tsta

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Development of an Explainable Random Forest-Based Algorithm for EEG-Based Sleep–Wake Classification Toward Sleep Apnea Detection

Authors :

Pargol Sharifi¹ Mohammad Fakharzadeh²

1- دانشگاه صنعتی شریف، تهران، ایران 2- دانشگاه صنعتی شریف، تهران، ایران

Keywords :

Polysomnography،EEG signal،Sleep stage classification،Sleep apnea detection،Random Forest،Machine learning

Abstract :

Automatic sleep stage classification allows separating sleep stages without human experts. Many existing algorithms rely on multi-channel physiological signals such as EEG, EOG, and EMG. However, because of the complex equipment and required expertise, these methods usually need specialized laboratory settings. Therefore, developing a high-accuracy classification algorithm using a single signal remains a key challenge in sleep research, as it could enable portable devices and home-based sleep monitoring systems. Sleep stage classification is essential for detecting and managing sleep disorders such as sleep apnea. This study presents an optimized and clinically interpretable pipeline for sleep stage classification and apnea detection using EEG signals. The proposed approach is based on a simple, interpretable Random Forest framework and is intended to serve as a valuable tool for both clinical and research-oriented applications in sleep apnea detection. It integrates optimized preprocessing, data cleaning, algorithmic optimization, and class balancing to enhance accuracy and interpretability. Notably, our optimized Random Forest pipeline outperforms more complex deep-learning models, especially on 6-class sleep staging, sleep–wake discrimination and Apnea detection. The proposed method achieved accuracy, sensitivity, and specificity of 99.68%, 97.59%, and 99.30%, respectively, for distinguishing sleep from wakefulness, and 87.18%, 85.19%, and 89.16%, respectively, for apnea detection.