Enhancing Photoplethysmography-Based Sleep Staging Models Through Temporal Context Optimization.

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Accurate sleep stage classification is crucial for diagnosing sleep disorders and assessing sleep quality.
• Polysomnography (PSG) is the gold standard, but photoplethysmography (PPG) offers a practical alternative for wearables.
• Current methods often require long signal acquisition, limiting wearable feasibility due to energy constraints.

Purpose of the Study:

• To develop an adapted sleep staging model optimized for shorter PPG signal segments.
• To balance accuracy and energy efficiency for wearable sleep monitoring applications.
• To improve sleep staging performance using contextual information from concatenated PPG segments.

Main Methods:

• Adapted a state-of-the-art sleep staging model.
• Concatenated 30-second PPG segments over 15-minute intervals to capture extended context.
• Evaluated the model's performance using accuracy, Cohen's Kappa, and F1-Weighted score.

Main Results:

• Achieved an accuracy of 0.75, Cohen's Kappa of 0.60, and F1-Weighted score of 0.74.
• The proposed method consistently outperformed models using only short PPG segments.
• Demonstrated improved sleep staging accuracy with context-aware approaches.

Conclusions:

• Context-aware approaches using concatenated PPG segments can enhance sleep staging accuracy in energy-constrained wearable devices.
• The proposed method offers a feasible solution for practical wearable sleep monitoring.
• Further research can explore optimizing segment concatenation for improved performance.