Evaluating AI-Enabled Technologies for Atrial Fibrillation Detection: A Systematic Review of Diagnostic Performance

Area of Science:

• Cardiology
• Digital Health
• Artificial Intelligence

Background:

• Atrial fibrillation (AF) is a common arrhythmia linked to stroke and cardiovascular issues.
• Current AF detection methods have limitations in accessibility and monitoring duration.
• AI and wearable technologies offer new avenues for early AF detection.

Purpose of the Study:

• To systematically review the diagnostic performance of AI-enabled and wearable technologies for AF detection.
• To evaluate these technologies in both clinical and real-world settings.

Main Methods:

• Comprehensive literature search of major databases (PubMed, Scopus, Web of Science).
• Inclusion of studies reporting sensitivity, specificity, predictive values, and AUC for digital AF detection tools.
• Data extraction and quality assessment using QUADAS-2, followed by random-effects meta-analyses.

Main Results:

• Twenty-four studies were included, with high-performing tools achieving sensitivity and specificity ≥94%.
• Consumer-grade devices showed lower specificity (46%) and positive predictive value (7.6%), indicating frequent false positives.
• Heterogeneity was observed, influenced by device type, signal source (PPG vs. ECG), algorithm, and population characteristics.

Conclusions:

• AI-enhanced and wearable technologies demonstrate significant potential for AF detection under specific conditions.
• Performance variability, particularly in consumer devices, highlights the need for external validation and algorithm refinement.
• Future research should focus on real-world effectiveness, cost-efficiency, and explainability for broader clinical adoption.