An AI-augmented ECG score independently predicted clinical and subclinical atrial fibrillation detection in patients with implantable cardiac monitors (HR 3.56; 95% CI 1.39-9.10; p=0.008).
Cohort (n=256)
Yes
Does AI-augmented ECG analysis predict clinical and subclinical atrial fibrillation in patients with implantable cardiac monitors?
AI-augmented 12-lead ECG analysis can independently predict the development of clinical and subclinical atrial fibrillation in patients with implantable cardiac monitors, providing a noninvasive tool for risk stratification.
Hazard Ratio: 3.56 (95% CI 1.39–9.1)
p-value: p=0.008
Abstract Background Implantable cardiac monitors (ICMs) are essential for long-term atrial fibrillation (AF) surveillance in patients with embolic stroke of undetermined source (ESUS). However, early identification of individuals likely to develop clinical or subclinical AF remains challenging. Artificial intelligence (AI) applied to standard 12-lead electrocardiograms (ECGs) may provide a novel, noninvasive tool for pre-emptive AF risk prediction. Purpose To evaluate whether AI-augmented ECG analysis can predict subsequent AF detection, including subclinical AF, during long-term ICM follow-up. Methods We prospectively enrolled 256 patients who underwent ICM implantation across four tertiary centers between September 2016 and January 2024. Baseline sinus rhythm 12-lead ECGs were analyzed using a validated AI algorithm to estimate AF probability. Patients were stratified into high-risk (n = 91) and low-risk (n = 165) groups based on the AI-ECG score. Follow-up device data were reviewed every three months for AF detection, including subclinical AF events. Kaplan–Meier survival and multivariate Cox regression analyses were performed to evaluate predictors of AF incidence. Results The AI-derived AF-probability scores were significantly lower in patients who developed AF compared with those who did not (p = 0.021; Figure 1). During a mean follow-up of 31.3 ± 15.9 months, AF (including subclinical AF) was detected in 28 patients (10.9%). Kaplan–Meier analysis demonstrated a significantly higher AF incidence in the AI-predicted high-risk group (log-rank p 0.0001; Figure 2). Multivariate Cox regression identified the AI-ECG score (HR 3.56, 95% CI 1.39–9.10, p = 0.008), history of stroke (HR 6.53, 95% CI 2.86–14.89, p 0.05), and age (HR 1.04, 95% CI 1.00–1.08, p = 0.04) as independent predictors of AF detection. Conclusion This multicenter prospective study demonstrates that AI-augmented ECG analysis can predict both clinical and subclinical AF in patients with ICMs. The AI-ECG model offers a noninvasive, cost-effective tool for early risk stratification and may guide personalized surveillance and anticoagulation strategies in ESUS and other high-risk populations.
Baek et al. (Mon,) conducted a cohort in Patients with implantable cardiac monitors (ICMs) (n=256). High-risk AI-ECG score vs. Low-risk AI-ECG score was evaluated on Atrial fibrillation detection (including subclinical AF) (HR 3.56, 95% CI 1.39-9.10, p=0.008). An AI-augmented ECG score independently predicted clinical and subclinical atrial fibrillation detection in patients with implantable cardiac monitors (HR 3.56; 95% CI 1.39-9.10; p=0.008).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: