Automatic processing of histological imaging to aid diagnosis of cardiac remodeling

Abstract

Despite the growing level of research in the medical field, heart diseases are still predominant and recurrent, regardless of the degree of economic and social development of different populations. The absence of oxygen and nutrient intake causes cardiac cells to die, which are replaced by nonfunctional fibrotic tissue, leading to an accumulation of proteins in the extracellular matrix, usually replaced by collagen. The amount of interstitial collagen in myocardial fibers plays an important role in identifying changes in the heart after cardiac remodeling. This paper aims to model, implement and evaluate a method for automatic analysis of microscopic images of cardiac tissue, aiming to quantify the presence of interstitial collagen in myocardial fibers, using image processing and feature extraction techniques, in order to train a classifier that provides a reference to assist in the identification of reactive fibrosis. The method was evaluated by statistical measures, taking as reference the quantification previously performed by specialists. In the experimental tests our method achieved accuracy rate of 92.4%, showing that it is reliable and promising.