Advances in AI-driven prediction of atrial fibrillation in hypertrophic cardiomyopathy: a systematic review

Traditional atrial fibrillation (AF) risk stratification models rely on a limited number of clinical variables and inadequately capture the complex pathophysiology of hypertrophic cardiomyopathy (HCM). Recent advances in artificial intelligence (AI) offer the potential to integrate large-scale multimodal data to improve individualized AF risk prediction. This systematic review aimed to evaluate the methodologies, performance, and clinical validity of AI-driven models for predicting AF in patients with HCM. This systematic review was conducted in accordance with PRISMA guidelines. PubMed, Embase, Scopus, CINAHL, and Web of Science were searched for studies evaluating AI-based prediction of AF in HCM from inception to December 2025. Eligible studies reported dataset characteristics, input data modalities, model type, and validation strategy. Two reviewers independently screened studies, extracted data, and assessed risk of bias using the Prediction model Risk Of Bias ASsessment Tool (PROBAST). Of 470 records identified, five observational cohort studies published between 2021 and 2025 met the inclusion criteria. Sample sizes ranged from several hundred to over 1,000 adult patients with HCM. AI models incorporated heterogeneous data modalities, including clinical variables, cardiovascular magnetic resonance (CMR) parameters, plasma biomarkers, and high-dimensional proteomics. Model discrimination ranged from moderate to high, with reported AUCs of 0.80–0.89. Multicentre studies with external validation demonstrated superior generalizability and consistently outperformed traditional risk scores. PROBAST assessment showed an overall low risk of bias. AI-driven models demonstrate promising performance for predicting incident AF in HCM, particularly when leveraging multimodal imaging and molecular data with external validation.