Career Center AES Connect

Abstracts

Automatic Seeg Contact Localization and Labeling Using Brainstorm

Abstract number : 1.293
Submission category : 3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year : 2024
Submission ID : 934
Source : www.aesnet.org
Presentation date : 12/7/2024 12:00:00 AM
Published date :

Authors :
Presenting Author: Chinmay Chinara, MS – University of Southern California

Takfarinas Medani, PhD – University of Southern California
Anand Joshi, PhD – University of Southern California
Raymundo Cassani, PhD – McGill University
François Tadel, MSc – Independent Research Engineer
Samuel Medina Villalon, Research Engineer – Aix-Marseille University, Marseille, France
Kenneth Taylor, PhD – Cleveland Clinic
Yash Vakilna, MS – The University of Texas Health Science Center
Johnson Hampson, MSBME – University of Texas, UT Health Houston
Christian Benar, Eng, PhD – Aix-Marseille University, Marseille, France
John Mosher, PhD – The University of Texas Health Science Center
Sylvain Baillet, PhD – McGill University
Dileep Nair, MD – Cleveland Clinic Foundation
Richard Leahy, PhD – University of Southern California

Rationale: Stereo-EEG (sEEG) is a powerful tool in neurosurgery that allows seizure analysis and accurate localization of the seizure onset zone (SOZ). Analysis of sEEG data is often done by acquiring pre-implantation MRI (pre-MRI), post-implantation CT (post-CT), and post-surgery MRI (post-MRI) scans. This analysis requires: (1) post-CT to pre-MRI coregistration and (2) sEEG contact localization and labeling from post-CT. Automating these steps can improve accuracy using a standardized method, reduce intra- and inter-rater variabilities, and reduce cost and time by streamlining the workflow. Here, we present automation tools from our open-source software Brainstorm for these tasks.

Methods: To link electrode contact coordinates to neuroanatomy, the post-CT (moving) is coregistered rigidly with the pre-MRI (reference), using the correlation ratio cost function. This is followed by pre-MRI-based skull-stripping of post-CT to remove extracranial regions as well as the wire-cluster of the sEEG electrodes. The post-CT is then thresholded to find candidate contacts. Brainstorm's automatic contact localization pipeline uses the GARDEL software integrated within Brainstorm as a plugin. Brainstorm includes interactive editing features and also provides an alternate semi-automatic approach to contact localization based on manual marking of electrode tip and skull entry points and individual contact editing. Automatic labeling of the anatomical locations of each contact can then be performed using one of the brain atlases integrated within Brainstorm.

Results: The result of this processing is a set of labeled contact locations identified in pre-MRI and post-CT space. Fig-1 (a) shows the post-CT registered to pre-MRI. Fig-1 (b) shows the thresholded mesh obtained post-CT that defines the candidate contacts. Fig-1 (c) shows the sEEG contacts detected automatically using GARDEL and the labels overlayed on the pre-MRI in 3D and 2D. Fig-2 shows the Brainstorm GUI that is used to perform the tasks.

Conclusions: The sEEG tools in Brainstorm provide an integrated workflow for sEEG related image and data processing and analysis. The (semi-)automated methods provide improved efficiency and consistency. Interactive features also provide flexibility for fine-tuning the results of the automated method.

Funding: This work is supported by NIH grant R01EB026299 and DoD grant HT94252310149.

Neurophysiology