Abstracts

Rationale: Epilepsy surgery evaluation generates large quantities of multimodal data including imaging and electrophysiology. It is critical to aggregate this data in a common space for research and clinical decision making. Existing brain atlases lack an intuitive naming or parcellation scheme that provides for easy cross-referencing of multimodal data. We define an atlas which addresses these issues in epilepsy surgery and test its use as a visual database with a cohort of patients.

Methods: We registered multimodal data from epilepsy patients onto the Yale Atlas; our cortical brain atlas parcellated to reflect the standard 1cm contact spacing of intracranial electrodes and defined relative to robust landmarks on the MNI 152 brain. Structural 3T MRI images underwent linear and nonlinear registration, and 3D electrode coordinates were transformed to the atlas space, which was then verified for accuracy using Yale BioImage Suite software. Electrophysiology data pertaining to seizure onset and spread regions was mapped, and bipolar stimulation mapping for language was incorporated.

Results: A total of 3386 intracranial electrode contacts from 20 temporal lobe epilepsy patients were transformed to the atlas: an average of 169 contacts per patient. The cohort was split in two: ten patients had a mesial temporal onset of seizures, and ten had a neocortical onset. In the mesial temporal group, the most common parcel of onset was H A (pes hippocampus) in 5/10 (50%) followed by H B (hippocampal body) in 4/10 (40%). In the neocortical group, the most common parcel of onset was TP2.2 A (anterior-inferior temporal pole) in 4/10 (40%) followed by T2.1 C (upper-middle temporal gyrus) in 3/10 (30%). In a comparison of illustrative cases (Figure 1), a patient with onset in H B underwent an extended temporal resection including the parcels involved at onset as well as the first and second spread and remained seizure free at one-year. The second patient had spread outside of the resected neocortical parcels and failed surgery. A third case is offered for illustration of language mapping superimposed on the atlas (Figure 2). Seizure onset was observed in the dominant temporal lobe neocortex. Mapping identified parcels that were involved in critical language function, and two RNS strips were subsequently placed to cover the parcels of onset and spread with an ILAE 2 outcome at one-year.

Conclusions: We present an epilepsy surgery atlas for aggregating multimodal data, including imaging, electrophysiology and functional stimulation mapping. The atlas allows for co-registration of multiple data sources onto the same visualization to allow for exact cross-referencing when used for clinical decision making. It also aids the aggregation of data from multiple patients for use in research. We intend to map over 100 patients to the atlas to compare PET, fMRI, stimulation mapping and electrophysiology for outcome correlation.

Funding: Please list any funding that was received in support of this abstract.: Department of Neurosurgery, Yale University School of Medicine.