Clustering of hypnographic parameters derived from human sleep Eeg

Clustering of hypnographic parameters derived from human sleep EEG. Certain EEG-derived parameters, averaged over time intervals of approximately one minute throughout the night , previde curves which can be immediately compared with the hypnogram. For this reason, these parameters have been often used in the quantitative and/or automatic analysis of sleep. On the other hand they assume continuous, not discrete, values and therefore supply additional information to that of the hypnogram. A classification of these parameters should be useful to evaluate their effectiveness in describing the various EEG rhythms and to check whether some of them give complementary information or are redundant. For this purpose, a c1ustering of 34 automatically computed parameters was performed for 12 nights on 4 subjects (3 per subject). The number of c1usters obtained was 7 for lO nights and 8 for the remaining 2. Five c1asses of pararneters, valid for alI subjects and for alI the nights, were recognized.