Statistical discrimination of controls, schizophrenics, depressives and alcoholics using local magnetoencephalographic frequency-related variables

Abstract

Atypically enhanced activity in the delta and theta EEG frequency bands has frequently been reported for schizophrenic patients, while alpha activity is often attenuated in these patients [2,9,10,12]. MEG and EEG data provide an advanced approach to analyze complex brain functioning and to examine differences between different psychiatric patient groups due to their brain activity. Past analyses using different physiological parameters to discriminate a psychiatric patient group from controls reached statistical correct classification rates of at maximum 80 percent. Results usually shifted to chance when adding a third group to the analysis. Winterer (2000) [13], for example, could discriminate between schizophrenic patients and controls with a correct classification rate of 77 percent when using delta power, signal power at Cz and power values of the high alpha range as variables in a discriminant analysis. Including a group of depressive patients in the analysis reduced the correct classification rate to 50 percent. Gallhofer (1991) [5] used 50 topographical frequency-related EEG-parameters in a discriminant analysis with schizophrenic and depressive patients and controls. He classified 49 out of the 50 subjects correctly.Strategies that try to describe the physiological substrate of psychiatric diseases with only a few parameters possibly over-simplify the nature of the phenomenon [see also 5]. More complex strategies are possibly more adequate to describe complex phenomena such like psychiatric diseases.The present study examined to what extent delta-, theta and alpha-band-related source space activity can separate controls, schizophrenics, depressives and alcoholics by discriminant analysis. The analyses are meant as a first step towards an evaluation of a set of physiological parameters that could possibly be representative of certain psychiatric gross groups. In order to explore possible methods sensitive to these physiological parameters, different strategies of MEG source space analysis and statistical procedures were performed on data obtained during three different mental modalities (rest, mental calculation and mental imagery).Enhancement in focal [1] as well as in multiple [4] slow wave activity has been reported for schizophrenic patients. A reduction of alpha activity has been reported for schizophrenic [10,12] and alcohol [3] patients as well. For the analysis of focal sources we performed the dipole density method that has been shown as a valid tool in the vicinity of the detection of pathological attributed slow wave activity for example around tumors [8] or lesions [11]. Multiple source activity in the slow wave and alpha range was detected by the minimum-norm method [6,7].