Spectral analysis of the EEG: comparison of various spectral parameters

One of the main problems in spectral analysis of the EEG consists of finding relevant parameters which are capable of summarizing all information included in the power spectrum. The mathematical properties and the clinical applicability of 4 sets of parameters were compared: (1) percentage of power within the conventional frequency ranges, (2) mobility and complexity (Hjorth), (3) the slope parameter (Dumermuth), (4) the coefficients of the autoregressive model. They were applied to the EEG changes in a child with brain stem tumour during cerebral radiotherapy. The statistical analysis suggests: (I) there are only minor differences in describing the EEG course between the conventional frequency ranges, mobility and the slope parameter, (II) the complexity is less satisfactory, (III) the first autoregressive coefficient of the AR model yields the best results.