.png){kind=icon}
.png){kind=icon}
Return
A Task-Generic High-Performance Unsupervised Pre-Training Framework for ECG
.png){kind=icon}
OA
AI
DOI：10.1109/JSEN.2024.3404141.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Electrocardiogram (ECG) feature extraction is an important step in applying machine learning for biomedical processing tasks. In deep learning studies centered on ECG data, established methods mainly rely on labeled data to automatically extract signal features. However, obtaining high-quality labeled records from open-source datasets is challenging, whereas many unlabeled data remain underused. Unsupervised pre-training methods have shown significant efficacy in using unlabeled datasets. Therefore, using unsupervised pre-training paradigms for ECG feature extraction could improve overall performance. This article presents an unsupervised pre-training framework based on a masked autoencoder (MAE) for extracting ECG signal features. An encoder-decoder framework is proposed to recover artificially masked ECGs. During the reconstruction process, the framework learns features from masked samples. Benefiting from the utilization of large-scale datasets, the framework enables learning generalized features and has a high performance for a wide range of deep learning tasks. In particular, the framework achieves an accuracy of 95.6% on the MITDB dataset for the ECG arrhythmia classification task and 98.8% on the ECGIDDB dataset for the human identification task. Evaluation using multiple downstream tasks and comparisons with the state-of-the-art confirm the validity of our proposed approach.
Keywords:
Electrocardiogram (ECG)
masked autoencoder (MAE)
unlabeled data
unsupervised pre-training