Spatio-temporal clustering of epileptic ECOG

The spatio-temporal mechanisms underlying the generation of epileptic seizures is not yet clearly understood. In this study, we attempt to quantify the spatio-temporal interactions of an epileptic brain by using a previously proposed SOM-based Similarity Index (SI) measure. We further show that spectral clustering approach can be appropriately used to determine the average spatial mappings in the brain at different stages of a seizure, by interpreting the SOM-SI values as affinity matrices. Results involving two pairs of seizures of an epileptic patient suggest that there may not be a regular pattern associated with channels's spatio-temporal dynamics during the inter-ictal to prepost ictal transition.