Synchronization Network-based Approach for Accurate Epileptogenic Zone Identification from Short Interictal Intracranial EEG Data
Abstract number :
2.195
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2024
Submission ID :
1149
Source :
www.aesnet.org
Presentation date :
12/8/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: Tonmoy Monsoor, PhD – University of California, Los Angeles
Atsuro Daida, MD,PhD – UCLA Mattel Children's Hospital
Prateik Shinha, BS – University of California, Los Angeles
Yipeng Zhang, Dr. – UCLA
Lawrence Liu, BS – University of California, Los Angeles
Naoto Kuroda, MD – Wayne State University
Shingo Oana, MD, PhD – University of California, Los Angeles
Sotaro Kanai, MD, PhD – Tottori University, Faculty of Medicine
Gaurav Shingh, BS – University of California, Los Angeles
Shaun Hussain, MD, MS – UCLA Mattel Children's Hospital, David Geffen School of Medicine
Raman Sankar, MD, PhD – University of California, Los Angeles
Aria Fallah, MD, MS – UCLA Mattel Children's Hospital
Richard Staba, PhD – University of California, Los Angeles
Jerome Engel Jr., MD, PhD – University of California, Los Angeles
Eishi Asano, MD/PhD – Wayne State University
Vwani Roychowdhury, PhD – UCLA
Hiroki Nariai, MD, PhD, MS – UCLA Mattel Children's Hospital
Rationale: Identifying the epileptogenic zone (EZ) in patients with medication-resistant epilepsy using brief interictal EEG segments remains a significant and unresolved challenge. Current clinical practice involves monitoring patients with intracranial EEG (iEEG) for periods of up to several weeks and analyzing multiple ictal segments to determine the seizure onset zone (SOZ). We hypothesize that all channels belonging to the EZ have subtle signatures of ictogenesis embedded within short interictal EEG segment, which can enable accurate estimation of the EZ.
Methods: We studied 159 pediatric patients at UCLA and Wayne State University who underwent chronic iEEG monitoring with grid/strip who underwent subsequent resection. All patients had SOZ annotations and postoperative seizure outcomes assessed after one year. We computed a directed and weighted synchronization coefficient between pairs of channels using power-phase coupling in high-frequency bands (gamma: 50-80 Hz, ripple: 80-250 Hz, and fast ripple: 250-300 Hz) from a 5-minute interictal EEG data (sampled at 1,000 Hz) obtained on the first night of the iEEG implant for each subject. This produced a synchronization network among channels for a given interictal segment and a sequence of such networks for the entire duration of the recording. For each channel, we computed a feature vector by leveraging the temporal dynamics of the network properties. These feature vectors were used to train and test a random forest model to estimate the likelihood of each channel belonging to the SOZ. During inference, predicted SOZ within the resected brain regions but not labeled as SOZ were defined as potential SOZ (PSOZ). This model was evaluated using leave-one-out cross-validation. Lastly, we built a prediction model incorporating the resection status of such estimated EZ (SOZ+PSOZ) to predict the likelihood of postoperative seizure freedom.
Results: The interictal synchronization networks frequently became hypersynchronous, driven by a single channel or several channels (Figure 1). The model identified the SOZ channels with an F1 score of 98% (comparison: marked SOZ vs. preserved brain regions marked as non-SOZ). Moreover, the model also identified PSOZ channels within the resected brain regions that were not labeled SOZ. Our postoperative seizure outcome prediction based on the resection status of estimated EZ yielded a mean F1 score of 86% (p = 2.5e-7) (Figure 2), which was significantly higher than the predictive power of current practice utilizing the resection status of the SOZ (F1 score = 78%, p=1.3e-5).
Conclusions: Since seizure episodes are characterized by large-scale neuronal synchronization, we hypothesize that the observed interictal hypersynchronous events are ‘mini seizures.’ This interictal synchronization network-based approach can potentially enable accurate estimation of the EZ from short interictal EEG data. This approach has the potential to reduce the duration of iEEG monitoring needed to decide on a resection plan.
Funding: The National Institute of Health, K23NS128318
Neurophysiology