A deep learning based approach for automatic cardiac events identification

Yuanshu Li, Kexin Hong, Xiaohu Shi, Wei Pang, Yubin Xiao, Peng Zhao, Dong Xu, Chunli Song, Xu Zhou, You Zhou

Abstract: Highlights•This study represents the effort to identify ED (end-diastolic) and ES (end-systolic) frames from 2D echocardiographic video.•In this work, we proposed the first classification-based method, which transforms a continuous problem into a simple binary classification problem.•The classification-based method can perform well on small datasets and it is easy to train, instead of using large and ideal datasets.•The classification-based method predicted A4C and A2C with the accuracy of above 95% and the AUCs of above 0.99.