Spectral Granger Causality for Ictal Network Study Using Magnetoencephalography Data
Abstract number :
3.254
Submission category :
5. Neuro Imaging / 5B. Functional Imaging
Year :
2022
Submission ID :
2204862
Source :
www.aesnet.org
Presentation date :
12/5/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:26 AM
Authors :
Natascha Cardoso da Fonseca, MD, PhD – University of Texas Southwestern Medical Center; Sasha Alick-Lindstrom, MD – Assistant Professor, Neurology, University of Texas SOuthwestern Medical Center; Amy Proskovec, PhD – Assistant Professor, Radiology, University of Texas SOuthwestern Medical Center; Pegah Askari, BS – Graduate Student Researcher, Radiology, University of Texas Southwestern Medical Center; Tyrell Pruit, PhD – Postdoctoral Researcher, Radiology, University of Texas Southwestern Medical Center; Irina Podkorytova, MD – Assistant Professor, Neurology, University of Texas Southwestern Medical Center; Andrea Lowden, MD – Assistant Professor, Pediatrics - Neurology, University of Texas Southwestern Medical Center; Afsaneh Talai, MD – Assistant Professor, Pediatrics - Neurology, University of Texas Southwestern Medical Center; Joseph Maldjian, MD – Professor & Division Chief, Radiology, University of Texas Southwestern Medical Center; Elizabeth Davenport, PhD – Assistant Professor, Radiology, University of Texas Southwestern Medical Center
This abstract has been invited to present during the Broadening Representation Inclusion and Diversity by Growing Equity (BRIDGE) poster session
Rationale: Granger Causality (GC) models can predict directional connections, wherefore it can be useful to better understand ictal networks. Few studies have performed spectral analysis of GC measures with Magnetoencephalography (MEG) data. MEG is a non-invasive technique that has been increasingly used during presurgical evaluation. In this study, we evaluate if Spectral Bivariate Granger Causality (SBGC) could be a valuable tool to analyze the epilepsy network in ictal MEG data.
Methods: Ictal MEG data from epilepsy patients were exported to Brainstorm Software with their corresponding T1 weighted magnetic resonance imaging. Preprocessing included notch (60 Hz and harmonics) and band-pass (0.5-100 Hz) filters. Visual inspection performed by clinical neurophysiologists determined the time and frequency range of interest. Following this, linear time-frequency (Morlet Wavelets) decomposition and power spectrum density (Fast Fourier Transform) were performed to refine the time segment and the frequency band of interest for each seizure. Source imaging was performed using Linearly Constrained Minimum Variance (LCMV) beamforming at the selected frequency band. SBGC connectivity maps were generated in source space using the initial 4 seconds of the seizure and the Desikan-Killiany atlas. We used a time segment without epileptiform discharge as baseline for noise and data convariance calculation. We correlated our results to intracranial EEG results and/or resection data when available. We averaged the intensity of outflow connections from each ROI (region of interest) and compared the regions with a stronger intensity in a lobar and sublobar (ROI) level.
Results: SBGC analysis was performed in 7 seizures (5 from neocortical and 2 from insular sources) from 6 patients (7 to 41 years) (Table 1). In 6/7 of the cases, the lobe with stronger averaged SBGC intensity was concordant with the epileptogenic zone, five of these with statistical significance (Kruskal Wallis rank sum, p< 0.005). The seventh case was a deep source in which the strongest outflow localized to contralateral deep structures. Also, the ROI with the strongest intensity was part of the epileptogenic zone and comprised or was near to the SOZ (Seizure Onset Zone) in 5/7 cases. Additionally, applying a threshold 85% of the maximum intensity to the graph maps, all the pediatric and insular seizures showed bilateral connections, while the three adult neocortical seizures were only connected within the hemisphere comprising the epileptic zone. In one case, we could observe the connections between two probable SOZs with the outflow from the most active source (Figure 1).
Neuro Imaging