Right Ventricle Segmentation in Cardiac MR Images Using Convolutional Neural Network Architecture

Abstract

Artificial intelligence approaches have vastly improved the health care field in several tasks such as image processing and information extraction. These techniques assist experts to predict and diagnose the risk of certain diseases earlier and more efficiently, such as cardiovascular disease through the evaluation of cardiac function. This paper presents a segmented private dataset for Right Ventricle, as well as associated the manual annotation method. In addition, we propose an adapted U-Net architecture for automatic RV contours segmentation. This approach ensures high accuracy while minimizing the manual evaluation time for this cavity.