Automatic Sleep EEG Classification with Ensemble Learning Using Graph Modularity

Sleep is of paramount importance to us, and it aids in healing mental as well as physical health of humans. Lack of sleep builds up the risk of serious diseases such as diabetes and cancer. The proficient technique of scoring the sleep stages in electroencephalogram (EEG) is an indispensable tool for medical practitioners in order to diagnose sleep disorders at an early stage. In this regard, correlation graphs are employed in order to classify the sleep stage EEG signals. The EEG segments are split to subsegments, and then by using the statistical methods, the dimensions of the subsegments are scaled down. Subsequently, each segment is mapped to a graph by taking particular subsegments as a vertices of the graph. The relation between the vertices of the graph is computed. The modularity of the graphs is taken as the features to a classifier based on ensemble learning. Depending on the outcomes of the ensemble classifier, the sleep EEG signals are classified. The experimental results show that the performance of the ensemble classifier for classifying the EEG sleep disorder achieves better results as compared to the individual classifiers.