Decision System Integrating Preferences to Support Sleep Staging.

Area of Science:

• Sleep Medicine
• Artificial Intelligence
• Biomedical Signal Processing

Background:

• Traditional sleep stage scoring relies on machine learning models trained on expert annotations.
• These methods often overlook established American Academy of Sleep Medicine (AASM) guidelines.
• Existing approaches struggle with ambiguous epochs where multiple sleep stage criteria are met.

Purpose of the Study:

• To develop a novel sleep stage scoring approach adhering to AASM guidelines.
• To formalize AASM rules, incorporating preferences for conflict resolution.
• To demonstrate the efficacy of the proposed method in accurately scoring doubtful sleep epochs.

Main Methods:

• Feature extraction from raw sleep signals.
• Segmentation of sleep recordings into 30-second epochs.
• Development of a rule-based system integrating AASM guidelines and preference logic for decision-making.

Main Results:

• The proposed approach successfully applied AASM guidelines for sleep stage classification.
• Formalized rules with preference integration effectively resolved conflicting criteria in doubtful epochs.
• The system demonstrated appropriate decision-making on challenging sleep data.

Conclusions:

• The novel approach provides a more accurate and guideline-adherent method for sleep stage scoring.
• Formalizing AASM rules with preference logic enhances machine-based sleep scoring robustness.
• This method offers a promising alternative to traditional machine learning models for sleep analysis.