Related Experiment Video
Updated: Jul 27, 2026
06:37
Artificial Intelligence-Based System for Detecting Attention Levels in Students
Published on: December 15, 2023
[A lightweight classification network for single-lead atrial fibrillation based on depthwise separable convolution
Yong Hong1, Xin Zhang1, Mingjun Lin1
1College of Biomedical Engineering, Southern Medical University, Guangzhou 510515, China.
Summary
A novel deep learning model, DSC-AttNet, efficiently diagnoses atrial fibrillation using ECG data. This lightweight model achieves high accuracy and precision, outperforming existing methods for wearable devices.
Area of Science:
- Cardiology
- Artificial Intelligence
- Biomedical Engineering
Context:
- Atrial fibrillation (AF) diagnosis is crucial for preventing stroke.
- Wearable ECG devices offer continuous monitoring but face limitations in computational power.
- Existing diagnostic models often lack the efficiency required for real-time analysis on resource-constrained devices.
Purpose:
- To develop a deep learning model (DSC-AttNet) for automated atrial fibrillation detection.
- To optimize the model for integration into wearable ECG monitoring devices by balancing complexity and performance.
- To enhance feature extraction and classification accuracy using depthwise separable convolution and channel-spatial attention.
Summary:
- A lightweight attention network, DSC-AttNet, was designed using depthwise separable convolution and channel-spatial attention fusion.
- The model was trained and validated on publicly available ECG datasets (LTAFDB, AFDB, NSRDB).
- DSC-AttNet achieved 97.33% average accuracy in 10-fold cross-validation and 92.78% on an external test set, outperforming comparison models with fewer parameters (1.01M) and lower computational volume (27.19G).
Impact:
- The proposed DSC-AttNet demonstrates superior classification performance and generalization ability for atrial fibrillation.
- Its reduced complexity and high efficiency make it suitable for deployment on wearable ECG devices.
- This facilitates automated, real-time AF diagnosis, potentially improving patient outcomes and reducing healthcare burdens.
