Prestimulus EEG classification by means of parametric methods to characterize evoked potential variabilities

The prestimulus ongoing EEG activities in pattern visual evoked potential (pVEP) recordings are adaptively segmented to search for short-term EEG segments. The prestimulus EEG segments are clustered by a nonsupervised hierarchical clustering algorithm, which yielded 5-6 classes in each dataset. The averaging of the sweeps in particular classes yielded pVEPs, which differed from each other in time and frequency domain characteristics. The results show the functional significance of the short-term EEG patterns in terms of their correlations with the variabilities of pVEPs evoked by the same type of stimulus.