Automatic Identification of K-complex in Sleep EEG Signal for Clinical Practice

Abstract

Sleep is one of the prime natural activities for human well-being in physical, emotional, and mental aspects. The assessment of sleep EEG requires in the diagnosis of some sleep-related disorders. Sleep transients (K-complex or sleep spindles) are one of the effective parameters in sleep studies. K-complex (KC) detection is a relentless task from the long-duration recorded sleep EEG signal. Manual visual scrutiny takes ample time, susceptible to errors, and needs a skilled person. A computer-aided automatic detection model has been designed to identify KC in the sleep EEG signal to conquer these difficulties. The overview of the chronological advancement in KC detection's automated methods is also enlightened in this paper. The demonstration of this study is carried out on the sleep-EDF EEG database. The proposed simulation model followed the process flow of signal template generation for KC using prior knowledge and matched filter based automatic KC identification using the defined template. This proposed model can give better performance even if the test signal is badly corrupted with noise, i.e., low SNR (signal to noise ratio). The sensitivity, specificity, and accuracy of the model using this intended method are 92.72%, 87.48%, and 91.2% respectively, with comparatively better results than reported in the literature. Hence, considering the achieved outcomes in the test signal assessment, this proposed method can be an alternative for automatic identification of K-complex that may help neurologists diagnose sleep-related problems for clinical practice and diminish neurologist's load to a greater extent.