Systematic Analysis of CMR Segmentation Using Deep Learning

Abstract

This review provides an overview of cardiac segmentation by using Deep learning for MR images. Cardiac MR is widely used due to its unique capability as non invasive imaging. CMR images are used to derive cardiac indices to diagnose various cardiac diseases by segmentation of heart chambers. There are number of challenges in automatic cardiac MR segmentation due to characteristics of MR images because of continuous movement of heart. Now days, deep learning become widely used technique to segment heart chambers for efficient and accurate results. In this review, we provide overview of publically available CMRI datasets, descriptions of currently available methods to segment Left ventricle, Right ventricle, and Myocardium. After reviewing various techniques, we describe limitations and possible solutions. We conclude deep learning based segmentation issues in the currentapproaches.