Ubiquitous Depression Detection Of Sleep Physiological Data By Using Combination Learning And Functional Networks

Nowadays, depression has become a common mental disorder with high morbidity and mortality. Due to the limitations of traditional interview-based depression detection, it has become an urgent problem to realize objective, convenient and fast detection. This study is to explore ubiquitous methods of depression detection based on combination learning and functional networks, using sleep physiological data. Sleep physiological data were collected using a portable physiological data instrument, and then preprocess and extract several related features. We applied combination learning to discover the best sleep stage, the optimal features subset, and the most effective classifier, which are hidden behind physiological features, to detect depression. Physiological features in the optimal feature subset based on Euclidean distance are mapped to nodes to construct the functional network. The optimal feature subset was combined with the functional network attributes as the input of the most effective classifier to get the ultimate performance of depression detection. Controlled trials based on ubiquitous sleep physiological data were conducted on different genders. Experiments show that the best results for male and female were derived from slow wave sleep (SWS) and rapid eye movement (REM), with performances of 92.21%; and 94.56%, AUC of 0.944 and 0.971, respectively. Thus, our study may provide an effective and ubiquitous method for detect depression.