Back

Deep Learning Prediction of Left Atrial Structure and Function from 12-lead Electrocardiograms

Brody, J. A.; Yogeswaran, V.; Wiggins, K. L.; Sitlani, C. M.; Bis, J. C.; Chen, L. Y.; Heckbert, S. R.; Lima, J. A. C.; Longstreth, W. T.; Soliman, E. Z.; Tison, G. H.; Ye, T.; Psaty, B. M.; Shojaie, A.; Floyd, J. S.

2025-09-30 cardiovascular medicine
10.1101/2025.09.29.25336490 medRxiv
Show abstract

Abnormal cardiac atrial structure and function (atrial cardiopathy)1 typically precedes atrial fibrillation (AF) and predicts other cardiovascular complications, yet detection is limited by the cost and limited accessibility of high-quality cardiac imaging. We trained a deep learning model (ECG-AI) using 12-lead electrocardiograms paired with 21,749 cardiac magnetic resonance scans to predict left atrial structure and function. ECG-AI measures of atrial cardiopathy were strongly associated with new-onset AF, heart failure, and ischemic stroke after adjustment for clinical risk factors and biomarkers in two external cohorts, outperforming imaging measures and clinical risk factors. The risk of cardioembolic stroke, the hallmark complication of AF, was 66% greater per SD left atrial volume. In a screening population, ECG-AI predicted subclinical AF from 14-day cardiac monitoring better than a clinical risk prediction tool. Our ECG-AI is an inexpensive, accessible tool that identifies individuals at high-risk for AF and related complications.

Published in Nature Communications (predicted rank #4) · training set

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.