Inferring Surgical Targets in the Epileptogenic Network Using Stereoelectroencephalography Spatiotemporal Dynamics
Abstract number :
2.185
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2024
Submission ID :
1120
Source :
www.aesnet.org
Presentation date :
12/8/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: Hana Farzaneh, MD – Mass General Hospital
Darya Frank, PhD – Universidad Politécnica de Madrid
Peter Hadar, MD, MS – Massachusetts General Hospital and Harvard Medical School
George Plummer, MD – Massachusetts General Hospital/Harvard Medical School
Steven Stufflebeam, MD – Athinoula A. Martinos Center for Biomedical Imaging
Noam Peled, PhD – Athinoula A. Martinos Center for Biomedical Imaging
Rationale: Using stereoelectroencephalography (sEEG) to localize surgical targets is a critical step in the surgical planning routine for patients with drug-resistant epilepsy (DRE). Identifying the seizure onset zone often requires intense manual inspection of the data to identify the specific area to be treated surgically. Therefore, the outcomes of this process can result in overlooking the epileptogenic network that may underlie a greater chance of seizure freedom. Here, we report the results of a retrospective study of five DRE patients, using sEEG contacts as nodes of the epileptogenic network, and analyzed the neurodynamics of each ictal event to reveal the most critical nodes within the network. We hypothesized that targeting these nodes during surgery would yield better surgical outcomes.
Methods: We selected epilepsy patients with DRE who had an sEEG study, resective surgery, postoperative MRI scan, and follow-up at least one year after the surgery. All patients were seizure-free. For the analysis of sEEG data, we used ictal events, as identified by an epileptologist. Following preprocessing, which included notch filtering and bipolar montage, we estimated the likelihood of Granger causality between channels - one channel’s time series has a statistically significant effect on the other after a certain lag in time. All channels were within gray matter, and neighboring channels were filtered. We estimated the criticality of each node in the network as a function of its significant Granger causality values to all other nodes. To obtain an estimate of criticality across events, we averaged the criticality of each node.
Results: Within event, we found that 75% of events across patients had at least one of the top five critical nodes identified in the clinical report. In Figure 1, we show an example of the epileptogenic network from a single event in one patient, with the nodes colored by their criticality and edged by their causality values. Using the mean criticality values across events, we found that at least one of the top five most critical nodes was identified by the epileptologist as a seizure onset zone in four out of five patients. Of the total number of seizure onset zones identified in the clinical report, 62% of the most critical nodes across events were matched. Finally, we validated the results using the postsurgical MRI. In Figure 2, we present an example from the same patient, showing that the electrode with the most critical node was located within the resection zone, demonstrating the effectiveness of this approach.
Conclusions: In this study, we present a novel approach for analyzing the spatiotemporal characteristics of the epileptogenic network, as captured by sEEG, identifying critical nodes, and validating them based on clinical reports and postsurgical MRIs. In the next step, we will add more patients to our sample, including those with no significant improvement, where we predict that our method will identify critical nodes outside the resection zone, potentially offering a revised approach to surgical planning.
Funding: This research was sponsored by NIH (R21NS101373, R01NS104116, S10OD030469)
Neurophysiology