Real-world Clinical Evaluation of an AI-enabled Atrial Fibrillation Detection System Using 7-day Holter ECG

Abstract

We have started using the Atrial Fibrillation Early Detection Support Service（AF Detector）, which incorporates a patch-type, 7-day, single-lead electrocardiogram（ECG recorder Heartnote^TM, JSR Corporation）and AI-enabled ECG analysis software（SmartRobin）after 2023. This study assesses SmartRobin’s performance and utility in detecting atrial fibrillation（AF）within real-world clinical settings. This retrospective study evaluated 100 consecutive patients who underwent AF detection using the AF Detector at our hospital from January to August 2023. To assess SmartRobin’s accuracy, an independent arrhythmia specialist confirmed AF diagnoses. The specialist analysis revealed that SmartRobin’s AF detection achieved the sensitivity of 100％ and the specificity of 69％ at the patient level. The critical need to ensure AF detection underlines SmartRobin’s significant utility as a diagnostic support system. Its integration with patch ECGs, noted for high patient adherence, brings its broader adoption. Improving specificity in SmartRobin’s decision is a challenge for future development.