Semi-supervised learning for ECG classification without patient-specific labeled data
作者：
Highlights：
• A high performance semi-supervised ECG classification system is presented.
• Including some patient-specific N beats for training is a practical and effective way to improve performance.
• An unsupervised method is designed to estimate these patient-specific N beats for training.
• A semi-supervised iterative label update method is designed to further improve the ECG classification performance.
• Sen and Spe for S and V beats prediction are comparable with several supervised methods.
摘要
•A high performance semi-supervised ECG classification system is presented.•Including some patient-specific N beats for training is a practical and effective way to improve performance.•An unsupervised method is designed to estimate these patient-specific N beats for training.•A semi-supervised iterative label update method is designed to further improve the ECG classification performance.•Sen and Spe for S and V beats prediction are comparable with several supervised methods.
论文关键词：Semi-supervised learning,Arrhythmia,CNN,ECG classification,Time series signal
论文评审过程：Received 27 August 2019, Revised 28 February 2020, Accepted 25 March 2020, Available online 7 May 2020, Version of Record 21 May 2020.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.113411
