Improving the excess kurtosis (g2) method for localizing epileptic sources in magnetoencephalographic recordings

Abstract

ObjectiveTo suggest ways to apply the excess kurtosis estimator g2, in the detection of epileptic activity with magnetoencephalography, while avoiding its bias towards detecting high-amplitude, infrequent events. MethodsSynthetic aperture magnetometry (SAM), combined with g2, was applied using window lengths ranging from 0.125s to 32s and with sum and maximum metrics on simulated data and recordings of two focal epilepsy patients. ResultsComparing sources with different spike rates (two per second and one per 2s), the sum metric was most efficient when using a window of 0.25s. Simulations showed that the sum metric is insensitive to spike frequency when the window includes more than one spike. SAM(g2) images from long segments with maximum metric resulted in misleading images, showing the strongest activity away from the lesions. ConclusionsUsing a sliding window and the sum metric is beneficial when imaging interictal spikes and status epilepticus. Windows should be short enough not to include more than one interictal event. For continuous events such as electrographic seizures windows should contain baseline data and the epileptic event. SignificanceThe sliding window and metric should be set according to the suggested guidelines when using SAM(g2) for presurgical evaluation.