Artificial intelligence for atrial fibrillation prediction while in sinus rhythm: a narrative review

Atrial fibrillation (AF) is a prevalent cardiac arrhythmia affecting over 50 million individuals worldwide, with serious complications including stroke and heart failure. Detecting AF in patients who present with normal sinus rhythm (SR) is still an open problem. This narrative review aims to explore the increasing role of artificial intelligence (AI) in predicting AF from SR based on different event horizons, highlighting advancements and ongoing challenges in the field. We examined a variety of AI methodologies divided into four main time frames: newly diagnosed AF, short-term (minutes-hours), mid-term (days-months), and long-term (years) prediction. The studies included in the review showcase applying AI models to electrocardiograms or RR intervals, clinical features, and other metadata. The AI methods outperformed the scoring systems and showed a higher accuracy in predicting AF from SR when the event horizon was shorter. AI models demonstrated promising capabilities in predicting AF from SR. However, the deployment of AI in clinical settings is not without challenges. Issues such as algorithm interpretability, the requirement for extensive clinical validation, and integration into existing healthcare frameworks remain significant barriers. Moreover, ethical considerations and data privacy concerns must be addressed to ensure the responsible use of AI technologies in healthcare. Future research should focus on enhancing the interpretability of AI models, improving the robustness of predictions through diverse datasets, and conducting comprehensive clinical trials to validate AI tools. By addressing these challenges, AI can play a critical role in advancing AF management and improving patient outcomes.