Abstracts

Rationale: Stereoelectroencephalography (sEEG) is increasingly used to identify ictal onsets and interictal epileptiform discharges (EDs) to define the epileptogenic zone in patients with medically-refractory epilepsy. At our institution, having completed long-term sEEG monitoring, intraoperative electrocorticography (ECoG) with surface grids is used to refine the sEEG-based surgical resection plan. Specifically, a resection that had been planned on the basis of long-term sEEG ictal onsets and/or EDs may be extended if EDs are identified outside of the planned resection on intraoperative surface ECoG. We recently showed that a statistical approach to analyzing interictal data, termed Granger causality (GC), may be a helpful adjunct to ED identification in localizing ictal networks. Here we extend this statistical approach and directly compare interictal long-term sEEG monitoring data to that recorded during intraoperative surface ECoG. Methods: Out of all sEEG cases conducted at our institution between April 2018 and May 2019, we identified five in which both long-term sEEG and pre-resection ECoG data were available. From these we analyzed five minute segments of interictal sEEG data and three to five minute segments of pre-resection ECoG data. Following the method reported in our earlier paper (Neurosurgery, 2018), we calculated highly-causal sEEG and ECoG nodes, using GC. We then utilized intraoperative brain surface photographs and 3D volume rendered brain images generated by Slicer and FreeSurfer programs to assess agreement between the ultimate resection, highly-causal sEEG nodes, and highly causal ECoG nodes. Results: In all five cases, the regional representation of high causality areas identified by ECoG was generally concordant with that seen in the sEEG recordings, accounting for technical differences. The tendency for clustering due to non-random grid placement could be tested by direct calculation of the pairwise Cartesian distances between ECoG electrodes and sEEG electrodes. These areas of high causality also tracked with the regions ultimately resected, so that on average 80% of the hightly-causal sEEG contacts and 92% of the highly causal ECoG contacts were within the resection zone.  Conclusions: GC analysis, independently applied to long-term sEEG and intraoperative ECoG interictal recordings reveals anatomical proximity of highly-causal nodes to the ultimate resection zone. These data suggest that GC analysis of pre-resection ECoG grid recordings and interictal sEEG recordings perform similarly in predicting the traditionally-identified resection zone and putative epileptogenic zone as in our earlier study (2018 AES Abstract ID: 502599). Funding: This work is funded in part by NIH (1R01NS069696 and U01EB023820-01) and in part by Wyss Institute for network data storage and management.