Laser-induced fluorescence spectroscopy of human colonic mucosa

GASTROENTEROLOGY 1990;99:150-157

Laser-Induced Fluorescence Spectroscopy of Human Colonic Mucosa Detection of Adenomatous

Transformation

CYRUS R. KAPADIA, FRANCIS W. CUTRUZZOLA, KENNETH M. O’BRIEN, MARK L. STETZ, ROSA ENRIQI JEZ, and LAWRENCE I. DECKELBAUM Departments of Medicine and Pathology, West Haven Veterans Administration and Yale University School of Medicine, New Haven, Connecticut

To evaluate the potential of laser-induced fluorescence spectroscopy for the detection of premalignant lesions of the gastrointestinal tract, the hypothesis that adenomatous transformation of colonic mucosa results in an alteration of laser-induced fluorescence that enables its differentiation from normal or hyperplastic tissue was tested. A fiberoptic catheter coupled to a helium-cadmium laser (325nm) and an optical multichannel analyzer were used to obtain fluorescence spectra (350-600nm) from 35 normal colonic specimens and 65 resected adenomatous polyps. A score based on six wavelengths was derived by stepwise multivariate linear regression analysis of the spectra. The mean score (+SEM) was +0.66 + 0.06for normal mucosa and -0.66 + 0.06 for adenomatous polyps (P < 0.001). Spectra from an additional 34 normal specimens, 16 adenomatous polyps, and 16 hyperplastic polyps were prospectively classified with accuracies of loo%, lOO%, and 64%, respectively. The mean score for hyperplastic polyps was significantly different from adenomatous (P < 0.001) but not from normal tissue. Thus, quantitative analysis of fluorescence spectra enables the detection of adenomatous transformation in colonic mucosa.

G

astrointestinal malignancies constitute a major worldwide health problem. In the United States, among internal cancers affecting both sexes, the estimated incidence and mortality for carcinomas of the colon and rectum are second only to lung cancer (11. Gastric carcinoma is a leading health problem in Japan (z), whereas esophageal cancer is common in several developing countries (3). It is now known that most of these cancers arise in premalignant lesions (4).

Me dical Center,

Because the regions of the gastrointestinal tract in which these malignancies arise are easily accessible via endoscopy, resources are being invested to develop more sensitive methods than are currently available for the detection of premalignant and early malignant lesions of the colon, stomach, and esophagus. The availability of such methods would enable a more rational approach to obtaining endoscopic biopsies during cancer surveillance procedures. At present, during such procedures, biopsies are obtained blindly at set locations or guided by visual impression of mucosal abnormalities. Additional diagnostic modalities might substantially improve the accuracy of endoscopic detection of premalignant lesions. The use of optical spectroscopy for tissue diagnosis is a new approach in medical diagnostics. Optical properties of tissue such as absorption and fluorescence may reveal valuable diagnostic information regarding tissue composition and pathology. It has been shown that low-power laser radiation is capable of inducing autofluorescence from tissues without causing tissue damage (5-12). Based on differences in laser-induced fluorescence (LIF) spectra, it has been possible to discriminate normal from atherosclerotic arteries (g-11), normal from carious teeth (121, and normal from malignant tissues (6,7]. Alfano et al. have reported preliminary studies on the ability of LIF spectroscopy to differentiate human lung and breast malignancies from their corresponding normal tissue (13). Similar ability to diagnose histological malignant

Abbreviation used in this paper: LIF, laser-induced fluorescence. 0 1990by the American Gastroenterological Association 0016-5665/96/63.00

COLONIC FLUORESCENCE

July 1990

or premalignant characteristics of gastrointestinal tissue based on LIF spectra would have a great impact on gastrointestinal endoscopy. The purpose of this study was to evaluate the LIF characteristics of normal and abnormal gastrointestinal tissue. The specific hypothesis was that fluorescence could be induced from normal, hyperplastic, and adenomatous colonic tissue by low-power ultraviolet laser irradiation and that analysis of the fluorescence spectra would enable discrimination of premalignant adenomatous transformation of colonic mucosa from normal or hyperplastic mucosa. Methods Tissue Acquisition Fluorescence spectroscopy was performed on polyps removed at colonoscopy and on specimens of normalappearing colonic mucosa obtained either at colonoscopy or at surgery. Specimens were stored in saline before spectroscopy, which was performed within 4 hours of tissue excision. Most specimens of normal mucosa were colonoscopic biopsies from the ascending, transverse, and descending colon and the rectum, obtained from patients undergoing elective colonoscopy for a variety of reasons other than inflammatory bowel disease or acute colitis. All patients gave their informed consent, and the study was approved by the medical center’s Human Investigations Committee. In every subject, two colonic mucosal biopsies were taken from each of the four sites mentioned above. Biopsy specimens were rinsed in saline to remove any blood from hemorrhage at the biopsy site. In addition to colonoscopic biopsies, normal mucosal specimens were also acquired from colons resected from patients with carcinoma. In these cases, spectra were obtained from the mucosal surface of visually normal appearing areas of the colon at a distance (>15 cm] from the tumor. The areas of normal-appearing colon from which spectra were obtained were then removed using a tissue punch. All colonoscopic biopsies of normal-appearing mucosa and all polyps were fixed following spectral acquisition in phos-

Figure 1. Experimental apparatus for fiberoptic IJF spectroscopy. Ultraviolet radiation from a helium-cadmium laser was transmitted by a single optical fiber to the coionic tissue site. The induced tissue fluorescence was collected and transmitted by the same optical fiber to a speetrograph. Tissue fluorescence was spectrally dispersed and imaged onto an intensffled diode array detector. The spectrum was plotted on a CRT display and stored and analyzed by a microcomputer.

Filter

SPECTROSCOPY

151

phate-buffered 10% formalin and processed routinely for light microscopic examination. Six-micron histological sections were stained with H&E, and each specimen was classified histologically in an independent and blinded fashion as normal mucosa, hyperplastic polyp, or adenomatous polyp.

Laser-Induced

Fluorescence

Spectroscopy

The experimental apparatus used to record arterial fluorescence is shown in Figure 1. A continuous-wave helium-cadmium laser (Omnichrome model 356-5M, Chino, CA] operating at a wavelength of 325 nm and a power output of 5-10 mW was coupled to a 400-Mm optical fiber (Diaguide Inc, New York, NY) and used to induce endogenous tissue fluorescence. A custom beam-splitter enabled transmission of the exciting laser radiation and collection of tissue fluorescence using the same fiber. A shutter (Vincent Associates, Rochester, NY] was positioned in the optical path to limit laser irradiation of the tissue to the period of fluorescence sampling (500 millisecond], thereby minimizing tissue photobleaching. The distal end of the optical fiber was positioned perpendicular to, and in light contact with, the tissue surface. Spectroscopy was performed with the specimen in air. Fluences used to induce tissue fluorescence ranged from 4-12 mJ/mm’. The resulting LIF was focused by an achromatic lens and imaged onto the slit of a spectograph (Jarrell-Ash Monospec 27, Waltham, MA). The tissue fluorescence was spectrally dispersed [grating 150 grooves/mm, blaze 450 nm] along the horizontal axis of the exit plane of the spectograph and imaged onto a 700-element intensified linear diode array detector (Princeton Instruments model IRY 700, Trenton, NJ) to analyze fluorescence over a wavelength range from 350-700 nm. Spectral data were digitized by the detector controller (Princeton Instruments model ST-100, Trenton, NJ] and transferred (via a National Instruments GPIB interface, Austin, TX) to a microcomputer (IBM PC AT, Armonk, NY) for storage and analysis. To improve the signal-to-noise ratio, 10 fluorescence spectra were recorded and averaged from each specimen site. To ensure a representative spectral analysis from each specimen, spec-

152

KAPADIA

ET AL.

GASTROENTEROLOGY

tra were recorded from multiple sites (range, Z-10, depending on specimen size). For the normal biopsies, each pair of biopsies from a single site were treated as a single specimen. Individual spectrum acquisition time was 33 milliseconds. A background fluorescence or baseline signal (obtained with the fiber positioned in air) was subtracted from each tissue spectrum to correct for nonzero background caused by dark noise in the linear diode array and to correct for the numerical offset inherent in the analog to digital conversion. Fluorescence intensity at each wavelength was measured in arbitrary units based on the amplitude of the signal detected by the linear diode array. Each tissue spectrum was spectrally and radiometrically calibrated using a mercury vapor lamp (UVP, San Gabriel, CA) and a National Bureau of Standards-traceable calibrated white light source (Optronics model 245A. Orlando, FL), respectively. Radiometric calibration was used to correct for spectral response nonuniformities inherent in the fluorimetry instrumentation.

Spectral Data Processing Computation

and Mean

Spectra

For the purpose of subsequent quantitative LIF score development, the initial 35 normal and 35 adenomatous specimens were grouped into a training set, and the remaining normal, adenomatous, and hyperplastic specimens were grouped into a validation set. Spectra were evaluated from 350-600 nm; truncation at 600 nm was performed to eliminate occasional fiber autofluorescence evident at 650 nm. Tissue spectra were normalized to the total integrated light intensity so that at each wavelength, the fluorescence intensity represented its fractional contribution to the total intensity. By normalization, dependence on procedural variables affecting light collection (e.g., surface topology, fiber-totissue geometry, and excitation and detection efficiency) was eliminated, and individual spectra could be quantitatively compared or averaged. The multiple spectra from each tissue sample were then averaged. The data were then smoothed using a low-pass filter. Resampling of tissue spectra at 15-nm intervals was performed in accordance with the Nyquist criterion (14). This processing reduced each spectrum from 700 spectral intensity values to 17 spectral intensity values (e.g., fluorescence intensity values at each of 17 wavelengths spaced every 15 nm from 350-600 nm) without loss of information. Mean normal and ademomatous spectra were computed from the individual normal and adenomatous spectra, respectively, of the training set by averaging the spectral intensity values at each of these 17 wavelengths.

Quantitative

Laser-Induced

Fluorescence

Score A quantitative LIF score was developed to discriminate adenomatous from normal tissue. The fluorescence spectra obtained from the initial 70 tissue specimens (35 normal, 35 adenomatous) were assembled into a training set and used to develop this tissue classification algorithm. Stepwise multivariate linear regression analysis was used to

Vol. 99, No. 1

relate the spectral intensity features to tissue type. The LIF score was formulated to classify normal colonic spectra with a positive score and adenomatous colonic spectra with a negative score. The stepwise regression used a partial F test (a to enter, 0.15; LYto remove, 0.15) to sequentially incorporate a subset of the 17 spectral intensity features into a discriminant function that formed the LIF score (15). The LIF score generated two distributions of scores, one for each colonic histology. A decision or threshold score was calculated optimally to classify colonic spectra into either of the two categories. The optimal threshold score was determined as the score that classified the normal and adenomatous tissue specimens in the training set with the greatest accuracy. Specimens with spectral LIF scores greater than or equal to the threshold were classified as normal, and specimens with scores less than the threshold were classified as adenomatous. Classification accuracy was defined as the number of specimens correctly classified by the LIF score. A similar fluorescence classification approach has been devised for arterial spectral classification (16). The LIF score was first used to classify the fluorescence spectra constituting the training set and then to classify prospectively the spectra from the normal, adenomatous, and hyperplastic specimens constituting the validation set.

Statistics An unpaired student’s t test was used to compare the LIF scores of the normal and adenomatous spectra (unpaired]. All values, unless otherwise stated, are expressed as the mean + 1 SEM.

Results A total of 47 adenomatous polyps, 16 hyperplastic polyps, 60 normal biopsies, and 9 normal specimens from resected colons were obtained for fluorescence spectroscopy. Histological analysis showed no substantial hemorrhage in any of the biopsy specimens. Adenomatous polyps ranged in size from 2-24 mm, whereas hyperplastic polyps ranged in size from 2-8 mm. The mean spectra from the 35 normal and the 35 adenomatous specimens constituting the training set are shown in Figure 2. The shape of the mean spectrum of normal colonic mucosa visually differs from that of adenomatous colonic mucosa and has a blue-shifted maximal intensity peak. The fluorescence spectra of the normal mucosal biopsies did not differ by site. Likewise, the spectra obtained from colonoscopic biopsies of normal mucosa were similar to those obtained from normal areas of surgically resected colons. A LIF score w&s derived to classify tissue spectra in the training set as normal or adenomatous mucosa. The LIF score derived from stepwise multivariate regression (as detailed in Methods) resulted in the

COLONIC

July 1990

500

FLUORESCENCE

SPECTROSCOPY

153

1

Figure 2. Mean LIF spectra of normal and adenomatoue coionic mucosa. Individual tfssue spectra were resampled at 15-nm increments and averaged to derive the mean spectra. The resultant 17 intensity features between 650 and 600 nm are shown kl SEh4 for normal specimens (A) and adenomatous polyps (A).

Wavelength (nm)

following equation: LIF score = -10.66 - (0.0025) X intensity at 350 nm + (0.0188) X intensity at 380 nm - (0.0088) X intensity at 395 nm + (0.0103) x intensity at 440 nm + (0.0955) x intensity at 560 nm - (0.071) x intensity at 590 nm.

The LIF scores for the spectra from the 35 normal and 35 adenomatous specimens in the training set are shown in Figure 3. The mean LIF scores were + 0.86 f 0.06 for normal colonic mucosa and -0.86 + 0.06 for adenomatous mucosa (P -C0.001). A value of +0.08 was chosen as the LIF score threshold; specimens with LIF scores r+0.08 were thus classified as normal colonic mucosa, and those with scores <+0.08 were classified as adenomatous mucosal transformation. Based on this decision threshold, all 35 normal mucosal specimens and all 35 adenomatous polyps were

. Normal

Adenomatous

Colonic Tissue Type

;

Figure 6. Laser-induced fluorescence scores of the normal and adenomatous colonic spedmens in the trafning set. For each colonic specimen, the stepwfse multfvarfate lfnear regression L&F score is plotted. The mean LfP scores were +0.66 f 0.06 for normal colonfc mucesa and -0.66 f 0.06 for adenomatous mucosa (P < 0.001). using all LIP Bcore of +0.08 a8 a cutoff, 100% of the 66 normal and 65 adenomatous spedmens were correctly classi5ed.

154 KAPADIA ETAL.

GASTROENTEROLOGYVol.99,No.1

correctly classified, thus giving a classification accuracy of 100% for normal mucosa and 100% for adenomatous polyps for the training set. Laser-induced fluorescence scores were obtained prospectively for the fluorescence spectra from the specimens in the validation set. Figure 4 shows the distribution of LIF scores in the validation set. All 34 normal mucosal specimens and all 16 adenomatous polyps were correctly classified, yielding prospective classification accuracies of 100% for normal mucosa and 100% for adenomatous polyps. Figure 4 also shows the LIF scores derived prospectively for the 16 hyperplastic polyps. The mean LIF score for hyperplastic polyps was not significantly different from the LIF score for normal mucosa (P = 0.9) but was significantly different from the mean LIF score for adenomatous polyps (P < 0.001). Fifteen of 16 hyperplastic polyps were correctly classified as normal (i.e., nonadenomatous), for a classification accuracy of 94%. Therefore, in the validation set, polyps could be classified as adenomatous or hyperplastic with 97% accuracy (31 of 32 polyps in the validation set classified correctly). Discussion In this study we have shown that normal coionic tissue emits characteristic fluorescence when excited by light from a helium-cadmium laser. This fluorescence appears to arise from the mucosal layer of the colon because the spectra obtained from colonic mucosal biopsies had characteristics similar to those

obtained from histologically normal areas in colons resected for colonic cancer. This observation is consistent with the report that the sampling depth of heliumcadmium laser-induced fluorescence is approxiThe spectra obtained from different mately 200 pm (17). areas of the colon (e.g., ascending colon, transverse colon, descending colon, and rectum) were indistinguishable, but all differed from the spectra from adenomatous polyps. To formulate a model to differentiate fluorescence spectra of normal mucosa from those of adenomatous polyps, spectra were analyzed from the first 35 normal mucosal specimens and the first 35 adenomatous polyps using stepwise multivariate linear regression analysis to develop a classification algorithm. Based on this model, all normal mucosal specimens and all adenomatious polyps were classified correctly. The next part of our study was designed to validate the fluorescence-based classification algorithm prospectively. To do this we prospectively analyzed spectra from an additional 34 normal mucosal specimens and 16 adenomatous polyps, as well as spectra obtained from 16 hyperplastic polyps to assess whether these benign lesions could be distinguished from adenomatous polyps based on LIF. Application of the algorithm to the recorded spectra classified 34 of 34 normal specimens and 16 of 16 adenomatous polyps correctly. This gave a prospective classification accuracy of 100% for both normal and adenomatous tissue. Furthermore, 15 of 16 hyperplastic polyps were classified correctly as normal (e.g., nonadenomatous) mucosa (94% accuracy). The ability to discriminate between

!

IA

3

-5-

-10

.

I

Adenomatous

,

I

Normal

Colonic Tissue Type

Hyperplastic

Figure 4. Laser-induced fluorescence scores of the adenomatous, normal, and hyperplastic colonic specimens in the validation set. The LIF scores of the spectra from 16 adenomatous polyps, 94 normal mucosal specimens, and 19 hyperplastic polyps are plotted. The LIP score threshold of +0.09 is represented by the horizontal dashed line. The overall prospective classification accuracy was 98%.

July 1990

hyperplastic and adenomatous polyps was not dependent on polyp size, as evidenced by the overlap in size distributions of these two polyp types. Although adenomatous polyps ranged from 2-24 mm in size, spectral features were similar; on the other hand, the spectral features of hyperplastic polyps were very different from those of adenomatous polyps of similar size. These data suggest that adenomatous colonic mucosa can be differentiated from normal and hyperplastic mucosa. Although adenomatous polyps could be differentiated from normal mucosa or hyperplastic polyps, the etiology of the difference in laser-induced characteristics was not determined in this study. The discriminating fluorescence characteristics may be due either to the presence of a specific unique fluorophore in the adenomatous tissue or to differing relative amounts of similar fluorophores. Diagnostic fluorescence spectroscopy in the arterial system has shown that unique fluorophores have been isolated from atherosclerotic plaque (18,19), yet the major difference in normal and atherosclerotic fluorescence appears to be attributable to differences in the relative concentrations of elastin and collagen (20). There are two approaches by which diagnostic tissue fluorescence may be used to detect abnormal tissue. One approach requires the prior administration of a fluorophore to the patient, which preferentially localizes to the target abnormal lesion. The lesion is subsequently identified by the fluorescence emitted by the localizing fluorophore when it is excited by light of an appropriate wavelength. In contrast, the surrounding normal tissue does not fluoresce similarly because localization of the administered fluorophore is absent. Analysis of this type of exogenous fluorescence has been used in preliminary studies for the identification of malignant lesions pretreated with hematoporphyrin derivative (21~231, tetracycline (24,251, or, more recently, N, N’-bis(B-ethyl-l, 3, dioxotane) kryptocyanine (EDKC) (B. Levin, M.D. Anderson Hospital, Houston, TX, September 1989, personal communication]. The second approach is based on analysis of the intrinsic fluorescence properties of tissue excited by light of a suitable wavelength. Analysis of autofluorescence eliminates the need to pretreat a patient with an exogenous agent that may have risks of toxicity such as the photosensitization present following administration of hematoporphyrin derivative. Previous studies from our laboratory have shown that laser-induced autofluorescence of arterial tissue can be used to differentiate atherosclerotic from normal arterial tissue (9). Autofluorescence has also been used to detect malignant kidney and prostate tumors in animal models (61, as well as malignant human lung and breast This report is the first published study of tumors (13). the use of diagnostic tissue autofluorescence to detect

COLONIC

FLUORESCENCE

SPECTROSCOPY

155

premalignant lesions and to evaluate the autofluorescence characteristics of gastrointestinal tissue. We chose to evaluate fluorescence characteristics of coionic tissue because the adenomatous polyp is the best-defined and the most abundant premalignant lesion of the gastrointestinal tract in Western countries. Furthermore, a visually similar but not premalignant lesion, the hyperplastic polyp, could be evaluated to test the validity of our method. In this study, ultraviolet laser irradiation was chosen to induce fluorescence from colonic tissue. The choice of a helium-cadmium-emitting laser at 325 nm was based on previous experience in our laboratory with this laser for the spectroscopic characterization of arterial tissue (9,161. In addition, ultraviolet irradiation provides the opportunity to induce fluorescence over the entire visible spectrum, which may provide greater diagnostic information than would be available from a partial spectrum induced by visible light excitation. It is possible that other laser wavelengths could be used for fluorescence excitation that might provide better discrimination between spectra obtained from normal and abnormal tissues. For example, different laser wavelengths may be necessary to provide optimum discrimination between different types of lesions or lesions in different parts of the gastrointestinal tract. Two preliminary reports using laser-induced autofluorescence to discriminate normal human colonic mucosa from adenomatous polyps (26) or cancer (27) have recently been published in abstract form. In the first report (261,the investigators evaluated the fluorescence spectra of normal and adenomatous colonic tissue using excitation wavelengths ranging from 290600 nm. Although spectroscopy was not performed fiberoptically with a laser source, the investigators found that good discrimination could be achieved between normal and adenomatous polyps. The best discrimination was obtained using ultraviolet excitation. In the second report (271,qualitative comparisons were made between emission spectra obtained from normal mucosa and cancerous tissue in nine surgically resected specimens of human colon. Fluorescent spectra induced by a nitrogen laser (337 nm) from normal colonic mucosa showed two discernible peaks, whereas spectra from eight of the nine cancer specimens showed only one peak. Therefore, ultraviolet laserinduced fluorescence may be capable not only of detecting adenomatous transformation, but also of detecting malignant colonic tissue. Further investigation is needed to evaluate whether the fluorescence of adenocarcinomas is similar to that of adenomatous tissue, as well as whether histological variation in adenomatous polyps (e.g., sessile vs. pedunculated, tubular vs. villous vs. tubulovillous) correlates with variation in fluorescence spectra. The purpose of extending the methodology of LIF

156 KAPADIA ET AL.

spectroscopy to the area of gastrointestinal endoscopy is not to replace histology, but rather to make it more effective. During an endoscopic examination, spectral information from a lesion would be instantly available to guide an endoscopist to take biopsies from areas most likely to yield positive results. Sampling error would thus be greatly minimized. Nowhere is such technology needed to a greater extent than in the detection of premalignant lesions of the gastrointestinal tract such as Barrett’s esophagitis and high-grade dysplasias of the colon occurring in inflammatory bowel disease. This technique would also be of use in allowing decisions about the removal of small polyps discovered during colonoscopy. If fluorescence signals suggest the polyp to be of the hyperplastic type, the endoscopist may decide against a polypectomy. The endoscopic removal of near sessile polyps is being carried out with increasing frequency (28). Laserinduced fluorescence spectroscopy could provide valuable information as to the type of polyp or, following polypectomy, as to the presence of residual adenomatous tissue at the polypectomy site. High-power lasers are currently used to ablate tumors of the esophagus and colon (32,33). Laser-induced (29,30), stomach (31), fluorescence spectroscopy might also provide valuable information as to whether neoplastic tissue remains or whether all of the tumor has been extirpated. Pseudopolyps, present in the colons of patients with inflammatory bowel disease, are frequently impossible to differentiate from adenomatous polyps by visual examination through an endoscope. Thus, it is not uncommon for an endoscopist to be faced with a dilemma when encountering numerous polypoid lesions in a patient with ulcerative colitis or Crohn’s disease. Colonoscopic polypectomy of all of these lesions to obtain a definitive histological answer is usually impossible because many of these lesions are sessile. Biopsies from such lesions would carry a very high sampling error. If it is possible to differentiate inflammatory tissue from tissue that has undergone adenomatous or malignant transformation on the basis of LIF, this technique would be an ideal tool to distinguish between pseudopolyps and adenomatous polyps. A vast number of spectra could be obtained and instantly analyzed from most of these lesions, and polypectomy or extensive biopsy could be reserved for those lesions with abnormal fluorescence spectra. A unique feature of the method used in our study is that a single optical fiber is used both to transmit the ultraviolet radiation from the laser to the tissue and to transmit the fluorescence emitted by the tissue to the spectrograph. Because the optical fiber is only 400 pm in diameter, it can readily pass through the biopsy channel of all commercially available fiberoptic or video endoscopes, thus enabling the system to be used during routine endoscopic procedures. Therefore, it is

GASTROENTEROLOGY

Vol. 99, No. 1

not necessary to design new endoscopes at high cost, as has been necessary for other noninvasive endoscopic diagnostic techniques such as ultrasonic endoscopy * Diagnostic fluorescence spectroscopy performed through a single optical fiber provided diagnostic spectral information from the tissue at the fiber tip. Technological advancements involving fiberoptic bundles and image processing could theoretically provide two-dimensional spectroscopic imaging (7,34,35). This would facilitate screening by simultaneously screening a larger surface area. It may be possible to use the same imaging bundle to provide a direct visual image (using white light illumination and recording of the reflected light) and a spectroscopic image (using ultraviolet illumination and recording of the emitted fluorescence) for diagnosis. The results of this study show that quantitative analysis of LIF spectra enables the detection of adenomatous transformation in colonic mucosa. Therefore, LIF spectroscopy holds promise as a potential tool for the endoscopic diagnosis of premalignant lesions of the gastrointestinal tract. However, clinical application of fluorescence spectroscopy to detect premalignant lesions will require in vivo confirmation of the results presented here. Comparison of in vitro and in vivo arterial spectra have shown only minimal differences (36) and suggests that in vivo colonic fluorescence spectroscopy will validate the results of this study. Evaluation of in vivo and in vitro colonic fluorescence in an animal model is currently under way in our laboratory. References 1. Silverg E. Lubera JA. Cancer statistics, 1989. CA 1989:39:3-20. 2. Nagayo T. Human gastric cancer. New York: Springer-Verlag,

1986:7-16. 3. Day NE. The geographic pathology of cancer of the esophagus. Br Med Bull 1984;40:329-334. 4. Morson BC, Jass JR, Sohin LH. Precancerous lesions of the gastrointestinal tract. London: Bailliere Tindall, 1985:3-15. 5. Deckelbaum LI. Laser-induced arterial fluorescence spectroscopy. In: Abela GS, ed. Lasers in cardiovascular medicine and surgery. Norwell, MA: Martinus Niihoff [in press). 6. Alfano RR, Tata DB, Corder0 J. Tomashefsky P, Longo FW, Alfano MA. Laser-induced fluorescence spectroscopy from native cancerous and normal tissues. IEEE J Quantum Electron 1984;QE-20:1507-1511. 7. Andersson PS, Montan S, Svanberg S. Multispectral system for medical fluorescence imaging. IEEE J Quantum Electron 1987; QE-23:1798-1805. 6. Leffell DJ. Stetz ML, Milstone LM, Deckelbaum LI. A technique for the evaluation of photoaging. Arch Dermatol1988;124:15141518. 9. Deckelbaum LI, Lam JK, Cabin HS, Clubb KS, Long MB. Discrimination of normal and atherosclerotic aorta by laserinduced fluorescence. Lasers Med Surg 1987:7:330-335. 10. Leon MB, Lu DY, Prevosti LG, Macy WW, Smith PD, Granovsky M, Bonner RF, Balaban RS. Human arterial surface fluores-

July 1990

cence: atherosclerotic plaque identification and effects of laser atheroma ablation. J Am Co11Cardiol1988:12:94-102. C, Ratliff NB, 11.Kittrell C, Willett RL, de 10s Santos-Pacheo Kramer JR, Malk EG. Feld MS. Diagnosis of fibrous arterial atherosclerosis using fluorescence. Appl Optics 1985;25:22802281. 12. Alfano RR, Lam W, Zarrabi H, Alfano MA, Corder0 J, Tata DB,

Swenberg CE. Human teeth with and without caries studied by laser scattering fluorescence and absorption spectroscopy. IEEE J Quantum Electron 1964;QE-20:1512-1516. 13. Alfano RR, Tang GC, Pradham A, Lam W, Choy DS, Opher E. Fluorescence spectra from cancerous and normal human breast and lung tissues. IEEE J Quantum Electron 1987;QE-23:18061811. 14. Oppenheim AV, Schafer RW. Digital signal processing. Englewood Cliffs, NJ: Prentice-Hall, 1975:28. 15. Bendel RA, Afifi AA. Comparison of stopping rules in forward “stepwise” regression. J Am Stat Assoc 1987;72:46-53. 18. O’Brien KM, Gmitro AF, Gindi GR, Stetz ML, Cutruzzola FW, Laifer LI, Deckelbaum LI. Development and evaluation of spectral classification algorithms for fluorescence guided laser angioplasty. IEEE Trans Biomed Engineering 1989;36:424-431. 17. Gmitro AF, Cutruzzola FA, Stetz ML, Deckelbaum LI. Measurement depth of laser-induced tissue fluorescence with application to laser angioplasty. Appl Optimol1988;27:1844-1849. 18. Banga I, Bihari-Varga M. Investigations of free and elastinbound fluorescent substances present in the atherosclerotic lipid and calcium plaques. Connect Tissue Res 1974;2:237-241. 19. Blankenhorn DH, Braunstein H. Carotenoids in man, III. The microscopic pattern of fluorescence in atheromas, and its relation to their growth. J Clin Invest 1958;37:160-165. 20. Laifer LI, O’Brien KM, Stetz ML, Gindi GR. Deckelbaum LI. Etiology of the fluorescence difference between normal and atherosclerotic arteries (abstr). Circulation 1988;78(suppl 2):11448. 21. Kato H, Aizawa K, Ono J, Konaka C, Kawate N, Yoneyama K, Kinoshita K, Nishimiya K, Sakai H, Noguchi M, Tomono T, Kawasaki S, Tokuda Y, Hayara Y. Clinical measurement of tumor fluorescence using a new diagnostic system with hematoporphyrin derivative, laser photoradiation, and a spectroscope. Laser Surg Med 1984;4:49-58. 22. Montan S, Svanberg K, Svanberg S. Multicolor imaging and contrast enhancement in cancer-tumor localization using laserinduced fluorescence in hematoporphyrin-derivative-bearing tissue. Optics Lett 1985:10:56-58. 23. Profio AE, Doiron DR, Balchum OJ, Huth GC. Fluorescence bronchoscopy for localization of carcinoma in situ. Med Phys 1983;10:35-39. 24. Klinger J, Katz R. Tetracycline fluorescence in the diagnosis of gastric carcinoma. Gastroenterology 1961;41:29-32. 25. Gupta S, Annamma ML, Gupta S. Tetracycline fluorescence in gastric malignancy. Am J Proctol Gastroenterol Colon Rectal Surg 1979;30:17-18.

COLONIC

FLUORESCENCE

SPECTROSCOPY

157

26. Richards-Kortum R, Tong L, Rava R, Fitzmaurice M, Petras R, Sivak M. Feld MS. Spectral studies of GI tissues: optimizing excitation and emission wavelengths for discrimination of normal and adenomatous tissues (abstr). Lasers Surg Med 1989;35: 184.

27. Yakshe PN, Bonner RF, Cohen P, Leon MB, Fleischer DE. Laser-induced fluorescence spectroscopy may distinguish colon cancer from normal human colon (abstr). Gastrointest Endosc 1989;35:184. 28. Christie JP. Colonoscopic excision of sessile polyps. Am J Gastroenterol1976;66:23-28. 29. Fleischer D, Kessler F. Endoscopic Nd:YAG laser therapy for carcinoma of the esophagus: a new form of palliative treatment. Gastroenterology 1983;85:600-606. 30. Lightdale CJ, Zimbalist E, Winawer SJ. Outpatient management of esophageal cancer with endoscopic Nd:YAG laser. Am J Gastroenterol1987;82:46-50. 31. Oguro Y, Hirashima T, Tajiri H, Yoshida S, Yamaguchi H, Yoshimori M, Itabashi M, Hirota T. Endoscopic treatment of early gastric cancer, polypectomy and laser treatment. Jpn J Clin Oncol1984;14:271-282. 32. Bowers J. Laser therapy of colonic neoplasms. In: Fleischer D, Jensen D. Bright-Asare P, eds. Therapeutic laser endoscopy in gastrointestinal disease. Boston: Martinus Nijhoff, 1983:139-150. 33. Mathus-Vliegen

EMH. Tytgat GNJ. Nd:YAG laser photocoagulation in colorectal adenoma. Gastroenterology 1986;90:1865-

1873. 34. Barrows GH, Sisken JE, Allegra JC. Grasch SD. Measurement

of fluorescence using digital integration of video images. J Histochem Cytochem 1984;32:741-746. 35. Sartori M, Sauerbrey R, Kubodera S. Tittel FK, Roberts R, Henry PD. Autofluorescence maps of atherosclerotic human arteries-a new technique in medical imaging. IEEE J Quantum Electron 1987;QE-23:1794-1797. 36. Deckelbaum LI. Sarembock IJ. Stetz ML, O’Brien KM, Cutruzzola FW, Gmitro AF. In-vivo fluorescence spectroscopy of normal and atherosclerotic arteries. Proc SPIE: Optical Fibers III 1988;906:314-319.

Received July 31,1989.Accepted January 22,1996. Address requests for reprints to: Lawrence I. Deckelbaum, M.D., West Haven Veterans Administration Medical Center/IIIB, West Spring Street, West Haven, Connecticut 06516. This work was supported by grants from the Whitaker Foundation and the National Institutes of Health (HL38723). Part of this work was initially presented at the Annual Meeting of the American Gastroenterological Association, May 1988. The authors would like to express their gratitude to Carol Schneider for preparation of the manuscript.