A Joint Optimization Guided Deep Learning Model based on CNN and Channel-Wise Transformers for Robust Sleep Stage...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Accurate sleep stage classification is vital for diagnosing sleep disorders.
• Manual scoring of polysomnography (PSG) data is laborious and error-prone.
• Current deep learning models face challenges with computational expense and real-time application.

Purpose of the Study:

• To develop a lightweight and efficient dual-branch deep learning model for automated sleep stage classification.
• To address limitations of existing models, including computational complexity and feature extraction.

Main Methods:

• A novel dual-branch deep learning architecture combining Convolutional Neural Networks (CNNs) and Transformers.
• Utilizes channel-wise attention mechanisms to capture signal dependencies efficiently.
• Model validated on four benchmark datasets: SleepEDF-20, SleepEDF-78, SleepEDFx, and SHHS.

Main Results:

• The proposed model demonstrates superior performance compared to baseline algorithms across all datasets.
• Achieved state-of-the-art results, indicating robustness and scalability.
• The model is computationally efficient and suitable for real-time applications.

Conclusions:

• The dual-branch CNN-Transformer model offers an accurate and efficient solution for sleep stage classification.
• This approach facilitates improved diagnosis and treatment of sleep disorders.
• Publicly available code promotes reproducibility and further research.