Abstract 16791: Preliminary Evaluation of a Deep Learning Approach for Echocardiographic Screening for Rheumatic Heart Disease

Lindsay A Edwards,Fei Feng,Mehreen Iqbal,Yong Fu,Amy Sanyahumbi,Shiying Hao,Doff B McElhinney,Xuefeng B Ling,Jiajia Luo

DOI: https://doi.org/10.1161/circ.138.suppl_1.16791

IF: 37.8

2018-11-06

Circulation

Abstract:Introduction: Echocardiography-based screening for rheumatic heart disease (RHD) in asymptomatic children in resource-limited settings results in early diagnosis, which, in combination with effective secondary prevention, may reduce morbidity and mortality. Physician-led screening, however, is not feasible in much of the developing world. Automated echocardiographic diagnosis via a novel deep learning approach may obviate the need for a physician, enabling more widespread echocardiographic screening, early diagnosis, and improved outcomes. In this study, we hypothesized that a deep learning model is capable of identifying mitral regurgitation (MR) in appropriate views. Methods: Echocardiogram clips were labeled by clip for view and presence of MR. Using the computational ability of a Nvidia GTX1080TI graphics processing unit, training dataset clips were used to build two convolutional neural networks to perform the stepwise tasks of classifying the clips 1) by view and 2) by the presence/absence of MR in appropriate color Doppler views. We trained the views classification model using 1857 clips of all views and the MR detection model using 686 parasternal long axis, parasternal short axis, and apical four chamber color Doppler clips. We then evaluated the performance of our models using a test dataset. Results are reported as an F1 score (an F1 score of 1 suggests perfect accuracy of the model). Results: Figure 1. a) View Classification Model Learning Curve, b) MR Detection Model Learning Curve Discussion: Our results suggest that a deep learning model is capable of discerning MR on echocardiogram clips. This is an encouraging step toward automated diagnosis of RHD, which will enable less skilled echocardiogram operators to make a real-time diagnosis of RHD.

cardiac & cardiovascular systems,peripheral vascular disease