Diastolic Function From Artificial Intelligence-Derived Electrocardiography in Mitral Annulus Calcification

To evaluate whether a validated artificial intelligence-enabled electrocardiography (AI-ECG) model for left ventricular diastolic function (LVDF) can provide prognostic risk stratification in patients with significant mitral annulus calcification (MAC).

We included patients who underwent transthoracic echocardiography and electrocardiography within 1 week between September 1, 2001, and April 30, 2023, and were not included in the AI-ECG LVDF training data set. Significant MAC was defined by consensus echocardiography criteria. Patients were assigned to 3 myocardial disease (MD) categories by the AI-ECG LVDF algorithm grading: normal/grade 1 LVDF (MD-1), grade 2 LVDF (MD-2), and grade 3 LVDF (MD-3). The analysis focused on prognostic stratification in this clinically distinct population. The primary end point was all-cause mortality.

Of 2506 patients with significant MAC (mean age, 77 years; 64.1% female), 709 (28.3%) were classified as MD-1, 1510 (60.3%) as MD-2, and 287 (11.5%) as MD-3. Advanced MD grade correlated with older age, male sex, lower left ventricular ejection fraction, higher rates of comorbidities, and worse LVDF echocardiographic parameters. During a median follow-up of 2.8 years, 1059 (42.3%) patients died. Multivariable survival analysis found MD-2 (adjusted hazard ratio, 1.40; 95% CI, 1.19 to 1.64) and MD-3 (adjusted hazard ratio, 1.52; 95% CI, 1.21 to 1.92) to be independently associated with higher mortality risk, also after adjustment for mitral inflow diastolic pressure gradients.

Myocardial disease grading derived from AI-ECG LVDF is independently associated with increased mortality risk in patients with significant MAC, underscoring its potential value in identifying high-risk patients in this challenging clinical population.