An ensemble method for improving robustness against the electrode contact problems in automated sleep stage scoring

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Artificial Intelligence in Healthcare

Background:

• In-home automated sleep scoring systems face challenges with electrode contact and signal quality, limiting clinical adoption.
• Current systems often yield unstable and inaccurate sleep stage scoring due to signal interference.
• Existing methods struggle to reliably handle data affected by electrode failures.

Purpose of the Study:

• To develop a robust automated sleep stage scoring method resilient to electrode contact failures.
• To enhance the accuracy and reliability of in-home sleep scoring for clinical applications.
• To mitigate the impact of problematic signals on overall scoring performance.

Main Methods:

• Proposed an ensemble method using multiple small sleep stage scoring models with diverse input signal sets.
• Excluded models utilizing problematic signals from the ensemble's voting mechanism to reduce error propagation.
• Investigated the efficacy of assigning different input signal sets to individual models within the ensemble.

Main Results:

• The proposed method significantly reduced the impact of electrode contact problems on scoring accuracy.
• Improved accuracy for epochs with problematic signals by 8.3 points compared to conventional methods.
• Decreased the overall deterioration in scoring accuracy from 7.9 to 0.3 points, demonstrating enhanced robustness.
• Confirmed that varied input sets for ensemble models increased overall efficacy.

Conclusions:

• The developed ensemble method enhances in-home sleep stage scoring accuracy while minimizing the effects of signal quality issues.
• This approach makes automated sleep scoring systems more reliable and suitable for widespread clinical use.
• The findings support the integration of advanced AI techniques for improved sleep disorder diagnosis and monitoring.