Histological Classification of Raman Spectra of Human Coronary Artery Atherosclerosis Using Principal Component Analysis

Abstract

We present a nonparametric method of analysis of Raman spectra of coronary artery tissue to classify atherosclerotic lesions. The method correlates the principal component scores of the Raman spectra with the tissue pathology. A data set composed of 97 samples of human coronary artery was used to develop the diagnostic algorithm, and a second data set composed of 68 samples was then used to test this algorithm prospectively. The results show that the algorithm can accurately classify coronary artery tissue into three classes: nonatherosclerotic, noncalcified plaque, and calcified plaque. The accuracy of this classification scheme is comparable to that previously achieved by means of a biochemical analysis of the Raman spectra using the same data.