UNSUPERVISED CLASSIFICATION OF EEG FROM SUBDURAL SEIZURE RECORDINGS

Whereas the visual EEG-inspection of epileptic seizures draws the atte ntion to the waxing and waning of specific graphoelements in multi-cha nnel recordings, the domain of computerized EEG-analysis for epilepsy diagnosis is detection of transients (i.e., spikes) and the quantifica tion of background activity (i.e., mapping procedures). We present an approach to identify relatively fast changes of background activity by use of an automatic classifier. This algorithm is independent of the occurrence of any specific single type of graphoelement. The EEG is se gmentated into short epochs of 0.64 sec duration each. For every segme nt a set of parameters (Hjorth, spectral power in classical frequency bands) is extracted, which taken together build elements of a vector-s pace. The elements are clustered in an automatic and unsupervised mann er by use of a cosine-classifier, such that every EEG-epoch belongs to one class. Changes of brain activity as seen with the EEG are marked by transitions from one class to another. The class occurrence density is defined as the number of different classes that occur within a pre -defined number of EEG-epochs. It gives a new measure of variability o f the EEG-signal. Comparing the epochs when class transitions take pla ce in different channels, the class transitions coincidence between tw o channels is a measure of functional coupling of brain areas.