Quantifying Transformations from Resting State to Stimulation Evoked Connectivity Networks
Abstract number :
2.192
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2024
Submission ID :
901
Source :
www.aesnet.org
Presentation date :
12/8/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: Caila Coyne, MS – University of Alabama at Birmingham
Helen Brinyark, BS – University of Alabama at Birmingham
Benjamin Cox, MD – University of Alabama at Birmingham
Sridevi Sarma, PhD – Johns Hopkins University
Sahaj Patel, PhD – University of Alabama at Birmingham
Arie Nakhmani, PhD – University of Alabama at Birmingham
Rachel Smith, PhD – University of Alabama at Birmingham
Rationale: This study aims to develop a framework for comparing resting state connectivity networks to stimulation evoked networks in drug resistant epilepsy patients. Single-pulse electrical stimulation (SPES) has become increasingly used to investigate functional and pathological connectivity in epilepsy and to probe cortical excitability. The stimulation-evoked responses, cortico-cortical evoked potentials (CCEPs), often differ in magnitude in epileptogenic and healthy brain regions. The development of dynamical network models to describe CCEP and interictal (resting state) data has allowed us to analyze model properties that have clinical significance. The resting state model takes the form: 𝒙[𝑡+1]=𝑨RS𝒙[𝑡] and is constructed from 500 ms windows where 𝒙[𝑡]∈ℝnx1 is the state vector describing the neuronal activity from each of the 𝑛 contacts, and 𝑨RS∈ℝ𝑛x𝑛 is the state transition matrix. To estimate 𝑨CCEP, an exogenous perturbation term is added to the model to account for the electrical input: 𝒙[𝑡+1]=𝑨CCEP𝒙[𝑡]+𝑩u[𝑡]. 𝑨CCEP and 𝑩 are then estimated simultaneously using average CCEP waveforms as the state vector.
Methods: To quantify similarity in connectivity between resting and stimulated networks, we calculate a transformation matrix T where 𝑨CCEP = T𝑨RS. We then calculate the root-mean-square (RMS) of T to define network similarity. We expect the RMS of T to be minimized with greater similarity between the resting state and stimulation evoked network. T is calculated for the comparison of numerous resting state windows to unique stimulated responses resulting in distributions of RMS values that are then normalized by the median interictal to interictal transformation RMS to account for the variation in artifact free contacts in the CCEP models.
Results: We find that the RMS of T can highlight different neural subnetworks. Differences in the metric generally align with the clinical annotations of contacts in epileptogenic tissue. In the representative patient (Figure 1), the RMS of T distributions for seizure onset and early propagation zone contacts are distinct, suggesting activation of a specific subnetwork. The separation also aligns with percent overlap in the channels that responded significantly to a given stimulation as defined by the N1 amplitude. For example, SOZ contact LAH1_2 has a mean RMS of -0.0324 and only ~33% overlap of significant response channels with non-SOZ contacts LAH3_4 (RMS of -0.0295) and LAH7_8 (RMS of -0.0288) which have 60% overlap with each other.
Conclusions: We are developing this framework to transition between resting state network models and stimulation evoked network models to leverage the clinical benefits of the CCEP response in seizure onset localization. Construction of a transformation matrix defined from only resting state data may allow insight into seizure propagation and effective connectivity and may reduce reliance on performing electrical stimulation during intracranial monitoring.
Funding: This work was funded by CURE Epilepsy Taking Flight Award (1061181).
Neurophysiology