Self-supervised inter-intra period-aware ECG representation learning for detecting atrial fibrillation

Xiangqian Zhu, Mengnan Shi, Xuexin Yu, Chang Liu, Xiaocong Lian, Jintao Fei, Jiangying Luo, Xin Jin, Ping Zhang, Xiangyang Ji

Abstract: Highlights•Innovative Self-Supervised Learning: We introduce a novel self-supervised ECG representation learning method, specifically tailored for detecting atrial fibrillation (AF).•Superior Performance: The method achieves outstanding AUC performance for AF detection on the BTCH dataset, reaching 0.953 for paroxysmal AF and 0.996 for persistent AF.•Scalability: Experiments validate the model’s exceptional performance in setups with partial leads, highlighting its potential for wearable devices and large-scale health monitoring.•Clinical Relevance: Integrating medical knowledge not only boosts model performance but also enhances interpretability, making the model more clinically viable and acceptable.