Classification of Sleep Consciousness based on Filter Bank Common Spatial Pattern

The aim of this research is to classify the level of consciousness within non-rapid eye movement sleep from electroencephalography data. We defined three categories of conscious experience based on subjective reports: dream experience, dream experience without recall of content, and no conscious experience. The filter-bank common spatial pattern and shrinkage linear discriminant analysis were used to classify sleep consciousness. We also evaluated conscious experience using 10-fold cross-validation. Classification accuracy was 88.8% for the three classes. In particular, the sensitivity for no conscious experience was 0.99, and the sensitivity for dream experience with and without recall of content was 0.86 and 0.73, respectively. These results suggest that it is easy to predict the brain dynamics between conscious and no conscious experience, but to distinguish dream experiences with and without the recall of content is difficult. These findings could help predict the levels of consciousness during sleep and may help in developing a passive biomarker of the levels of consciousness in humans.