Physiological-Interaction-Aware Net for Automatic Sleep Staging from Concurrent Spatio-Time-Frequency View.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Sleep Medicine

Background:

• Accurate sleep staging is vital for monitoring sleep and diagnosing disorders.
• Physiological interactions across systems (brain, heart, respiratory) vary with sleep stages.
• Current methods often overlook these interactions, limiting classification accuracy.

Purpose of the Study:

• To develop an advanced model for automatic sleep staging.
• To incorporate physiological interactions into sleep stage classification.
• To improve the accuracy of sleep monitoring and disorder diagnosis.

Main Methods:

• Developed the Physiological-Interaction-Aware Net (PIANet) for automatic sleep staging.
• Utilized a concurrent spatio-time-frequency approach.
• Integrated functional connectivity, spatio-temporal, and time-frequency feature extraction, fusing cross-view information.

Main Results:

• PIANet demonstrated superior performance compared to existing methods on multiple datasets (ISRUC-S1, ISRUC-S3, DOD-H).
• Achieved high classification accuracies: 0.825 (ISRUC-S1), 0.836 (ISRUC-S3), and 0.891 (DOD-H).
• Obtained competitive F1 scores: 0.814 (ISRUC-S1), 0.826 (ISRUC-S3), and 0.839 (DOD-H).

Conclusions:

• The proposed PIANet effectively leverages physiological interactions for improved sleep staging.
• The model's ability to integrate diverse signal views enhances classification accuracy.
• PIANet offers a promising advancement for sleep monitoring and clinical diagnosis.