Classification of atherosclerotic rabbit aorta samples by mid-infrared spectroscopy using multivariate data analysis

Abstract

Atherosclerotic and normal rabbit aorta samples show a marked difference in chemical composition governed by the water, lipid, and protein content. The strongly overlapping infrared absorption features of the different constituents, and the complexity of the tissue matrix, render tissue classification by direct evaluation of molecular spectroscopic characteristics obtained from IR reflectance or attenuated total reflectance (ATR) measurements virtually impossible. We apply multivariate analysis and classification techniques based on partial least squares regression (PLS) and linear discriminant analysis to IR spectroscopic data obtained by IR-ATR measurements and reflectance IR microscopy with high predictive accuracy during blind testing. Training data are collected from atherosclerotic and normal rabbit aorta samples. These results demonstrate the potential of IR spectroscopy combined with multivariate classification strategies for the in-vitro identification of normal and atherosclerotic aorta tissue. The prospect for future in-vivo measurement concepts is also discussed.