Authors :
Presenting Author: Takfarinas Medani, PhD – University of Southern California
Anand Joshi, PhD – University of Southern California
Chinmay Chinara, MS – University of Southern California
Yash Vakilna, MS – The University of Texas Health Science Center
Wayne Mead, MS – The University of Texas Health Science Center
Raymundo Cassani, PhD – McGill University
Sylvain Baillet, PhD – McGill University
John Mosher, PhD – The University of Texas Health Science Center
Richard Leahy, PhD – University of Southern California
Rationale:
Scalp EEG is a non-invasive technique for studying the brain's electrical activity with high temporal resolution. EEG techniques are used in research and clinical applications, such as diagnosing Alzheimer’s disease and epilepsy. EEG allows the monitoring and recording of electrical brain activity from multiple electrodes placed on the scalp. EEG-based cortical current density mapping requires accurate knowledge of the locations of the electrodes on the scalp. The number and placement of electrodes vary from a few to high-density models with hundreds of electrodes. Researchers and clinicians have developed some solutions for precise electrode localization. The most common approach uses an electromagnetic digitizer (e.g. Polhemus). However, these methods are typically not easy to use, require skilled technicians, and the procedures are time-consuming and subject to errors. This study presents a structured light approach for mapping the scalp surface and automatically identifying each electrode's 3D location and label. We have implemented and integrated an automatic pipeline within our open-source Brainstorm software.Methods:
In this study, we used Revopoint, an affordable and advanced 3D scanner that uses structured-light technology. We first perform 3D scanning of the participant wearing an EEG cap with electrodes, Fig-1(a). The acquired scans are then minimally processed within the RevoScan software and imported into Brainstorm. The output of the scan is a 3D mesh of the scalp shown in Fig-1 (a). The locations of the electrodes on the mesh are identified by segmentation and texture mapping of the scalp and cap image to the mesh. The mesh is then flattened to 2D using a spherical to cartesian transform, Fig-1(b). The flattened 2D map of the electrode locations is mapped to the spatial configuration of the EEG cap using the manufacturer’s diagram, Fig-1(c). This is achieved by first matching the orientation of the 2D map to the cap using three anatomical fiducial points. Matching contacts between the 2D map and the diagram is then performed by nonlinear warping using an Iterative Closest Point (ICP) algorithm. A bending energy regularizer keeps the deformation smooth, Fig-1(d) and (e). Upon convergence, the electrode labels are transferred from the diagram to the scalp, resulting in the identified and labeled electrodes on the scalp, as shown in Fig-1 (f).Results:
This automatic process results in the 3D location and labels of the EEG electrodes, Fig-1(f). All processes are integrated within Brainstorm accessed through an interactive GUI panel, Fig-2. This interface is similar to the interface previously developed in Brainstorm for the Polhemus and Fastrack digitizers, allowing an easy transition for previous users and a fast learning process through online tutorials.Conclusions:
This study presents a structured-light approach using a 3D scanner for 3D scalp electrode localization and labeling. The method is implemented as an automated approach within the Brainstorm software. This represents a new low-cost, fast, easy-to-use approach for digitizing heads and electrodes.Funding: This work is supported by NIH grant R01EB026299 and DoD grant HT94252310149.