A Novel Sleep Stage Scoring System: Combining Expert-Based Rules with a Decision Tree Classifier.

Area of Science:

• Sleep Medicine
• Computational Neuroscience
• Biomedical Engineering

Background:

• Polysomnography (PSG) is the standard for sleep analysis and disorder diagnosis.
• Manual sleep stage scoring by specialists is time-consuming and subjective.
• Automating sleep scoring can improve efficiency and consistency.

Purpose of the Study:

• To develop and validate an automated algorithm for sleep stage scoring using PSG data.
• To ensure the algorithm adheres to the American Academy of Sleep Medicine scoring rules.
• To assess the accuracy, robustness, and interpretability of the automated scoring system.

Main Methods:

• Developed a likelihood ratio decision tree classifier.
• Extracted time and frequency domain features from EEG, EMG, and EOG signals in 30-second epochs.
• Trained and tested the algorithm on PSG data from 38 healthy individuals.

Main Results:

• Achieved an overall scoring accuracy of 80.70% on the test set.
• Demonstrated comparable accuracy between training and test sets, indicating robustness.
• The algorithm's performance was consistent with visual scoring and highly interpretable.

Conclusions:

• Automated sleep stage scoring is feasible and accurate, adhering to expert rules.
• The developed algorithm offers a robust, fast, and interpretable alternative to manual scoring.
• This technology has the potential to enhance sleep disorder diagnosis and research.