AUTOMATIC SLEEP AROUSAL DETECTION AND ANALYSIS OF THERAPEUTIC TECHNIQUES FOR SLEEP DISORDER.

Sleep arousals are the most common among the sleep disorders. Sleep arousals are characterized by the sudden shift in electroencephalogram (EEG) frequency during sleep. These sleep arousals cause fragmented sleep, which are the most important factors for day time sleepiness and cause for various sleep disorders. To diagnose the sleep arousals, a polysomnographic (PSG) test is to be done in a sleep laboratory. Normally, a single PSG recording is taken for an average duration of eight hours. This makes the manual detection of sleep arousals not only error prone but also a tedious task. Hence automation of this process is needed. In this work a novel automation method is presented to detect the presence of arousals. At first two EEG signals and an EMG signal are extracted from the PSG recordings then the statistical features are extracted, followed by grouping and classification using artificial neural network (ANN). The novelty of this work relies on the usage of both Hjorth and Power Spectral Density difference spectrum features leading to an improved accuracy than either of them alone. In addition the work also demonstrates an effective therapeutic technique for insomnia, which is a modification and improvement of certain steps of the existing Cognitive Behavioral Therapy (CBT), tested in the sleep lab, at IIT Kanpur. The method, showed positive changes in the sleep quality of the subjects. 
Statistical analyses indicate that the method leads to significant increase in the average value of sleep efficiency, percent REM sleep etc, followed by an average decrease in the mean value of REM latency, Arousal Index, Apnea-Hypopnea related arousal indices etc, proving the effectiveness of the method used.