Cardiologent AI System Aids Patient-Level Arrhythmia Decisions

Current AI approaches for cardiac rhythm analysis often stop at identifying an arrhythmia from a single recording, failing to provide a comprehensive patient-level assessment or guide clinical decisions. The significance of an arrhythmia, such as atrial fibrillation, varies greatly depending on the patient's overall health and history, necessitating a more holistic approach to diagnosis and management. To address this gap, researchers have developed Cardiologent, a multi-agent system that offers patient-level arrhythmia decision support, spanning from detection to clinical decision-making. This system employs individual agents for each signal source, such as a single ECG lead and photoplethysmogram data from wearables. These agents ground their readings in measured features rather than simple labels. Cardiologent then assembles these readings into a patient's rhythm profile, combining it with the patient's personal data. It reasons against retrieved clinical guidelines for the specific case, with a critic agent verifying each conclusion against the cited guideline. The system was evaluated on integrated diagnosis, clinical significance, urgency, and management, outperforming existing methods. Crucially, its conclusions are traceable to cited guidelines and validated by expert cardiologists, offering auditable decisions for clinicians rather than blind recommendations, paving the way for use in continuous monitoring.