AI-Based Prediction of Atrial Fibrillation in ESUS Patients With ICM

Embolic Stroke of Undetermined Source (ESUS) accounts for a significant proportion of ischemic strokes, and occult Atrial Fibrillation (AF) is considered a major etiology. While Implantable Cardiac Monitors (ICMs) are the gold standard for long-term rhythm monitoring, identifying patients at the highest risk for AF remains a clinical challenge.

This multicenter, prospective study aims to validate the clinical utility of an artificial intelligence-based electrocardiogram analysis algorithm, "SmartECG-AF," in this specific population. The algorithm analyzes 12-lead ECGs recorded during sinus rhythm to detect subtle signs of electrical remodeling associated with paroxysmal AF.

Enrolled patients with ESUS who have undergone ICM implantation will have their baseline ECGs analyzed by the SmartECG-AF algorithm. Based on the AI-generated probability score, patients will be stratified into a "High Risk" group and a "Low to Intermediate Risk" group. The study will longitudinally track these patients to compare the time-to-event for ICM-detected AF between the two groups. Additionally, the study will evaluate the correlation between the AI risk score and the incidence of Major Adverse Cardiovascular Events (MACE), providing evidence for AI-guided risk stratification in cryptogenic stroke management.

Inclusion criteria

• Patients aged 30 years or older.
• Patients diagnosed with Embolic Stroke of Undetermined Source (ESUS) who have undergone or are scheduled for Implantable Cardiac Monitor (ICM) implantation.
• Patients who have undergone at least one 12-lead ECG examination within 2 weeks before or after the date of ICM implantation.
• Patients maintaining Sinus Rhythm on ECG at the time of enrollment.
• Patients who have voluntarily signed the informed consent form.

Exclusion criteria

• Patients diagnosed with Atrial Fibrillation (AF) at least once prior to the date of enrollment.
• Patients whose ICM battery status is at Elective Replacement Interval (ERI), making recording impossible.
• Patients whose ECGs cannot be analyzed by the AI algorithm (SmartECG-AF) due to severe artifacts or noise, or are incompatible with digital analysis.