Schedule of Events | Symposia

Sketchpad Series

Comparison of ERP inverse solutions using MRI-informed versus standard forward models

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Mark Pflieger¹ (mark.pflieger@gmail.com), Kathryn Toffolo², John Foxe²; ¹San Diego State University, ²University of Rochester Medical Center

Any inverse method for estimating the brain current sources of EEG measurements requires a forward model which comprises: (i) a source domain model that specifies feasible locations, orientations, and prior probabilities of primary currents; (ii) a conductivity geometry model based on segmented head tissues with assigned conductivities; (iii) a numerical solver of Poisson’s equation that calculates how any given primary current generates electric potentials on the head surface; and (iv) a sensor configuration that specifies electrode locations and an EEG reference. Individual structural MRIs can inform elements (i), (ii), and (iv). Alternatively, study participants can share a standard forward model derived from average MRI data. We hypothesize that MRI-informed (versus standard) forward models spatially sharpen inverse solutions, thereby increasing both spatial specificity and statistical power for discriminating between experimental conditions. To address these hypotheses, we are using structural MRIs of 20 participants in a 128-channel ERP language study (PMID: 39369943) to construct finite element method (FEM) forward models with 5-tissue (skin, skull, CSF, gray matter, white matter) conductivity geometries, and three different source domains (intracranial lattice, gray-matter volume, cortically constrained) for further comparisons. The standard FEM model is derived from ICBM 2009b data (nonlinear, asymmetric) and the New York Head. For each case we shall construct statistical nonparametric maps of congruent versus incongruent (word ending) sLORETA inverse solutions. Then the negative log p-value maps obtained for MRI-informed versus standard models shall be compared at a second level. The results should assist ERP researchers to assess MRI benefits versus costs.

Topic Area: METHODS: Electrophysiology