An investigation of the contextual distribution of false positives in a deep learning-based atrial fibrillation detection algorithm
作者：
Highlights：
• Showed how an AF detection model results in high FPR under free-living conditions.
• Investigated the influence of user’s ambulatory contexts on FPR on 215 days long ECG.
• Sudden Acceleration, activity & body position changes cause 78% of short FP segments.
• True positive AF segments were clustered around the morning and late evening hours.
• Provided the implications of context-awareness for improving ambulatory AF detection.
摘要
•Showed how an AF detection model results in high FPR under free-living conditions.•Investigated the influence of user’s ambulatory contexts on FPR on 215 days long ECG.•Sudden Acceleration, activity & body position changes cause 78% of short FP segments.•True positive AF segments were clustered around the morning and late evening hours.•Provided the implications of context-awareness for improving ambulatory AF detection.
论文关键词：Atrial fibrillation (AF),Electrocardiogram (ECG),Context-aware ECG,Deep learning (DL),False positive (FP),Arrhythmias
论文评审过程：Received 12 May 2021, Revised 13 May 2022, Accepted 12 August 2022, Available online 18 August 2022, Version of Record 26 August 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.118540
