DESCRIPTION (provided by applicant): An expert software system is proposed that will analyze digital polysomnographic data and identify EEG arousals based on standard criteria (ASDA report: EEG Arousals: Scoring Rules and Examples, Sleep, 1992). The algorithm logic closely follows the scoring rules so that information from intermediate steps leading up to the fmal decision will be made available to the operator. Phase I research will test three innovative techniques for marking abrupt changes of EEG frequency corresponding to arousal: techniques based on spectrogram, "Time Frequency Distribution", and "Instantaneous frequency". Optimal criteria will be established for discrimination of polysomnogram sections with and without arousal. A database of sixteen manually scored full night sleep records will be established with at least 1200 arousal events. Six records will be used as training set for algorithm development and optimization, and ten records as an independent test set for the assessment of algorithm performance. An effective software tool will be developed by this project for the sleep lab to assess sleep fragmentation.