Eeg-based Dynamic Network Analysis of Seizures
Abstract number :
2.18
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2024
Submission ID :
685
Source :
www.aesnet.org
Presentation date :
12/8/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: James Chen, MD, PhD – VAGLAHS/UCLA
Cindy Le, BS – VAGLAHS
Rationale:
Independent component analysis (ICA) can decompose EEG signals and statistically separate the independent EEG generators that are spatially fixed but temporally independent. Using the algorithms of ICA, dipole fitting, and Granger causality with an EEG epoch of 2.5 seconds to ensure that the data is stationary, an EEG-based dynamic network analysis can be constructed with multiple linked epochs to represent acute seizures and other EEG patterns, such as alpha rhythms, sleep spindles, vertex sharp transients, triphasic waves, and status epilepticus.
Methods:
The EEG data from the clinical study was processed using the open-source EEGLAB toolbox for ICA and dipole fitting. Our custom-developed MATLAB scripts were used to compute the directional F factors (in the F statistics) from Granger causality analysis among each pair of independent dipoles, which was plotted as a 100-ms snapshot. A video clip for reviewing the dynamic network was assembled from the sequential 100-ms snapshots. An integrated view of all the epileptic dipoles was gathered and further analyzed using cluster analysis with projections onto the brain MRI images.
Results:
Based on the above computational model of dynamic network analysis, the scalp EEG patterns during a seizure seem to arise from complex evolving interplays among the epileptic dipoles (n=20). The spreading of the seizure to different cortical and subcortical regions can be evaluated over time. The epileptic dipoles can be divided into neocortical vs. subcortical clusters. In all the seizures analyzed, there is always a presence of both the cortical and subcortical clusters, suggesting that most seizures require the interplay of the cortical epileptic zone with subcortical volleys.
Conclusions:
EEG-based dynamic network analysis is created from established algorithms of ICA, dipole fitting, Granger causality analysis, and cluster analysis. It provides a new method of inspecting acute seizures from the perspective of a dynamic epileptic network. Further study will elucidate its clinical applications.
Funding:
EP190044, CDMRP
Neurophysiology