Wearable Single-Lead ECGs for Large-Scale Screening of Valve Disease: Feasibility and Diagnostic Accuracy

Abstract

Background: Valvular heart disease (VHD) is one of the most challenging disorders that cannot be noticed early, and only after the structural damage has taken place, the signs of it will be disclosed. Wearable one-lead ECG technologies offer a population scale, noninvasive screening device though their capability to diagnose VHD has not been established yet.

Objective:To evaluate the feasibility and diagnostic accuracy of wearable single-lead ECGs when detected in large (community based) cohorts in predicting clinically significant valve disease.

Method:It was a multi-centered, prospect trial carried out on patients with an age of 4085 years subjected to 14-day ECG wearable and transthoracic echocardiography as a gold standard of reference. Deep-learning models were conditioned on the patterns informed about aortic stenosis (AS), aortic regurgitation (AR) and mitral regurgitation (MR). Such measurements as compliance of the devices, quality of signals, and percentage of recording time using devices that had been analysed were considered feasibility measures. Sensitivity, specificity, AUC, and positive predictive value were the measures of diagnostics.

Results:Compliance (>92%), as well as overall monitoring time (87% of total time) was found to provide analyzable data in the ECG. The AUCs of moderate-severe AS, MR, and AR, 0.89, 0.84, and 0.81, respectively, were found using the AI-ECG model. Their array of sensibilities was 72-84 and specificity was 80-91. The signs of AS might be observed to 9 months prior to echocardiographic diagnosis.

Conclusion:Possible and rather effective solution to valve disease screening on large-scale with scalable and wearable single-lead ECGs and AI data analysis will allow one to diagnose it at earlier stages and sent patients through the designated channels.