Expert Knowledge-based Sleep Stage Determination by Conditional Probability Integrated with Amendment Function

抄録

An expert knowledge-based method integrated with amendment function is presented in order to develop an effective automatic sleep stage determination technique for clinical practice. The main methodology is based on the expert knowledge of visual inspection by qualified clinician. A knowledge database is consisting of probability density functions of parameters for various sleep stages. The sleep stage is decided automatically based on conditional probability. An amendment process is developed to modify the decision on sleep stages. We applied the methodology to the sleep data of patients with sleep disorder under usual recording condition in hospital. The determination result was satisfactory. With the amendment function, the determination system showed better performance on stage change and continuity detection. The developed expert knowledge-based method can be an assistant tool for clinics.