A Sleep Disorder Detection Model based on EEG Cross-Frequency Coupling and Random Forest

Study objectives Sleep disorders are medical disorders of the sleep architecture of a subject, and based on their severity they can interfere with mental, emotional and physical functioning. The most common ones are insomnia, narcolepsy, sleep apnea, bruxism, etc. There is an increment risk of developing sleep disorders in elderly like insomnia, periodic leg movements, rapid eye movement (REM) behaviour disorders,sleep disorder breathing, etc. Consequently, their accurate diagnosis and classification are important steps towards an early stage treatment that could save the life of a patient. The Electroencephalographic (EEG) signal is the most sensitive and important biosignal, which is able to capture the brain sleep activity that is sensitive to sleep. In this study, we attempt to analyse EEG sleep activity via complementary cross-frequency coupling (CFC) estimates that will further feed a classifier, aiming to discriminate sleep disorders.

Methods We adapted an open EEG Database with recordings that were grouped into seven sleep disorders and a healthy control. The EEG brain activity from common sensors has been analysed with two basic types of cross-frequency coupling (CFC). Finally, a Random Forest (RF) classification model was built on CFC patterns, that were extracted from non-cyclic alternating pattern (CAP) epochs.

Results Our RF^CFC model succeeded a 74% multiclass accuracy. Both types of CFC, PAC and AAC patterns contribute to the accuracy of the RF model, thus supporting their complementary information.

Conclusion CFC patterns, in conjunction with the RF classifier proved a valuable biomarker for the classification of sleep disorders.

Statement of Significance In this study, we developed an efficient model that is able to perform sleep disorder diagnosis by analysing the EEG sleep activity under the framework of cross-frequency coupling (CFC) with the support of RF. CFC has been proven an important mechanism that supports the integration of neural activity of different frequency content through a nested hierarchy of their oscillatory pattern inherent to distinct neural functions. Our results suggest that CFC can reflect aberrant physiological interactions during sleep stages, which are sensitive to differentiate sleep disorders. Therefore, our RF^CFC model can be a valuable mental health tool for an accurate classification of those somnipathies.