Synthetic generation of cardiac tissue motion from surface electrocardiograms.

Abstract

Cardiac tissue motion is a sensitive biomarker for detecting early myocardial damage. Here, we show the similarity, interpretability and diagnostic accuracy of synthetic tissue Doppler imaging (TDI) waveforms generated from surface electrocardiograms (ECGs). Prospectively collected ECG and TDI data were cross-matched as 9,144 lateral and 8,722 septal TDI-ECG pairs (463 patients) for generating synthetic TDI across every 1% interval of the cardiac cycle. External validation using 816 lateral and 869 septal TDI-ECG pairs (314 patients) demonstrated strong correlation (repeated-measures r = 0.90, P < 0.0001), cosine similarity (0.89, P < 0.0001) and no differences during a randomized visual Turing test. Synthetic TDI correlated with clinical parameters (585 patients) and detected diastolic and systolic dysfunction with an area under the curve of 0.80 and 0.81, respectively. Furthermore, synthetic TDI systolic and early diastolic measurements generated from an external ECG dataset (233,647 patients) were associated with all-cause mortality during both sinus rhythm and atrial fibrillation, underscoring their potential for personalized cardiac care.