An Ensemble Learning-Assisted Obstructive Sleep Apnea Detection Model Using EEG Physiological Signals and Improved Extra Tree Classifier

Obstructive Sleep Apnea (OSA) is a sleep disorder where people experience repeated interruptions in their breathing during sleep. Electroencephalogram (EEG) data can provide valuable insights into the disruptions caused by OSA during sleep. This letter proposes an ensemble expert system for obstructive sleep apnea detection (EESOSAD-I) using a single-lead EEG signal. The EESOSAD-I leverages discrete wavelet transform with db8 for EEG signal decomposition into sub-bands and formulates essential statistical features for information encapsulation. A state-of-the-art data preprocessing and feature fusion pipeline has been designed and developed for feature refinement and underlying pattern enhancement toward OSA detection. With the developed data refinement and incorporated ensemble extra tree classifier, the proposed EESOSAD-I achieved a notable performance accuracy of 84%, outperforming all the incorporated benchmarks with optimal margins. Further, the proposed model has also been compared with multiple state-of-the-art models, and it demonstrated superior performance in various evaluation schemes with the highest performance of 84% and an average performance accuracy of 82%.