Application of correlation dimension and pointwise dimension for non-linear topographical analysis of focal onset seizures

For many patients who are candidates for epilepsy surgery, non-invasive evaluation fails to provide sufficient information to permit surgical treatment. Since there are also definite risks and considerable costs associated with invasive procedures, new (non-invasive) techniques are required. This study provides empirical evidence that a non-linear approach applied to ictal surface electroencephalograms (EEGs) can help to delineate the area of seizure onset and may prove useful in complementing visual analysis of the EEG. Multichannel EEGs, recorded from eight patients with different drug-resistant localisation-related epilepsies, were analysed using the concept of correlation dimension and two extensions based on the pointwise dimension. The latter also provided results in cases where assessment of the correlation dimension was not feasible. Comparative values between 2 and 6 were accepted as the result of the algorithms, mostly 3-4 for the EEG channels strongly reflecting epileptic activity, and 4-6 for the other signals. The proportion of accepted pointwise values was usually 200-800% for strong epileptic EEG activity compared to the other data. The approach permitted the characterisation of the scalp area reflecting epileptic activity. The results obtained were in perfect concordance with those obtained during pre-surgical work-up and confirmed by the post-operative outcome.