Analysis of mutli-spectral images from cardiovascular tissue by means of blind source separation methods

Abstract

This article presents the use of blind source separation methods for the decomposition of cardiovascular tissue multi-spectral images in its main morphological components. The evaluated images were acquired with two kinds of systems: one based in a confocal configuration and another based in interference filters. The implemented source separation algorithms are based on a multiplicative coefficient upload and on Principal Component Analysis (PCA) techniques. The goal is to represent a given multi-spectral image as the weighted sum of different components. The resulting weighted coefficients are used to quantify the content of the main components in a given multi-spectral image. The methodology is validated on cardiovascular bovine tissue. The results show that PCA not only allows image reduction but also an increase in the image contrast. This fact allows for a better determination of the tissue's structure. Also, the result of applying NMF shows that the method allows for maps that quantify the principal chromophores that compose cardiovascular tissue.