Approximate Subject Specific Pseudo MRI from an Available MRI Dataset for MEG Source Imaging

Abstract

Computation of headmodel and sourcemodel from the subject's MRI scan is an essential step for source localization of magnetoencephalography (MEG) (or EEG) sensor signals. In the absence of a real MRI scan, pseudo MRI (i.e., associated headmodel and sourcemodel) is often approximated from an available standard MRI template or pool of MRI scans considering the subject's digitized head surface. In the present study, we approximated two types of pseudo MRI (i.e., associated headmodel and sourcemodel) using an available pool of MRI scans with the focus on MEG source imaging. The first was the first rank pseudo MRI; that is, the MRI scan in the dataset having the lowest objective registration error (ORE) after being registered (rigid body transformation with isotropic scaling) to the subject's digitized head surface. The second was the averaged rank pseudo MRI that is generated by averaging of headmodels and sourcemodels from multiple MRI scans respectively, after being registered to the subject's digitized head surface. Subject level analysis showed that the mean upper bound of source location error for the approximated sourcemodel in reference to the real one was 10 ± 3 mm for the averaged rank pseudo MRI, which was significantly lower than the first rank pseudo MRI approach. Functional group source response in the brain to visual stimulation in the form of event-related power (ERP) at the time latency of peak amplitude showed noticeably identical source distribution for first rank pseudo MRI, averaged rank pseudo MRI, and real MRI. The source localization error for functional peak response was significantly lower for averaged rank pseudo MRI compared to first rank pseudo MRI. We conclude that it is feasible to use approximated pseudo MRI, particularly the averaged rank pseudo MRI, as a substitute for real MRI without losing the generality of the functional group source response.

Reduced Auditory Steady State Responses in Autism Spectrum Disorder

“…MEG data were co-registered with participants' structural MRIs by matching the digitised head-shape data with surface data from the structural scan [32]. Two control participants did not complete a T1 structural MRI and therefore a size-matched, template-MRI was used [33,34]. The aligned MRI-MEG images were used to create a forward model based on a single-shell description of the inner surface of the skull [35], using the segmentation function in SPM8 [36].…”

Dysregulated Oscillatory Connectivity in the Visual System in Autism Spectrum Disorder

“…MEG data were co-registered with participants MRI structural scan by matching the digitised head-shape data with surface data from the structural scan . Two control participants did not complete a T1 structural MRI and therefore a pseudo-MRI was used, see Gohel et al, (2017) for full procedure. The aligned MRI-MEG images were used to create a forward model based on a single-shell description of the inner surface of the skull (Nolte, 2003), using the segmentation function in SPM8 (Litvak et al, 2011).…”

Dorsal Root Ganglion Stimulation Modulates Cortical Gamma Activity in the Cognitive Dimension of Chronic Pain

“…The implanted DRG stimulators used were not MRI compatible and, as such, individual structural MRIs (pre-or post-operative) were not available. Therefore, the ICB152 MRI template in Brainstorm was warped to fit the head model of each participant by co-registering the nasion, left and right pre-auricular fixed points acquired during head shape digitization [26]. Each subject-specific template was then used to calculate a lead field matrix based on a single shell model.…”