Wavelet-based sparse source imaging in localizing epileptic sources for partial epilepsy

In the present study, we investigated the performance of a recently proposed sparse source imaging algorithm, i.e. wavelet-based sparse source imaging (W-SSI), in estimating epileptic sources using magnetoencephalography (MEG) interictal spikes. Three patients with medically refractory partial epilepsy were studied. Spontaneous MEG data were recorded using 148-channel magnetometers. Realistic boundary element (BE) head model and cortex model were built from segmenting each subject's structural magnetic resonance imaging (MRI) scans. W-SSI was performed on epileptic spikes selected by experienced doctors and results were compared with clinic data from pre-surgical evaluations, such as clinic diagnosis, MRI lesions, and surgical resections when available. Estimated sources in Patient 1 were located at surrounding areas of MRI lesions, and estimated source in Patient 2 were located within resection areas. Estimated sources in Patient 3 with Landau-Kleffner syndrome (LKS) were distributed on perisylvian regions involving auditory and language cortices, which is consistent with the language ability loss in LKS patients. The results demonstrated the capability of W-SSI in localizing cortical sources from clinic epilepsy data, which suggests that it is a promising tool for pre-surgical evaluations in partial epilepsy. Moreover, the dynamic patterns of recovered cortical sources further provided important information in understanding the formation and propagation of epileptic discharges.