Pattern recognition of multiple excitation autofluorescence spectra for colon tissue classification.

Objectives: The aim of this study was to explore the usefulness of multiple excitation autofluorescence (AF) and a spectral feature-based pattern recognition in classification of colon tissues.

Materials And Methods: Under four different excitation wavelengths (337, 375, 405 and 460 nm), AF spectra of freshly excised normal and adenocarcinoma colon tissues were measured. Pattern recognition method including features extraction, data reduction using principal component analysis (PCA) and Fisher's discriminant analysis (FDA) were performed for classification.

Results: There was a significantly difference between spectral patterns of normal and adenocarcinoma tissues. Compared with the other three excitation wavelengths, the AF spectra obtained under 337 nm excitation provided more diagnostic information, but also more sensitive to the trivial change resulted from neoplastic transformation. For discriminating normal from adenocarcinoma tissues, the sensitivity, specificity and accuracy using 337 nm excitation in the present study were 88.9%, 80.0% and 83.9%, respectively. Compared these values with those determined from multispectral data analysis, our findings indicate that the latter has higher specificity while maintaining the same sensitivity (sensitivity 88.9% vs. 88.9%, specificity 91.4% vs. 80.0%, and accuracy 90.3% vs. 83.9%).

Conclusion: This study suggests that the pattern recognition of the multiple excitation AF spectra is an effective algorithm for improving the diagnostic accuracy of adenocarcinoma.