Multi-valued decision making of sleep stages determination based on expert knowledge

This paper introduced an expert knowledge-based automatic sleep stage determination system working on statistical signal processing. The main methods included two processes. One was expert knowledge base construction, which was developed in terms of probability density functions (pdfs) of parameters for each sleep stage. Here, the visual inspection by a clinician was utilized rather than stage scoring criteria for expert knowledge base construction. Another was multi-valued decision making of sleep stages, where stages were determined automatically according to the conditional probabilities. Totally, four subjects were participated, They are patients from Toranomon hospital, Japan. The automatic sleep stage determination results showed close agreements with the visual inspection in stage awake, light sleep stages and deep sleep stages. The constructed expert knowledge base reflected the distributions of characteristic parameters corresponding to each stage. The proposed method may have strong performance to be an assistant tool for clinicians enabling further inspection of sleep disorder cases.