Diagnostic accuracy of ai-enabled wearables for detecting atrial fibrillation recurrence after catheter ablation: a systematic review and meta-analysis
J Giorgi, R Huntermann, P G Batista, M E Molinari, J Peres de Oliveira, R R A Santos Silva, M V Montenegro, L A Lucena, G C Gaelzer, M Malik, C Fischer BaccaAbstract
Introduction
Artificial intelligence (AI)–enabled monitoring has emerged as a promising approach for detecting atrial fibrillation (AF) recurrence after catheter ablation, providing continuous or high-frequency rhythm assessment beyond traditional ECG- and Holter-based follow-up. However, the diagnostic performance of AI-driven wearables and handheld devices in this context has not been systematically evaluated.
Purpose
We performed a systematic review and meta-analysis to quantify the diagnostic accuracy of AI-based and device-enabled monitoring strategies for identifying post-ablation AF recurrence.
Methods
A systematic review and meta-analysis were conducted in accordance with PRISMA guidelines. Four studies comprising 333 participants were included, evaluating handheld single-lead ECG systems with AI algorithms, deep learning–enabled symptom-triggered recorders, implantable loop recorders, and wearable photoplethysmography-based approaches, using conventional ECG or continuous monitoring comparators. Pooled diagnostic performance was assessed using random-effects models, univariate sensitivity and specificity analyses with 95% confidence intervals (CIs), and a bivariate SROC model were applied.
Results
The overall sensitivity was 85.9% (95% CI 76.1–92.1) and specificity was 98.8% (95% CI 84.4–99.9). The pooled positive predictive value was 98.1% (95% CI 81.3–99.8), while the negative predictive value was 89.9% (95% CI 81.3–94.8). The summary area under the receiver operating characteristic curve was 0.93 (95% CI 0.84–0.98), indicating high overall discriminative ability. Substantial heterogeneity was observed across studies, reflecting differences in monitoring modality, patient populations, and recurrence definitions.
Conclusion
In this meta-analysis of four studies, AI- and device-enabled monitoring approaches showed good overall diagnostic performance for detecting atrial fibrillation recurrence after catheter ablation, with particularly high specificity. Although these findings suggest potential utility as adjunctive tools within post-ablation follow-up, the limited evidence base and between-study heterogeneity highlight the need for larger prospective studies to clarify their role in contemporary surveillance pathways.Graphical Abstract