Performance Analysis of Deep Learning Models for Heart Segmentation in Chest X-ray Images on a Small Dataset

Abstract

The widespread practice of screening of the lungs by radiography makes it possible to analyze the chest area for the presence of extrapulmonary pathologies, such as cardiac pathologies. In many cases, it is advisable to assign the process of solving the problem of analyzing and marking up images to automated algorithms. This paper discusses the performance comparison of multiple deep learning models for heart segmentation on chest x-ray images. The information obtained can be used to improve the algorithms for recognizing pathologies in chest X-ray images.