ANALYSIS OF SIGNAL COMPLEXITY VERSUS SPEC TRAL PARAMETERS IN EEG TIME-SERIES OF PSYCHIATRIC-PATIENTS - A RETROSPECTIVE PILOT INVESTIGATION USING ROUTINE EEG RECORDINGS IN PSYCHIATRY

In order to confirm other findings that the EEG from schizophrenics an d depressives is different from usual controls and differ from each ot her we performed a retrospective pilot investigation using routine EEG recordings in psychiatry. From 123 schizophrenic and 84 depressive pa tients closed-eye-EEG's were registered after admission and compared w ith EEG's of 27 healthy volunteers. 32 of the schizophrenic patients w ere unmedicated, 24 received haloperidol, 39 were treated with perazin e and 28 with clozapine. 30 of the depressive patients were treated wi th amitriptyline, 54 patients didn't receive any medication. EEG data were digitally stared and analyzed by using common spectral parameters and the algorithms of Hjorth's global frequency analysis: complexity, mobility and mean variation of activity. Further more, the correlatio n exponent of Grassberger and Procaccia was determined. These paramete rs, with exception of mobility and the spectral parameters, have in co mmon that they describe the degrees of freedom in a EEG time series, h ence, give us an impression about the complexity of signal in differen t physical domains. Finally, a classification of the different diagnos tic and medicated groups was carried out, employing all measured EEG-p arameters. Thus, we wanted to find out whether classification of psych iatric diseases is possible this way and especially, which variables a re discriminative. For untreated schizophrenics versus healthy control s we found an average correct reclassification of 93%. Best discrimiti ation was achieved by the parameters measuring signal complexity, wher eas spectral parameters only contributed a minor part. Discriminating unmedicated depressive patients versus healthy controls and depressive s versus schizophrenics was possible in 82 and 85% of cases, respectiv ely. Here, both power spectral and signal complexity parameters contri buted almost aquivalent to the classification. Finally, it could be de monstrated that medicated patients can be classified correctly to a le sser extend and show a strong tendency to a ''normal'' configuration o f EEG-variables. In a forward classification taking three groups (schi zophrenics, depressives and normal controls) about half of the depress ive and schizophrenic patients were correctly allocated, thus showing some speciality of the EEG effects but also unexplained false classifi cations and fuzziness. Under medication there is a shift towards norma lity: Some of the schizophrenic and depressive patients are classified as normal controls, indicating EEG-changes towards normality under th e treatment with psychopharmaca. The EEG of normal controls and psychi atrially ill unmedicated patients seems to be entirely different. Howe ver, between schizophrenics and depressives there is some overlap whic h is - so far - not explained.