Scalp EEG Functional Connectivity Reveals Precise Biomarkers of the Epileptogenic Zone in Children with Drug-resistant Epilepsy
Abstract number :
1.191
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2022
Submission ID :
2204254
Source :
www.aesnet.org
Presentation date :
12/3/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:24 AM
Authors :
Georgios Ntolkeras, MD – Baystate Medical Center, UMass Chan Medical School & Boston Children's Hospital, FNNDSC, Harvard Medical School; Navaneethakrishna Makaram, PhD – Newborn Medicine – Boston Children's Hospital, FNNDSC, Harvard Medical School; Matteo Bernabei, MS – Campus Biomedico University of Rome; Aime Cristina De La Vega, MS – Newborn Medicine – Boston Children's Hospital, FNNDSC, Harvard Medical School; Joseph Madsen, MD – Neurosurgery – Boston Children's Hospital, Department of Neurosurgery, Harvard Medical School; Phillip Pearl, MD – Boston Children's Hospital, Department of Neurology, Harvard Medical School; Christos Papadelis, PhD – Cook Children's Hospital, Texas Christian University School of Medicine; Ellen Grant, MD – Radiology – Boston Children's Hospital, FNNDSC, Harvard Medical School; Eleonora Tamilia, PhD – Newborn Medicine – Boston Children's Hospital, FNNDSC, Harvard Medical School
Rationale: For children with drug-resistant epilepsy (DRE) undergoing epilepsy surgery, delineating the epileptogenic zone (EZ) is critical. This can be estimated via various diagnostic tools, among which conventional scalp EEG is key, being the most well-established tool to characterize seizures, low-cost, noninvasive, and widely used in all centers. EEG is traditionally reviewed seeking visible biomarkers, like seizures or spikes. Yet, seizures are unpredictable and may need lengthy recordings and spikes are local biomarkers, while epilepsy is increasingly theorized as a brain network disorder. Thus Functional Connectivity (FC) between brain regions cannot be ignored as it helps reveal hypersynchronous seizure networks. _x000D_
Our goal is to introduce a novel EEG source imaging method to study the functional brain networks of children with DRE (at various frequencies), reveal new FC biomarkers of the EZ and develop a machine-learning (ML) tool that precisely locates the EZ before surgery using short interictal EEG data.
Methods: We studied 32 children (5.9±5.2 years) who had successful epilepsy surgery (Engel 1A; >1 year follow-up). We analyzed 5-min sleep scalp EEG data separating spike and silent EEG epochs (with or without spikes; Figure 1A). We reconstructed the activity of ~1,200 cortical vertices across the whole brain (source imaging, Figure 1B) and computed whole-brain FC in 6 frequencies (Figure 1B), obtaining 6 frequency-specific networks for silent and spike EEG data separately. We defined the EZ - as the patient’s resection - and three non-epileptogenic zones (NEZ) in the same and opposite hemisphere as Figure 1C shows. We defined regional FC features (Figure 1D) for each frequency: To identify which feature characterized the EZ, we compared values between the different zones (Wilcoxon sign-rank). We computed whole-brain FC features from each cortical point (Figure 1E) and used them to develop a ML classifier able to pinpoint the cortical points within the EZ (kNN; 10-fold cross-validation). We ran all analyses using either silent or spike EEG data.
Results: As Figure 2A shows, regional FC features (in various frequencies) differ between EZ and NEZs, demonstrating higher FC in the EZ whether estimated from silent or spike epochs. Analysis on spikes shows increased FC (in all frequencies) in the whole EZ hemisphere with higher amount in the EZ; while for silent EEG, increased FC (in fewer frequencies) was specific to the EZ not the hemisphere. Using whole-brain FC biomarkers, our ML system (Figure 2B) correctly identified 72%-73% of the EZ (sensitivity) with 99% specificity (74% precision and 74-73% accuracy) when using silent or spike EEG respectively.
Conclusions: We presented a novel FC approach that localizes the EZ with 74% precision using interictal scalp EEG with, or even without, spikes. Our data show that noninvasive EEG biomarkers of increased FC are unique signatures of the EZ in pediatric DRE: they allow us to pinpoint the EZ (area to resect) with 73% sensitivity and 99% specificity. This could boost the value of routine presurgical EEG in pediatric DRE, adding to the visual identification of seizures or spikes.
Funding: BCH/Harvard Faculty Career Development Fellowship; PI: Tamilia
Neurophysiology