A Robust Sleep Apnea-hypopnea Syndrome Automated Detection Method Based on Fuzzy Entropy Using Single Lead-EEG Signals

Sleep apnea-hypopnea syndrome is a relatively common disease, characterized by a repetitive reduction or cessation of respiratory airflow, which seriously affects sleep quality and in the long term, can lead to heart disease, hypertension, diabetes, and stroke. Polysomnography is currently the standard method for diagnosing apnea/hypopnea. However, accurate diagnosis can be difficult due to the complex process of multi-signals acquisition in polysomnography. Instead, this paper presents a novel automated fuzzy entropy-based method for detecting apnea/hypopnea using single-lead electroencephalogram signals. The method consists of four steps: (1) The electroencephalogram signals corresponding to respiratory events are partitioned into five sub-bands according to frequency; (2) Features of fuzzy entropy in each sub-band are extracted; (3) The extracted features are evaluated using statistical methods; (4) The features are classified using a classifier, such as the support vector machine, k-nearest neighbor, and random forest algorithms. In this study, data were obtained from a total of 55 subjects with sleep apnea-hypopnea syndrome from both a public and clinical database. The experimental results indicated that all of the selected metrics, including accuracy, sensitivity, and specificity were close to or above 90% for both publicly available and clinical data. Moreover, this approach is sensitive to all types of sleep apnea/hypopnea, an important aspect that is rarely explicitly discussed in the literature.