Overview of Deep Learning Based Cardiac MR Image Segmentation Methods

Abstract

Cardiac image segmentation has great significance of making correct diagnoses of cardiac diseases, and providing useful information for clinical treatment and surgery. Among cardiac images, magnetic resonance imaging (MRI) is the most common modality. With the rapid development of deep learning, it is widely used in various fields such as computer vision, natural language processing and visual recognition. In the meantime, a large number of deep learning based methods have been proposed for Cardiac MR image segmentation. In this paper, we attempt to provide a survey on such methods published in peer-reviewed journals and conferences between 2017 and 2021, including their motivations and segmentation strategies. We then explore the advantages, disadvantages and the challenges such as generalization ability, interpretability and so on. In addition, several publicly available cardiac image datasets and evaluation metrics are given. This paper will offer useful guidelines to study MR image segmentation for researchers and practitioners working in related domains.