Abstracts

Rationale: Recently we have shown that Granger causality (GC) analysis has the potential to reveal ictal networks from interictal baseline data obtained not only from subdural grids but stereotactically implanted depth electrodes. While GC analysis of extraoperative interictal data appears to predict ictal targets, the question of whether recordings under anesthesia during the placement of stereoelectroencephalography (sEEG) depth electrodes (potentially useful for modifying the implantation strategy) could do the same has not been resolved. In this study, we tested whether the regions of high GC based on intraoperatively recorded sEEG would statistically resemble the topography of the seizure onset zone and resection as seen in GC maps calculated from extraoperatively recorded interictal sEEG data. Methods: We analyzed four to five minutes of intraoperative interictal data collected immediately after the completion of implantation and five minutes of baseline data recorded postoperatively (n=9). Granger causality maps were statistically compared to conventionally-constructed surgical plans and resections, by using non-parametric, rank-order statistics as described in our earlier paper (Neurosurgery 2018). Electrode rankings were also compared between intra- and extraoperative GC directly. Results: In 7 out of 9 cases, the intraoperative interictal GC rankings of the electrodes mapped to the seizure onset zone had higher causality than predicted by chance (range: p < 10-5 to 0.0004), which is comparable to extraoperative interictal GC rankings (range: p < 10-5 to 0.004 in all 9 cases). In all 9 cases of extraoperative data sets, causality in the resection zone was significantly increased (range: p < 10-5 to 0.003). In 8 of 9 cases for intraoperative data sets, causality in the resection was also similarly increased (range: p < 10-5 to 0.02). The intra- and extraoperative GC electrode rankings were statistically similar in all 9 cases (p < 0.005 in each case, Spearman). The aggregate probability of such a match is very small for both intra- and extraoperative data (range: p < 10-27 to p < 10-19), suggesting that the networks highlighted in interictal GC maps from both data sets correlate with surgically-relevant seizure networks. Conclusions: GC analysis applied to sEEG interictal data collected intraoperatively at the time of implantation can reveal ictal networks, which is comparable to GC maps of postoperatively recorded interictal data. This may suggest that real-time GC analysis of sEEG could be useful in surgical decision-making: data obtained real-time after placement of a subset of sEEG electrodes may suggest modifications to the remaining sEEG electrode implantation plan, guiding subsequent electrode placement to better sample key areas needed to define the epileptogenic zone. Funding: This work is funded in part by NIH (1R01NS069696) and in part by Wyss Institute for network data storage and management.