Abstracts

Rationale: Current data show that 20-30% of patients with epilepsy are refractory to medical therapy, and are therefore possible candidates for resective neurosurgery. For these patients, pre-surgical evaluation is performed to localize both epileptogenic and eloquent cortex so that surgery is safe and effective. Comprehensive pre-surgical workup includes brain imaging, long-term electroencephalography (EEG) monitoring, Wada testing, and neuropsychological testing. Despite rigorous evaluation, resective surgery results in seizure freedom in only two-thirds of cases. Epilepsy has recently been re-thought of as a network-level disorder, which may explain why many patients fail focal resection. Since current pre-surgical assessments do not evaluate network architecture, a new tool is needed. Recent studies have demonstrated the value of resting state functional MRI (rsfMRI) in mapping brain networks, and epileptogenic cortex can be identified by applying EEG source-localization algorithms. Here, EEG source localization and rsfMRI connectome mapping were combined to create a 3D network map of each patient. Methods: Five minutes of continuous blood oxygenation level dependent (BOLD) MRI data were acquired while the patient was supine with their eyes closed. On a different day, scalp EEG recordings were recorded for up to five days and a neurologist identified all ictal and inter-ictal discharges. Open-source BOLD MRI analysis software was used to motion correct, smooth, and normalize the MRI data. Then, BOLD time-series was extracted for ninety functional regions of interest (ROIs) (W. R. Shirer and others, ‘Decoding Subject-Driven Cognitive States with Whole-Brain Connectivity Patterns’, Cerebral Cortex, 22.1 (2012), 158–65). Connection strength between ROIs was quantified and used to plot the strongest connections in a 3D map. Then, EEG data was cropped, filtered, and co-registered to the BOLD MRI. Finally, an inverse method of source localization was used to map ictal and inter-ictal discharges. Results: The algorithm was first validated in a healthy control brain. Consistent with previous studies, the map showed an abundance of interhemispheric connections and was highly symmetric. Average correlation of ROIs in the left and right hemispheres were not significantly different (Fig 1a). The connections measured as above average in strength were equally distributed between left and right hemispheres. Next, three patients with temporal lobe epilepsy were mapped. In two cases, the connection map showed reduced strong connections in the affected hemisphere (-33%, -19%) (Fig 1b). In the third case, connection strength was not significantly different between sides. In all three cases, EEG source localization was consistent with the neurologist’s assessment. Conclusions: This modelling algorithm has been validated in a healthy control and was successful in visualizing epilepsy networks. The algorithm localized epileptogenic brain and showed characteristic network asymmetry consistent with previous connectomic studies of epilepsy. In two patients, network architecture was asymmetric, which suggests a paucity or reorganization of networks adjacent to foci. In the third patient, however, the network was symmetrical despite the lateralized epileptogenic source. Future work will examine the use of network symmetry measurements to predict surgical outcomes. Funding: NREF Medical Student Summer Resarch Funding