Pathological and Physiological High-frequency Oscillations on Whole-brain Noninvasive Recordings: Comparing Healthy Children and Patients with Epilepsy
Abstract number :
3.027
Submission category :
1. Basic Mechanisms / 1C. Electrophysiology/High frequency oscillations
Year :
2022
Submission ID :
2204015
Source :
www.aesnet.org
Presentation date :
12/5/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:23 AM
Authors :
Lorenzo Fabbri, BS – The University of Texas at Arlington; Cecilia Liberati, MS – Universitá Campus Bio-medico di Roma; Shannon Conrad, MS – Research Assistant, Jane and John Justin Neurosciences Center, Cook Childrens's Hospital; Deniz Aygun, BS – Clinical Research Assistant, Boston Children's Hospital; Steven Stufflebeam, MD – Massachusetts General Hospital; Phillip Pearl, MD – Director, Epilepsy and Clinical Neurophysiology, Boston Children's Hospital; Scott Perry, MD – Medical Director, Neurology, Cook Childrens's Hospital; Eleonora Tamilia, PhD – Harvard Medical School; Christos Papadelis, PhD – Director of Research, Jane and John Justin Neurosciences Center, Cook Childrens's Hospital
This abstract is a recipient of the Young Investigator Award
This abstract has been invited to present during the Basic Science Poster Highlights poster session
Rationale: High-frequency oscillations (HFOs) are among the most promising interictal biomarkers of epileptogenicity in drug-resistant epilepsy (DRE). Yet, their clinical value is still debated since the HFO-generating area is often larger than the actual epileptogenic zone due to the generation of physiological HFOs (physHFOs) in healthy brain regions. Previous studies have used intracranial EEG (iEEG) data to separate physHFOs from pathological HFOs (pathHFO) but suffered from the iEEG’s inability to obtain whole-brain coverage and to record physiological data from typically developing (TD) controls. Here, we aim to use full-head noninvasive methods to: (a) generate the first cortical map of physHFOs in the normal human brain; (b) identify differences between HFOs generated by the healthy and the epileptic brain (comparing TD and DRE data); and (c) develop an automated classifier of physHFOs.
Methods: We analyzed magnetoencephalography (MEG; 306 sensors) and high-density EEG (HD-EEG; 256 channels) data from 19 TD (11.9 ± 3.5 y, 8 male) and 14 children with DRE (13.5 ± 3.3 y, 8 male). We detected HFOs (ripples: 80-160 Hz) on MEG and HD-EEG separately using automated detection followed by visual review and localized their cortical sources using electric and magnetic source imaging (Figure 1A). We then extracted a broad set of temporal, spatial, morphological and spectral features from each HFO as listed in Figure 1B. HFO features were compared between TD and DRE group (Wilcoxon rank-sum test) and used to train and test (10-fold cross-validation) a k-nearest-neighbor (kNN) classifier for discriminating physHFOs and pathHFOs.
Results: We found more HFOs on HD-EEG than MEG in both the TD (218 vs. 53 HFOs; 80 vs 20%) and DRE group (120 vs. 37 HFOs; 76 vs. 24%). We generated 3D maps of physHFOs in the normal pediatric cortex demonstrating high rates in the somatosensory area for both HD-EEG and MEG (Fig 2A). For HD-EEG, all the HFO features showed differences between TD and DRE group (Figures 2B-E). HFOs in the TD group had a higher and more variable frequency than DRE group (Figure 2B, p< 0.01), weaker power content (Figure 2B, p< 0.001), shorter and less variable duration (Figure 2C, p< 0.01), lower and less variable amplitude (Figure 2D, p< 0.01), and larger spatial extent at the sensor and source level (Figure 2E, p< 0.05). Our kNN classifier showed a 73% accuracy in discriminating the EEG-HFOs from the TD and DRE group (Figure 2F; p< 0.001) and correctly classified 82% of the physHFOs. No differences were seen for MEG likely due to the lower number of HFOs, but MEG-HFOs showed smaller spatial extent and higher and more variable frequency than HD-EEG (p < 0.05) in both DRE and TD groups.
Basic Mechanisms