Deepaware: A hybrid deep learning and context-aware heuristics-based model for atrial fibrillation detection

• DOI: 10.1016/j.cmpb.2022.106899
• Publication Date: 2022-05-19T17:28:01Z
• Authors: Devender Kumar; Abdolrahman Peima...; Kamal Sharma; Helena Domínguez; Sadasivan Puthuss...; Jakob E. Bardram
• Keywords: Benchmark (surveying)

State-of-the-art automatic atrial fibrillation (AF) detection models trained on RR-interval (RRI) features generally produce high performance standard benchmark electrocardiogram (ECG) AF datasets. These models, however, result in a significantly false positive rates (FPRs) when applied ECG data collected under free-living ambulatory conditions and the presence of non-AF arrhythmias.This paper proposes DeepAware, novel hybrid model combining deep learning (DL) context-aware heuristics (CAH), which reduces FPR effectively improves participant-operated from conditions. It exploits RRI P-wave features, as well contextual ECG.DeepAware is shown to be very generalizable superior state-of-the-art unseen Most importantly, able detect efficiently recordings has achieved sensitivity (Se), specificity (Sp), accuracy (Acc) 97.94%, 98.39%, 98.06%, respectively. Results also demonstrate effect activity analysis (via P-waves detection) CAH reducing over features-based model.The proposed DeepAware can substantially reduce physician's workload manually reviewing positives (FPs) facilitate long-term monitoring for early AF.