CHEMOMETRICAL ANALYSIS OF ENDOMETRIAL TISSUE FLUORESCENCE SPECTRA A.Vaitkuviene a *, E.Auksorius a, D. Fuchs b, V.Gavriushin a a Vilnius University, Institute of Materials Science and Applied Research, b Innsbruck University, Institute of Medical Chemistry and Biochemistry, and Ludwig Boltzmann Institute of AIDS-Research. ABSTRACT An effort has been made to detect neopterin spectrum in fluorescence of premalignant endometrial tissue and to estimate the number of fluorophores naturally existing in the tissue with fluorescence present above the noise level. Endometrial Tissue fluorescence was measured in vitro by excitation with the third harmonic of Nd YAG laser. Multivariate curve resolution was used for testing neopterin presence in endometrial tissue. Fluorescence spectra of neopterin was measured and used as a target spectrum for testing. Seven factors fluorescence of natural fluorophores of endometrial tissue were found to be present above the noise level in the overall autofluorescence. Neopterin concentration may be too low in endometrial tissue to make its fluorescence above the noise level because neopterin spectrum was not found to be among the spectra resolved by multivariate curve resolution. An intensity increase in the neopterin spectrum spectral region in hyperplastic endometrial samples might be associated with neopterin concentration increase. Keywords: Biomarker, Laser induced fluorescence spectroscopy, Multivariate curve resolution. 1. INTRODUCTION Diagnostic techniques based on optical spectroscopy have the potential to link the biochemical and morphological properties of tissues to individual patient care. If applied successfully, optical spectroscopy has the potential to represent an important step forward toward advances in diagnostic and therapeutic medical applications.currently, one of the most widely explored applications of fluorescence spectroscopy is the detection of endoscopically invisible, early neoplastic growth in epithelial tissue sites. There are no effective diagnostic techniques for these early tissue transformations. If fluorescence spectroscopy can be applied successfully as a diagnostic technique in this clinical context, it may increase the potential for curative treatment and thus reduce complications. Fluorescence spectroscopy is ideally suited for this application because of its ability to examine tissue surfaces rather than tissue volumes, and its adaptability to an endoscopic device. Biological molecules that exhibit endogenous fluorescence are amino acids, structural proteins, enzymes and coenzymes, vitamins, lipids and porphyrins. Their excitation maxima lie in the range 250 450 nm, whereas their emission maxima lie in the range 280 700 nm [1]. The endogenous fluorophores that are speculated to play a role in transformations that occur with carcinogenesis are the amino acids tryptophan and tyrosine, the structural proteins collagen and elastin, the coenzymes NADH and FAD and porphyrins biological processes. * Correspondence: Women Clinics, Vilnius University, Saulėtekio al. 6 III 2040 Vilnius, Lithuania Laser Florence 2001: A Window on the Laser Medicine World, Leonardo Longo, Alfons G. Hofstetter MihaiI Lucian Pascu, Wilhelm R. A. Waidelich Editors Proceedings of SPIE Vol 4903 (2002) 240 2002 SPIE - 1605-7422/01/$15.00
The altered biochemical and morphological state that occurs as tissue progresses from a nondiseased to a diseased state is detected in the spectral characteristics of the measured fluorescence. Neopterin is a marker of stimulation of celular immune response. Tumor necrosis factor-α enhances interferon induced neopterin production. Neopterin concentration is higher in patients with a malignant tumor. Therefore, neopterin seems to be an important component of spectral analysis of premalignant and malignant tissue [2]. Information about biological molecules presence or absence in tissue and their concentration might be of particular importance. Multivariate curve analysis (MCR) provides the number of chemical species present in the sample, the spectrum of each species for identification, and the concentration for each species. Curve resolution is chemometrical tool based on the determination of qualitative information and the recovery of response profiles. A lot of modifications of curve resolution techniques and applications have been published [3,4]. The traditional curve resolution techniques can be adapted to analyse fluorescence spectra in order to resolve data into individual concentration and pure spectra. The additional benefit of spectra noise reduction also results from the MCR techniques. 2. MATERIALS AND METHODS Materials. Sixty patients were included in the study. Endometrial tissues were biopsied and classified by routine histopathology as Normal, Hyperplastic (HP) or Cancerous (CN). The 36 normal, 19 hyperplastic and 5 cancerous samples were under study. Laser induced fluorescence measurements in vitro were performed with laser as excitation source. NAD:YAG laser third harmonic (3,51eV) providing 70 ns pulses and 1Hz repetition rate was used. Data matrix - fluorescence spectra of 60 samples consisting of 220 points each were obtained. Thus, Data matrix consisting of 220 columns (wavelength) and 60 rows (samples) were analysed. A fluorescence spectrum of neopterin (Schircks-Laboratories, Jona, Switzerland) was measured. Chemometrics. Chemometrics is a part of chemistry that develops mathematical and statistical methods for analysing chemical data. Multivariate curve resolution is a chemometrical tool based on the determination of qualitative information and the recovery of response profiles. Multivariate curve analysis provides the number of chemical species present in the sample, the pure spectrum of each species for identification, and the concentration for each species. The pure spectra obtained by multivariate curve resolution can be used to identify chemical species using library spectra.in the curve resolution methods, an iterative least squares optimisation procedure is implemented. Multivariate curve resolution consisting of principal component analysis combined with non-oblique factor rotation and alternating least-squares optimisation was used for fluorophores present in endometrial tissue spectral and concentration profiles resolution. Principal component analysis creates a set consisting of a small number of specially chosen basis spectra, called principal components, which can accurately characterize all of the spectral changes in a set of spectral data. Statistically significant principal components are called factors. However, factors are not yet directly related to any real chemical spectra (they, for example, are negative). Non-oblique factor rotation transforms factors to physically meaningful spectral profiles. Alternating least squares optimised the profiles. 3. RESULTS Laser induced fluorescence spectra were obtained from sixty samples. In Figure 1, the normalized spectra are presented. 241 Proc SPIE Vol. 4903
Fig. 1. Normalized fluorescence spectra. Principal component analysis was used to derive principal components - basis spectra responsible for the spectral variation. Derived principal components were tested with the Malinowski s indicator function [4] to select statistically significant principal components - factors. Remaining principal components after factor selection describes only noise and they are discarded from further analysis. Seven factors were found to be statistically significant. In Figure 2, the first three principal components - factors are shown. Fig. 2. First three principal components - Factors considered as basis spectra characterizing main spectral changes in the data set. In Figure 3, seventh and eighth principal components are shown. The eighth and further principal components describe only noise. These last principal components are discarded and not included in the further analysis. Proc. SPIE Vol. 4903 242
Fig. 3. Seven principal components are considered. The eighth and further principal components describe only noise. These last principal components are discarded and seven principal components factors are left for further analysis The factors are not directly related to any real chemical spectra (they, for example, are negative). Thus, non-orthogonal oblique rotation (SYSTAT 10, SPSS) was used to rotate factors. In Figure 4, seven rotated factors are shown. The rotation has no unique solution in attempt to resolve spectral profiles. Fig. 4. Spectra derived by factors non-orthogonal oblique rotation. Ambiguities presented in recovering profiles using factor rotation are solved by use of iterative optimisation methods and of chemical constraints. Alternating least squares algorithm with codes written in MatLab (MatLab version 6, Mathworks) was used for the optimisation. The rotated factors were used as initial estimates and non-negativity of spectra and concentrations, unimodality and linear additivity of spectra were assumed as chemical constraints. In Figure 5, resolved spectral profiles by multivariate curve resolution are shown. 243 Proc SPIE Vol. 4903
Fig. 5. Spectra derived by multivariate curve resolution. An attempt might be made to identify resolved spectral profiles (Fig. 5). However curve resolution is restricted to spectra overlapping. Resolved spectra can be compared to fluorophores spectra (Fig. 6) used in our previous work [6]. Fig. 6. Spectra of collagen, elastin, neopterin, NADH, Vitamins, Caroten an porphyrins respectively, found elsewhere, derived from spectra or measured during the research (neopterin). Multivariate curve resolution was used to resolve concentration profiles as well. Average concentrations of resolved spectral components are shown by the average of spectra intensity in Fig. 7. No one of the pure spectra (Fig. 5) resolved by multivariate curve resolution from fluorescence data (Fig. 1) was identified as a neopterin spectrum (Fig. 6). However, there is an intensity increase in the neopterin spectra spectral region (about 400-500nm) in hyperplastic endometrial samples. Proc. SPIE Vol. 4903 244
Fig 7. Average concentrations of resolved spectral components are shown by the average of spectra intensity. 4. CONCLUSION AND SUMMARY Seven factors fluorescence of natural fluorophores of endometrial tissue were found to be present above the noise level in the overall autofluorescence spectra. The seven pure spectra obtained by multivariate curve resolution can be used to identify chemical species using library spectra. Neopterin concentration may be too low in endometrial tissue to make its fluorescence above the noise level because neopterin spectrum was not found to be among the spectra resolved by multivariate curve resolution. There was observed an intensity increase in the neopterin spectrum spectral region (about 400-500nm) in hyperplastic endometrial samples which might be associated with neopterin concentration increase but the changes are subtle and are not statistically significant. REFERENCE 1. R. Richards-Kortum, E. Sevick-Muraca, Quantitative Optical Spectroscopy for Tissue Diagnosis, Annu. Rev. Phys. Chem., 47, 555 606 (1996) 2. D. Fuchs, G. Weiss, H. Wachter, Neopterin, biochemistry and clinical use as a marker for cellular immune reactions. Int Arch Allergy Immunol 101:1-6 (1993) 3. E. Casassas, I. Marqus, R. Tauler. Study of acid-base properties of fulvic acids using fluorescencespectrometry and multivariate curve resolution methods. Anal. Chem. Act. 45: 473-484 (1995) 4. R. Tauler, B.R. Kowalski and S. Flemming. Multivariate curve resolution applied to spectral data from multiple runs of an industrial process. Anal. Chem., 65 2040-2047 (1993) 5. E.R. Malinowski, Factor Analysis in Chemistry, 2nd ed., Wiley, Chichester, 1991. 6. V. Gavryushin, J. Vaitkus, A. Vaitkuviene. Methodology of medical diagnostics of human tissue by fluorescence. Lithuanian J. Physics, Vol. 40, No.4, 232-236 (2000) 245 Proc SPIE Vol. 4903