A paradigm shift toward full-cycle management of atrial fibrillation: integrating digital twins and artificial intelligence.

Background: Clinical management of atrial fibrillation (AF) often remains fragmented, with decisions largely based on information available at a single time point rather than on the continuous evolution of the disease. Emerging technologies such as digital twins and artificial intelligence (AI) offer new possibilities to change this paradigm by enabling dynamic, patient specific modelling and data driven decision support across the full disease cycle. Objective: This review comprehensively examines the technical pathways and clinical value of integrating digital twins and artificial intelligence across the full cycle of atrial fibrillation, and explores the potential impact of this integration on diagnostic and treatment paradigms. In the present review, the concept of virtual closed-loop AF management encompasses both: (1) real-time or near-real-time decision support during in-hospital electrophysiological procedures; (2) longitudinal post-recording analysis based on wearable devices, ambulatory ECG monitoring, imaging datasets, and electronic health records. Therefore, the proposed framework spans outpatient screening, in-hospital intervention planning, intra-procedural guidance, and post-procedural follow-up management. Methods: A structured narrative literature review was conducted to summarize recent advances in digital twin technology and artificial intelligence for atrial fibrillation management. Particular attention was given to multimodal data integration, patient-specific modeling, in silico simulation, and AI-assisted clinical decision support. Results: A relatively complete technical chain for constructing patient-specific cardiac digital twins has been established, achieving anatomical Dice coefficients of 93% or higher and a correlation coefficient for activation time prediction exceeding 0.96. Artificial intelligence permeates the entire process of AF management: the use of AI-ECG increased AF detection rates by 2.3-fold, models for predicting post-ablation recurrence achieved area under the curve (AUC) values generally ranging from 0.72 to 0.85, and intra-operative three-dimensional reconstruction time was reduced to 65 s. The integration of the virtual closed-loop framework: "data sensing → model construction → in silico testing → clinical decision → feedback optimisation" has been preliminarily validated in scenarios such as drug screening, ablation planning, and thrombus risk assessment. Conclusion: The integration of digital twin and artificial intelligence provides a novel pathway for full-cycle atrial fibrillation management, and holds the potential to shift the current diagnostic and therapeutic paradigm from fragmentation towards continuity. The first prospective multicentre randomised controlled trial has provided the first prospective evidence supporting Al-assisted personalized AF management; however, its clinical application still relies on further validation through multicentre studies and high-quality evidence. As technology matures and evidence accumulates, this model is expected to be gradually introduced into clinical practice.