Self-supervised ECG pre-training. (September 2021)

• Record Type: Journal Article
• Authors: Liu, Han; Zhao, Zhenbo; She, Qiang
• Is Part Of: Biomedical signal processing and control. Volume 70(2021)
• Publication Date: 2021-09
• Subjects: Electrocardiogram -- Deep learning -- Self-supervised learning; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103010
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract: Background: Real-world medical data, such as electrocardiogram (ECG), often show a long-tail distribution and severe category imbalance, and severely imbalanced data generate bias in deep learning models. In this work, we investigate how to alleviate the problems of label imbalance and inadequate labelling faced by deep learning models when applied to ECG data. Methods: We constructed a short-duration twelve-lead ECG dataset, containing more than 300, 000 samples, for morphological recognition based on the actual distribution to evaluate and compare the recognition ability of humans and computers regarding ECG morphology. Two unique ECG data augmentation methods were designed and were combined with a variety of current mainstream self-supervised learning methods, and ultimately, the pre-trained weights were transferred to an 8-class multi-label ECG classification task for evaluation. Results: The experiments showed that self-supervised pre-training relying on negative sample pairs could achieve significantly better ECG representation than baseline, which was significantly effective for alleviating the imbalance in ECG data and reducing the labels of supervised samples. This method effectively utilized a large number of normal ECG samples. Additionally, with the diagnosis of the expert team as ground truth, under the condition of accessing only a small number of labelled samples, these models even performed better than the human ECG doctors participating in the test. Conclusion: The combination of self-supervised learning and unique data augmentation methods in the recognition of ECG morphology can effectively alleviate the long-tail problem and severe data imbalance and can significantly reduce the need for labelled samples in the downstream task. Highlights: Designed two data augmentation methods unique to ECG contrastive learning. Assessed the performance of mainstream self-supervised learning methods on ECG data. An Image dataset of over 300, 000 short-duration 12-lead ECG samples was exposed.