PSC-Net: Integration of Convolutional Neural Networks and transformers for Physiological Signal Classification
Biomedical Signal Processing and Control · 2024-05 · 91 · 106040
DOI: 10.1016/j.bspc.2024.106040 ↗Kurz gefasst
Fachartikel in der Zeitschrift Biomedical Signal Processing and Control, 2024-05. Verfasst von Qichao Liu und 10 weiteren Autor:innen. Die Arbeit wurde bisher 16-mal zitiert (Crossref) und nennt 58 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Objective
Physiological signals, such as electrocardiogram (ECG) and wrist pulse signals, play a crucial role in assessing the overall health condition of patients. Hence, accurate identification of physiological signal categories has emerged as a critical component for prompt diagnosis. However, this field still confronts several prominent challenges, including limited availability of data, imbalanced datasets, convergence issues with loss functions, and the need for model architectures capable of accurately detecting waveform patterns.
Method
This study presents a model named Physiological Signal Classification Network (PSC-Net), which integrates the benefits of convolutional neural network (CNN) and transformer in the context of medical artificial intelligence. Specifically, we adopt the GRU-LSTM network (LGNet) with causal convolution, gated convolution, gate recurrent unit (GRU), and long short-term memory (LSTM) to extract temporal features from the physiological signal. In the transformer, we leverage two layers of GRU to extract global features and introduce residual connections to incorporate frequency domain features. To overcome the challenges posed by small and imbalanced datasets, we propose a new data augmentation algorithm and a novel loss function. Addition, we employ the dung beetle algorithm for automatic learning rate optimization.
Result
Our proposed method achieves superior accuracies of 83.33%, 100.0%, 95.74%, and 98.85% on four physiological signal datasets (including one clinical dataset): five types of pulses, coronary heart disease, MIT-BIH Arrhythmia, and MIT-BIH ST Changes. Furthermore, the evaluation of diverse datasets showcases the remarkable generalization capability of the presented method, thereby validating its effectiveness across various scenarios.
Zitieren
APA
Liu, Q., Feng, Y., Xu, H., Li, J., Lin, Z., Li, S., Qiu, S., Wu, X., Ma, Y., Xu, Y., & Li, F. (2024). PSC-Net: Integration of Convolutional Neural Networks and transformers for Physiological Signal Classification. Biomedical Signal Processing and Control, 91, 106040. https://doi.org/10.1016/j.bspc.2024.106040
Vancouver
Liu Q, Feng Y, Xu H, Li J, Lin Z, Li S, et al. PSC-Net: Integration of Convolutional Neural Networks and transformers for Physiological Signal Classification. Biomedical Signal Processing and Control. 2024;91:106040. doi:10.1016/j.bspc.2024.106040
BibTeX
@article{Liu2024,
author = {Liu, Qichao and Feng, Yue and Xu, Hong and Li, Jia and Lin, Zhuosheng and Li, Shengke and Qiu, Shihan and Wu, Xin and Ma, Yuangang and Xu, Ying and Li, Fufeng},
title = {PSC-Net: Integration of Convolutional Neural Networks and transformers for Physiological Signal Classification},
journal = {Biomedical Signal Processing and Control},
year = {2024},
volume = {91},
pages = {106040},
doi = {10.1016/j.bspc.2024.106040}
}
EUVIMED