Application of principal component analysis and linear discriminant analysis on elastic light single-scattering spectroscopic data for classification of prostate tissue: A preliminary study

Abstract

Elastic Light Scattering Spectroscopy (ELSS) system with a single optical fiber probe was employed to detect positive surgical margins of prostate cancer in combination with principal components and linear discriminant analysis. ELSSS spectra in the 450-750 nm wavelength regions were obtained from the total of 299 tissue samples from 27 patients. The ELSSS spectral data were compared against the “gold standard” histopathology results. Data analysis was done using principal components analysis, followed by linear discriminant analysis. Receiver Operating Characteristic (ROC) curve was calculated for diagnostic performance. Classification based on the discriminant score provided a sensitivity of 100% and a specificity of 95.2%, in differentiating benign from malign surgical margins of prostate tissues, and the area under the ROC curve of 0.98. In this study, it was shown that ELSSS system can accurately distinguish benign and malign surgical margins of prostate tissues with high sensitivity and specificity. In our study, diagnostic accuracy showed that the applicability of in-vivo fiber optic tissue classification.