Prediction of new-onset atrial fibrillation in patients with hypertrophic cardiomyopathy using artificial intelligence-enhanced electrocardiogram

Background Atrial fibrillation (AF) occurs in around 25% of patients with hypertrophic cardiomyopathy (HCM), elevating their risk of thromboembolism and exacerbating symptoms. There lack predictive models for development of AF in individuals with HCM. Purpose We sought to develop artificial intelligence-enhanced electrocardiogram (AI-ECG) models to predict new-onset AF in patients with HCM. Methods The AI-ECG models were developed using a dataset of 3,068 patients with HCM who had no previous history of AF. Patients were divided into non-overlapping data sets: 70% for training, 10% for validation, and 20% for testing. The AI-ECG models employed a residual convolutional neural network (CNN). The primary outcome was incident AF at 5 years. Results The AI-ECG model for the primary outcome achieved an area under receiver characteristics curve (AUC-ROC) of 0.828 (95%CI: 0.786-0.865). Integration of age and sex with the ECG-AI model as a logistic regression model achieved an AUC-ROC of 0.845 (95%CI: 0.806-0.879). Patients stratified based on the AI-ECG model output demonstrated significant difference in risk of developing AF during follow-up (P<0.001). Conclusions AI-ECG may function as a simple and efficient tool for predicting new-onset atrial fibrillation in patients with HCM to enable prompt identification and intervention.