EEG Source Localization of Stereotactic EEG Using Cortico-cortical Evoked Potentials: Methods and Proof of Concept
Abstract number :
3.167
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2023
Submission ID :
933
Source :
www.aesnet.org
Presentation date :
12/4/2023 12:00:00 AM
Published date :
Authors :
Presenting Author: Benjamin Cox, MD – University of Alabama at Birmingham
Rachel Smith, PhD – Assistant Professor, Engineering, University of Alabama at Birmingham; Ismail Mohamed, MD – Professor, Pediatric Neurology, Children's of Alabama; Jerzy Szaflarski, MD, PhD – Professor, Neurology, University of Alabama at Birmingham
Rationale: EEG source localization (ESL) of scalp EEG is an established technique for localization of focal seizures. 1-4However, ESL has not been adequately studied using intracranial EEG. Intracranial EEG (icEEG) data has an inherently higher signal to noise ratio and is also free from assumptions regarding conductivity of intervening skull, CSF, and subcutaneous tissue, however, it also offers more targeted and non-homologous spatial sampling of electrocerebral activity. Many of the commonly used ESL algorithms were originally studied for scalp EEG and their translation to icEEG is not certain, given differences in sensor location and spatial sampling. Cortico-cortical evoked potentials present an ideal technique for evaluating ESL algorithms for icEEG given the precise source location is already known.
Methods: We preformed cortico-cortical evoked potentials (CCEPs) on a single patient with stereo EEG in gray matter not involved in seizure onset or propagation. Evoked potentials were averaged for each stimulation pair using Curry 8 (Compumedics, Charlotte, NC) and filtered to remove shock artifact. The stimulated electrodes were deselected and the evoked potential was analyzed at the earliest peak. The forward solution was solved with the finite element method (FEM) using the MPRAGE sequence of the patient’s MRI. The inverse solution was solved using the following: simple dipole solution, minimum norm, sLORETA, L1 norm, LP norm, SWARM, eLORETA, swLORETA, ssLOFO, FOCUSS, LAURA, and LORETA algorithms. The distance from the ESL result (or maxima, for current density algorithms) to the midpoint between the two stimulated electrodes was measured and was compared using the Friedman test.
Results: A total of nine stimulation locations were used involving hippocampus, temporal pole, cingulate, amygdala, and inferior temporal gyrus in a patient with a bitemporal sEEG implantation. Mean distance to the midpoint of the stimulation pair was 9.9mm for dipole solution, 17.0 for minimum norm, 13.5 for sLORETA, 13.7mm for L1 norm, 17.4mm for Lp Norm, 15.9mm for SWARM, 20.2mm for eLORETA, 15.2mm for swLORETA, 15.6mm for ssLOFO, 15.3mm for FOCUSS, 20.6mm for LAURA, and 24.9mm for LORETA (see Fig. 1 for examples) with no statistical difference (p=0.34), although there was a trend towards greater discordance with LORETA, eLORETA, and LAURA algorithms. LAURA , LORETA, and LP norm required multiple adjustments of lambda during regularization, making these calculations more time consuming. Minimum norm, L1 norm, SWARM, SSLOFO, and FOCUSS tended to have more focal, constrained solutions, whereas sLORETA, eLORETA, and swLORETA had more diffuse solutions.
Conclusions: This proof of concept study demonstrates that CCEPs are a potentially useful tool for evaluating EEG source localization algorithms for icEEG with the known location of electrical source providing an ideal gold standard. While there was no difference in terms of distance from stimulation source in our small series, larger studies across multiple patients are needed. Certain algorithms provide a more constrained solution, whereas others tended to have a wider solution, which may be important for applying these algorithms to ictal activity.
Funding: None
Neurophysiology