A multicenter, randomized, controlled study to evaluate the efficacy of an artificial intelligence-based electrocardiography for the detection of atrial fibrillation in patients with suspected symptoms (DEEP-AF)

Background Atrial fibrillation (AF) is frequently paroxysmal and asymptomatic, leading to delayed diagnosis and missed opportunities for stroke prevention. Artificial intelligence (AI) applied to sinus rhythm 12-lead electrocardiography (ECG) may enable risk enrichment and more efficient allocation of rhythm-monitoring resources. Objective To determine whether an AI-ECG-guided strategy improves detection of newly diagnosed AF compared with usual care in patients with symptoms suggestive of AF and no prior AF diagnosis. Methods DEEP-AF is a prospective, multicenter, randomized, controlled superiority trial conducted at 13 tertiary hospitals in South Korea. A total of 1,230 adults aged 30 years or older with symptoms suggestive of AF and no AF on screening 12-lead ECG will be randomized 1:1 to an AI-ECG–guided strategy or usual care. In the AI-ECG group, the SmartECG-AFrisk algorithm will generate a risk score (high risk ≥50; low risk <50) to support clinician-directed rhythm monitoring within guideline-concordant practice. Usual care will follow standard clinical evaluation without access to the AI score. Results The primary outcome is the incidence of newly diagnosed AF within 6 months, confirmed on standard ECG modalities. Secondary outcomes include time to AF diagnosis, AF detection according to AI risk strata, healthcare resource utilization and diagnostic efficiency, initiation and timing of anticoagulation or antiplatelet therapy, AF burden, detection of other clinically relevant arrhythmias, arrhythmia-related procedures, clinical events, and patient satisfaction. Conclusion DEEP-AF will provide pragmatic randomized evidence on whether AI-ECG–guided risk enrichment can improve AF detection and diagnostic efficiency in routine care.