Computer-assisted sleep staging based on segmentation and clustering

In this paper, we present a method that can be used to automatically classify sleep states in an all-night polysomnogram (PSG) to generate a hypnogram for the assessment of sleep-related disorders. The method is based on ideas of segmentation and classification (clustering) using sleep related features. Segments are clustered to generate groups of similar patterns that can subsequently be labeled as one of the accepted clinically relevant sleep stages. Each PSG is processed independently to generate classes of similar patterns in an unsupervised manner, thus achieving pseudo-natural classes that are independent of any classification criterion. Overall performance as compared to manual scoring of 12 subject is shown to be 61.1%.