indicative of a degree of tissue abnormality;
detecting a second fluorescence intensity spectrum emitted from said tissue sample as a result of illumination with said second wavelength;
calculating from said first fluorescence intensity spectrum, a probability that said tissue sample is normal or abnormal; and calculating from said second fluorescence intensity spectrum a degree of abnormality of said tissue sample.

2. The method of claim 1, each of said calculating steps comprising:

conducting principal component analysis of said first and second spectra, relative to a plurality of preprocessed spectra obtained from tissue samples of known diagnosis.

3. The method of claim 1, each of said calculating steps comprising:
normalizing said first and second spectra, relative to a maximum intensity within said spectra.

4. The method of claim 3, each of said calculating steps further comprising:

mean-scaling said first and second spectra as a function of a mean intensity of said first and second spectra.

5. A method of detecting tissue abnormality in a diagnostic tissue sample, comprising:
illuminating said tissue sample with an illumination wavelength of electromagnetic radiation selected to cause said tissue sample to emit a Raman spectrum comprising a plurality of wavelengths shifted from said illumination wavelength;
detecting a plurality of peak intensities of said Raman spectrum at wavelength shifts selected for their ability to distinguish normal tissue from abnormal tissue;

comparing each of said plurality of detected peak intensities at said wavelength shifts with intensities of a Raman spectrum from known normal tissue at corresponding wavelength shifts;
detecting abnormality of said tissue sample, as a function of said comparison.

6. The method of claim 5, further comprising:
calculating a ratio between selected intensities of said Raman spectrum; and detecting abnormality of said tissue sample, as a function of said ratio

7. A method of detecting tissue abnormality in a diagnostic tissue sample, comprising:
illuminating said tissue sample with an illumination wavelength of electromagnetic radiation selected to cause said tissue sample to emit a Raman spectrum comprising a plurality of wavelengths shifted from said illumination wavelength;
detecting a plurality of peak intensities of said Raman spectrum at wavelength shifts selected by their ability to distinguish normal tissue from abnormal tissue;
calculating a ratio between at least two of said plurality of peak intensities; and detecting abnormality of said tissue sample, as a function of said ratio.

8. The method of claim 7, further comprising:
calculating a second ratio between two of said plurality of peak intensities; and detecting a degree of tissue abnormality as a function of said second ratio.

9. A method of detecting tissue abnormality in a diagnostic tissue sample, comprising:
illuminating said tissue sample with electromagnetic radiation having a plurality of wavelengths, a first subset of said plurality of wavelengths having been selected to cause tissue to emit fluorescence spectra indicative of tissue abnormality, and a second set of said plurality of wavelengths having been selected to cause tissue to emit Raman spectra indicative of tissue abnormality;
detecting a fluorescence intensity spectrum from the tissue sample;
detecting a Raman spectrum from said tissue sample;
and assessing abnormality of said tissue sample as a function of said detected fluorescence spectrum and as a function of said detected Raman spectrum.

10. An apparatus for detecting and quantifying tissue abnormality in a tissue sample, comprising:
a controllable illumination source for emitting a plurality of electromagnetic radiation wavelengths selected to cause said tissue sample to produce fluorescence intensity spectra indicative of tissue abnormality;
an optical system coupled to said illumination source for applying said plurality of radiation wavelengths to a tissue sample;
a fluorescence intensity spectrum detecting device for detecting an intensity of fluorescence spectra emitted by said sample as a result of illumination by said plurality of electromagnetic radiation wavelengths;
a data processor, connected to said detecting device, for analyzing detected fluorescence spectra to calculate a probability that said sample is abnormal.

11. A Raman spectroscopy apparatus for detecting tissue abnormality in a tissue sample, comprising:
a controllable illumination device for generating at least one illumination wavelength of electromagnetic radiation selected to cause a tissue sample to emit a Raman spectrum including a plurality of wavelengths shifted from said illumination wavelength;

a Raman spectrum detector for detecting a plurality of peak intensities of said Raman spectrum at selected wavelength shifts; and a programmed computer connected to said Raman spectrum detector, programmed to compare each of said plurality of detected peak intensities with corresponding peak intensities of a Raman spectrum from known normal tissue, to detect tissue abnormality.

- 62 -

12. An apparatus for detecting and classifying tissue abnormality at a tissue site, comprising:
a first source of electromagnetic radiation of a first wavelength that excites different first fluorescence intensity spectra in normal and abnormal tissue;
a first receiver sensitive to the first fluorescence intensity spectra;
means coupled to the first receiver for calculating from the first fluorescence intensity spectra a first probability that the tissue site is normal or abnormal;
a second source of electromagnetic radiation of a second wavelength that excites different second fluorescence intensity spectra in tissues having different types of abnormality a second receiver sensitive to the second fluorescence intensity spectra;
means coupled to the second receiver and to the first probability calculating means for calculating from the second fluorescence intensity spectra a second probability that the tissue site is of a particular type of abnormality, when the first probability is indicative of abnormality at the tissue site;
and a tissue site probe coupled to the first and second sources and to the first and second receivers.

13. An apparatus as in claim 12 wherein the first probability calculating means comprises:
means for conducting a first principle component analysis of the first fluorescence intensity spectra; and means coupled to the first principle component analysis means for conducting a logistic discriminant analysis to obtain the first probability;
and wherein the second probability calculating means comprises:
means for conducting a second principle component analysis of the second fluorescence intensity spectra; and means coupled to the second principle component analysis means for conducting a logistic discriminant analysis to obtain the second probability.

14. An apparatus as in claim 13 wherein:
the first probability calculating means further comprises means for normalizing the first fluorescence intensity spectra relative to respective maximum intensities thereof, prior to conducting the first principle component analysis in the first principle, component analysis means; and the second probability calculating means further comprises means for normalizing the second fluorescence intensity spectra relative to respective maximum intensities thereof, prior to conducting the second principle component analysis in the second principle component analysis means.

15. An apparatus as in claim 13 wherein:
the first probability calculating means further comprises means for mean-scaling the first fluorescence intensity spectra as a function of a mean intensity thereof, prior to conducting the first principle component analysis in the first principle component analysis means; and the second probability calculating means further comprises means for mean-scaling the second fluorescence intensity spectra as a function of a mean intensity thereof, prior to conducting the second principle component analysis in the second principle component analysis means.

16. An apparatus as in claim 12 wherein:
the first and second sources comprise a pulsed nitrogen pumped dye laser respectively operated to generate pulses at the first wavelength having a power level, pulse duration, and repetition rate that excite the first fluorescence intensity spectra in normal and abnormal tissue, and to generate pulses at the second wavelength having a power level, pulse duration, and repetition rate that excite the second fluorescence intensity spectra in tissues having different types of abnormality;

and the first and second receivers comprise a polychromator coupled to an intensified diode array controlled by a multi-channel analyzer, respectively synchronized to the first and second sources.

17. An apparatus for detecting and classifying tissue abnormality at a tissue site, comprising:
a controllable illumination source for emitting a plurality of electromagnetic radiation wavelengths selected to cause tissue to produce fluorescence intensity spectra indicative of tissue abnormality;
an optical system coupled to the illumination source for applying the plurality of electromagnetic radiation wavelengths to the tissue site;
a fluorescence intensity spectrum detecting device for detecting an intensity of fluorescence spectra emitted by the tissue site as a result of illumination by the plurality of electromagnetic radiation wavelengths; and a first data processor coupled to the detecting device for calculating from the detected fluorescence intensity spectra a probability that the tissue site contains a particular type of abnormal tissue.

18. A method for detecting and classifying tissue abnormality in a tissue site, comprising:

applying a plurality of electromagnetic radiation wavelengths to a tissue site to cause the tissue site to produce fluorescence intensity spectra indicative of tissue abnormality;
detecting an intensity of the fluorescence spectra emitted by said site as a result of illumination by said plurality of electromagnetic radiation wavelengths; and calculating from the detected fluorescence spectra a probability that said tissue site contains a particular type of abnormal tissue.

19. A method of detecting and classifying abnormal tissue at a tissue site of a patient, comprising:
applying electromagnetic radiation to the tissue site at a first wavelength that excites different fluorescence intensity spectra within a first wavelength range in normal and abnormal tissue;
applying electromagnetic radiation to the tissue site at a second wavelength that excites different fluorescence intensity spectra within a second wavelength range in tissues having different types of dysplasia;
obtaining measurements of fluorescence intensities from the first wavelength and second wavelength applying steps within the first and second wavelength ranges;
calculating from the measurements of fluorescence intensities a first probability that the tissue site is normal or abnormal and a second probability that the tissue site is of a particular type of dysplasia; and classifying the tissue site as normal or abnormal and if abnormal, as to as to type of dysplasia, in accordance with the first and second probabilities.

20. A method as in claim 19 wherein the calculating step comprises:
preprocessing the measurements of fluorescence intensities within the first wavelength range to remove known and diagnostically immaterial variations therefrom;
applying a first set of orthogonal elements that account for diagnostically significant differences between different fluorescence intensity spectra within the first wavelength range of normal and abnormal tissue in a population inclusive of the patient, to the preprocessed measurements of fluorescence intensities within the first wavelength range, to obtain first scores for the tissue site;
calculating the first probability based on the first site scores;
preprocessing the measurements of fluorescence intensities within the second wavelength range to remove known and diagnostically immaterial variations therefrom;
applying a second set of orthogonal elements that account for diagnostically significant differences between different fluorescence intensity spectra within the second wavelength range of normal and abnormal tissue in a population inclusive of the patient, to the preprocessed measurements of fluorescence intensities within the second wavelength range, to obtain second scores for the tissue site;
and calculating the second probability based on the second site scores.

21. A method as in claim 20 wherein:
the first set of orthogonal elements is a set of principal components;
the first scores for the orthogonal elements are principal component scores;
the second set of orthogonal elements is a set of principal components; and the second scores for the orthogonal elements are principal component scores.

22. A method as in claim 20 further comprising generating the first and second sets of orthogonal elements, wherein the step of generating the first set of orthogonal elements comprises:
selecting a training set comprising a plurality of subjects having a known distribution of normal and histo-pathological tissue types for tissue morphologically similar to the tissue site;

applying electromagnetic radiation at the first wavelength to the tissue of the training set;
measuring first training fluorescence intensities within the first wavelength range resulting from the step of applying electromagnetic radiation to the tissue of the training set;
preprocessing the first training fluorescence intensities to remove known and diagnostically immaterial variations therefrom; and transforming the normalized measurements of first training fluorescence intensities into the first set of orthogonal elements, wherein the first set of orthogonal elements is limited to diagnostically useful orthogonal elements that account for significant variations between the normalized measurements of first training fluorescence intensities in the normal and histo-pathological tissue types;
and wherein the step of generating the second set of orthogonal elements comprises:
applying electromagnetic radiation at the second wavelength to the tissue of the training set;
measuring second training fluorescence intensities within the second wavelength range resulting from the step of applying electromagnetic radiation to the tissue of the training set;
preprocessing the second training fluorescence intensities to remove known and diagnostically immaterial variations therefrom; and transforming the second normalized measurements of training fluorescence intensities into the second set of orthogonal elements, wherein the second set of orthogonal elements is limited to diagnostically useful orthogonal elements that account for significant variations between the normalized measurements of second training fluorescence intensities in the histo-pathological tissue types.

23. A method as in claim 22 further comprising:
calculating first training scores for the first set of orthogonal elements from the first training fluorescence intensities;
generating first probability distributions of the first training scores for each of the normal and histo-pathological tissue types;
calculating second training scores for the second set of orthogonal elements from the second training fluorescence intensities; and generating second probability distributions of the second training scores for each of the histo-pathological tissue types; and obtaining prior probabilities based on the training set;
wherein the first probability calculating step comprises:
developing from the first probability distributions first conditional joint probabilities that the normal and histo-pathological tissue types of the training set will exhibit the first site scores; and calculating the first probability as posterior probabilities of normality and abnormality for the tissue site from the first conditional joint probabilities and the prior probabilities;
and wherein the second probability calculating step comprises:
developing from the second probability distributions second conditional joint probabilities that the histo-pathological tissue types of the training set will exhibit the second site scores; and calculating the second probability as posterior probabilities of the type of histo-pathological tissue for the tissue site from the second conditional joint probabilities and the prior probabilities.

24. A method of classifying a cervical tissue site as normal or of a particular histo-pathological type based on autofluorescence, comprising:
selecting a training set comprising a plurality of subjects having a known distribution of normal and histo-pathological cervical tissue types;
applying electromagnetic radiation at plural excitation wavelengths to excite fluorescence in the various normal and histo-pathological tissue types of the training set;

detecting sets of training fluorescence intensity spectra respectively excited in the various normal and histo-pathological tissue types of the training set at the plural excitation wavelengths;
preprocessing the training fluorescence intensity spectra to remove known and diagnostically immaterial variations therefrom;
transforming the sets of preprocessed training fluorescence intensity spectra for each of the plural excitation wavelengths into a set of principal components that account for significant variations therein;
limiting the sets of principal components to diagnostically useful principal components for each of the plural excitation wavelengths in accordance with the known distribution of normal and histo-pathological tissue types;
calculating respective training scores for the limited sets of principal components from the preprocessed training fluorescence intensity spectra;
generating probability distributions of the calculated training principal component scores for each of the normal and histo-pathological tissue types;
applying electromagnetic radiation at the plural excitation wavelengths to excite fluorescence in the cervical tissue site;
detecting first and second fluorescence intensity spectra excited in the cervical tissue site at the plural excitation wavelengths;
preprocessing the first and second fluorescence intensity spectra to remove known and diagnostically immaterial variations therefrom;
calculating first and second scores for the limited sets of principal components based on the first and second fluorescence spectra;
developing from the probability distributions first and second conditional joint probabilities that the various normal and histo-pathological tissue types of the training set will exhibit the first and second principal component scores;
obtaining prior probabilities based on the training set;
calculating posterior probabilities of normality and abnormality for the cervical tissue site from the first conditional joint probabilities and the prior probabilities;
classifying the cervical site as normal or abnormal based on which of the posterior probabilities of normality and abnormality is highest;
calculating posterior probabilities of type of histo-pathological abnormality for the cervical tissue site from the second conditional joint probabilities and the prior probabilities; and where the cervical site is classified as abnormal in the normal or abnormal classifying step, classifying the cervical site as to type of type of histo-pathological abnormality based on which of the posterior probabilities of type of histo-pathological abnormality is highest.

25. A method as in claim 24 wherein the cervical tissue site is in vivo.

26. A method as in claim 24 wherein the cervical tissue site is in vitro.

27. A method of identifying a set of orthogonal elements to score based on a measurement of tissue fluorescence intensities from a site of unknown tissue, and obtaining probability distributions of scores from which a probability that the unknown tissue is a particular type of tissue may be calculated, comprising:
selecting a training set comprising a plurality of subjects having a known distribution of tissue types for tissue morphologically similar to the unknown tissue;
applying electromagnetic radiation that excites different fluorescence intensity amplitudes within a range of wavelengths in the tissue types of the training set;
measuring fluorescence intensities resulting from the step of applying electromagnetic radiation to the tissue of the training set;
preprocessing the measured fluorescence intensities to remove known and diagnostically immaterial variations therefrom;

transforming the preprocessed measurements of fluorescence intensities into a set of orthogonal elements that account for significant variations between the preprocessed fluorescence intensities from the tissue types;
limiting the set of orthogonal elements to diagnostically useful orthogonal elements in accordance with the known distribution of the tissue types;
calculating scores for the limited set of orthogonal elements; and generating probability distributions of the calculated scores for each of the tissue types.

28. A method of identifying a set of orthogonal principal components to score based on a measurement of tissue fluorescence intensities from a site of unknown tissue, and obtaining probability distributions of principal component scores from which a probability that the unknown tissue is of a particular type, either normal or abnormal or of a particular histo-pathological type, may be calculated, comprising:
selecting a training set comprising a plurality of subjects having a known distribution of tissue types in tissue sites for tissue morphologically similar to the unknown tissue;
applying to the tissue sites electromagnetic radiation that excites different fluorescence intensity amplitudes within a range of wavelengths in the tissue types of the training set;

measuring fluorescence intensities resulting from the step of applying electromagnetic radiation to the tissue sites;
preprocessing the measured fluorescence intensities to remove known and diagnostically immaterial variations therefrom;
transforming the preprocessed measurements of fluorescence intensities into a set of principal components that account for significant variations between the preprocessed fluorescence intensities from the tissue types;
limiting the set of principal components to diagnostically useful principal components in accordance with the known distribution of tissue types;
calculating scores for the limited set of principal components; and generating probability distributions of the calculated principal component scores for each of the tissue types.

29. A method as in claim 28 wherein.
the applied electromagnetic radiation has a wavelength of 337 nm;
the tissue morphology is cervical tissue and the tissue types are squamous normal, columnar normal, low grade squamous intraepithelial lesions, and high grade squamous intraepithelial lesions; and the set of principal components is limited to principal components that demonstrate the statistically most significant differences between normal squamous epithelia and low and high grade squamous intraepithelial lesions.

30. A method as in claim 28 wherein:
the applied electromagnetic radiation has a wavelength of 380 nm;
the tissue morphology is cervical tissue and the tissue types are squamous normal, columnar normal, low grade squamous intraepithelial lesions, and high grade squamous intraepithelial lesions; and the set of principal components is limited to principal components that demonstrate the statistically most significant differences between normal columnar epithelia and low and high grade squamous intraepithelial lesions.

31. A method as in claim 28 wherein:
the applied electromagnetic radiation has a wavelength of 460 nm;
the tissue morphology is cervical tissue and the tissue types are squamous normal, columnar normal, low grade squamous intraepithelial lesions, and high grade squamous intraepithelial lesions; and the set of principal components is limited to principal components that demonstrate the statistically most significant differences between low grade squamous intraepithelial lesions and high grade squamous intraepithelial lesions.

32. A method as in claim 28 wherein the preprocessing step comprises normalizing the measured fluorescence intensities for each of the tissue sites within the range of wavelengths to a peak intensity of one.

33. A method as in claim 28 wherein the preprocessing step comprises:
calculating a mean spectrum for each subject from the measured fluorescence intensities of the tissue sites of said each subject within the range of wavelengths therefor; and subtracting the mean spectrum for said each subject from the measured fluorescence intensities of the tissue sites of said each subject within the range of wavelengths therefor.

34. A method in claim 28 wherein the transforming step comprises:
forming a data matrix D from the fluorescence intensities;
calculating a covariance matrix Z from the matrix D
in accordance with the expression calculating a variance V accounted for by the first Z = n eigenvalues in accordance with the expression retaining eigenvalues and corresponding eigenvectors in a matrix C that account for 99% of the variance V.

35. A method as in claim 34 wherein the transforming step further comprises:
calculating, component loadings C for each of the principal components in accordance with the expression CLij = eliminating fluorescence intensities at wavelengths within the range of wavelengths that are not most highly correlated with the component loadings, prior to the limited step.

36. A method as in claim 34 wherein the limited step comprises:

calculating a principal component score matrix R in accordance with the expression R = D x C; and calculating a diagnostic contribution of each principal component from the matrix R using a two-sided unpaired student's t-test.

37. A method as in claim 36 wherein the probability distributions generating step comprises modeling the probability distributions using a gamma function defined by the expression f(x;.alpha.,.beta.) =