Conditional probability of Cauchy distribution in automatic sleep stage determination for sleep data with artifacts

An automatic sleep stage determination system dealing with the sleep data contaminated by artifacts is developed, which is working on an expert knowledge-based multi-valued decision making method. The knowledge database is consisted of probability density functions of parameters for various sleep stages according to the visual inspection by a qualified clinician. The probability density functions are approximated by Cauchy distribution on histograms. Sleep stages are determined automatically according to the maximum value of conditional probability. Due to the infinite variance of Cauchy distribution, the effect of mis-determination caused by artifacts can be abated. The result of automatic sleep stage determination was satisfactory. The presented automatic sleep stage determination can be an assistant tool for clinical practice.