High resolution patient specific models of the EEG and MEG using 3D segmentation

Abstract

Epilepsy is a common and severe neurological disorder. In 20% of cases the condition is unaffected by drugs and surgery is considered. In the surgery the area where the seizures start is removed. Recent work using phase synchronization utilizes short duration high density (256 channel) interical scalp EEG to localize the epileptic area on the scalp. To translate that area to the neuronal sources below mathematical modeling of the electrical activity of brain is needed. The neuronal activities which mainly contribute to the EEG are the postsynaptic potentials of the pyramidal neurons in the cortex. A 3D model of the head is needed to model the electrical activity of the brain which can be developed from segmented MR images. Ten tissues were segmented in the MRI data of an adult female subject using the software platform Mimics. Normal vectors were computed on the white matter surface. For each normal vector a dipole was computed representing the activity of a group of pyramidal neurons in the cortex. The dipole intensity was from 0-40 µA/cm2 with a uniform random distribution. The tissues were assigned conductivities from the literature. The dipolar sources and tissues were imported into an adaptive FEM solver which solves the potential and current flux in the whole head. The scalp potentials were extracted and referenced to a common average reference. The MEG sensor coils were assumed to be 1.0 cm above the scalp surface. The magnetic fields at the sensor locations were computed by use of Biot-Savart law. The above model was tested to study the differential contributions of the left and right hemisphere and the Cerebellum to the EEG. For the left brain, the positive and negative peaks were towards the left side of the plot. For the right brain the positive and negative peaks were towards the right side of the plot. This suggests that one possibly may infer lateralization of the left and right cortical source volumes from the scalp potential plots. Simulations of stroke were performed which suggest that stroke location may be inferred from EEG. Later on this model will be applied for epileptic stroke localization.