[Automatic detection of grouped patterns in EEG].

Abstract

The detection and classification of grouped (so called paroxysmal) activity and of artefacts of similar appearance is a necessary part of the automated description of clinical EEGs. An algorithm for this task is presented. It is based on short-time spectra and on parameters extracted from them. The rate of agreement with the visual assessment is nearly the same as that between well trained EEGers. This is true for the classification and to nearly the same degree for the detection. Nevertheless these patterns must be grouped and validated if the method is to be used in practice. A modification useful for a fast analysis combined with the quantitative description of the background activity is discussed.