Myocardial Function Imaging in Echocardiography Using Deep Learning

Deformation imaging in echocardiography has been shown to have better diagnostic and prognostic value than conventional anatomical measures such as ejection fraction. However, despite clinical availability demonstrated efficacy, everyday use remains limited at many hospitals. The reasons are complex, but practical robustness questioned, a large inter-vendor variability demonstrated. In this work, we propose novel deep learning based framework for motion estimation echocardiography, fully automate myocardial function imaging. A estimator was developed on PWC-Net architecture, which achieved an average end point error of (0.06±0.04) mm per frame using simulated data from open access database, par or compared previously reported state the art. We further demonstrate unique adaptability image artifacts signal dropouts, made possible trained models that incorporate relevant augmentations. Further, automatic pipeline consisting cardiac view classification, event detection, segmentation used estimate left ventricular longitudinal strain vivo. method showed promise by achieving mean deviation (-0.7±1.6)% semi-automatic commercial solution N=30 patients with disease, within expected limits agreement. thus believe learning-based can facilitate extended practice.