Using AI-ECG to Stratify Long-Term Mortality Risk and Prognosis in TAVR Patients

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Long-term mortality after transcatheter aortic valve replacement (TAVR) remains high.
• Conventional risk models inadequately capture subclinical electrophysiological changes predicting prognosis.
• Electrocardiograms (ECGs) offer a readily available source of prognostic information.

Purpose of the Study:

• To develop and validate an AI-ECG model for predicting long-term mortality in TAVR patients.
• To assess the AI-ECG model's performance in independent internal and external validation cohorts.
• To evaluate the clinical utility of AI-ECG for risk stratification.

Main Methods:

• A cohort of 711 patients undergoing TAVR was analyzed.
• Preoperative ECGs were processed using a Residual Network-18 model for risk stratification.
• The primary endpoint was 3-year all-cause mortality, with validation across two centers.

Main Results:

• The AI-ECG model showed comparable discrimination in internal (AUC 0.767) and external (AUC 0.712) validation cohorts.
• High-risk patients identified by AI-ECG had significantly higher 3-year mortality (61.5%) compared to low-risk patients (16.5%).
• AI-ECG risk classification independently predicted mortality (aHR 3.49) and demonstrated clinical net benefit.

Conclusions:

• The AI-ECG model provides accurate, noninvasive long-term risk stratification for TAVR patients.
• This AI tool has significant potential for personalized follow-up and management strategies.
• AI-ECG enhances prognostic assessment beyond traditional risk factors.