PATTERNS

EEG spectral patterns

Based on the partial tone model of EEG and taking into account the specific properties of time-dependent spectra derived from non-stationary time series, we developed our concept of spectral patterns (Stassen 1980; Stassen et al. 1982). This approach generalizes the notion "spectrum" in the sense that corresponding spectral intensities are regarded as fluctuating rather than being fixed-valued, thus incorporating the non-stationary nature of EEG time series into the model.
      Indeed, spectral patterns are a "logical" extension of previous approaches to modelling EEG individuality: Since any parameterization of spectra in terms of frequency bands necessarily reduces inter-individual differences and imposes a certain conformity in EEG characteristics, subsequent authors based their analyses on the spectrum as a whole and introduced, for example, the spectrum difference ratio (Lykken et al. 1974) in order to more reliably assess EEG differences between individuals. The practice, however, of handling time-varying EEG spectral densities either by averaging over many successive short epochs (the resolution of spectral analyses is inversely proportional to the underlying epoch length of time series, so that epoch lengths cannot be arbitrarily chosen. A 4-second epoch, for example, is required to get a 1/4 Hz resolution in the frequency domain) or by basing spectral analysis on the longest possible periods (which implicitly has an averaging effect) typically results in a considerable loss of information. In particular, single epoch evaluations of the EEG, of whatever length, are of limited use with regard to reproducibly measuring inter-individual EEG differences. In consequence, the concept of spectral patterns is a way to overcome these difficulties.
      In this model, the various brain regions are assumed to exhibit basic electrical activities which are continuously modified in the course of performing tasks or of reacting to events external to the central nervous system. Accordingly, in the absence of such tasks and events, reactive changes of electrical brain activity are expected to reach a minimum. Under this condition, the "natural" variability of brain wave patterns can apparently be assessed without great difficulties. If, in addition, we postulate that the generation of brain waves occurs under strong internal regularities (rather than being random), the natural variability inherent to brain waves can be estimated from a finite number of consecutive epochs.
      Once EEG spectral patterns have been designed and a suitable similarity measure has been selected, the fundamental questions concerning individual EEG characteristics can be formulated in terms of the partial tone model:

Moreover, this method of approach makes it also possible to directly control the goodness of EEG quantification (with respect to measuring the individual EEG characteristics) on the basis of the between-subject discriminating power provided by the model. However, in order to successfully establish the method, we need to determine the free parameters inherent to the approach: (1) an appropriate recording time containing enough information for a reliable estimation of spectral patterns, (2) a subdivision of the recording time into epochs large enough that the actual composition of partial tones is revealed, yet small enough that the individual variability of each spectral component can be estimated from a sequence of consecutive epochs and, (3) a subdivision of the frequency domain into intervals compatible with the relative importance of EEG bands in their dependence upon the actual state of consciousness.
      This task is a typical problem of pattern recognition and, accordingly, can be carried out by means of trainable algorithms in connection with a design sample set, a test sample set, and an appropriate criterion function ("supervised learning"). Such procedures merely assume that if all training samples are correctly processed, few mistakes will be made on the test sample set. Following this design, we realized a normative study based on a sufficiently large and representative sample of male and female volunteers, whose EEGs were recorded twice at an interval of 14 days. Subsequently, the recordings of the first measurement served as training samples, whereas the recordings taken 14 days later were referred to as test samples in order to derive sample-independent and reproducible calibration parameters.
      Once all calibration parameters are determined, the overall resolution of the method can be tested experimentally, for example, by cluster analysis, so that the goodness of measured EEG individuality can be quantified in an objective way. Ideally, patterns derived from repeated measurements on the same individual should cluster together, with clusters of different, unrelated persons completely separated from each other. In non-ideal cases, on the other hand, the between-subject overlap of clusters is a measure of de facto discriminating power.


Feedback

If you have questions or comments concerning this program package send e-mail to one of the following addresses:

[ HOME Psychiatric University Hospital Zurich ]
[ HOME Psychiatric University Hospital Zurich, German Pages ]