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

Area of Science:

• Neuroimaging
• Biophysics
• Computational Neuroscience

Background:

• Accurate head and source models are crucial for magnetoencephalography (MEG) and electroencephalography (EEG) source localization.
• When a subject's MRI is unavailable, pseudo MRI models are often created using standard templates or pooled scans.
• This study focuses on approximating pseudo MRI for MEG source imaging.

Purpose of the Study:

• To evaluate two methods for approximating pseudo MRI: first rank and averaged rank.
• To compare the accuracy of these pseudo MRI methods against real MRI for MEG source localization.
• To assess the impact of pseudo MRI on the identification of functional group source responses.

Main Methods:

• Two pseudo MRI approaches were developed: first rank (lowest registration error) and averaged rank (averaging registered models).
• Models were registered to the subject's digitized head surface using rigid body transformation and isotropic scaling.
• Subject-level and functional group analyses were performed to quantify source localization errors.

Main Results:

• Averaged rank pseudo MRI showed a significantly lower mean upper bound of source location error (10 ± 3 mm) compared to first rank pseudo MRI.
• Source distribution for functional group responses to visual stimulation was similar across first rank pseudo MRI, averaged rank pseudo MRI, and real MRI.
• Averaged rank pseudo MRI yielded significantly lower source localization error for functional peak responses than first rank pseudo MRI.

Conclusions:

• Approximated pseudo MRI, especially the averaged rank method, is a feasible substitute for real MRI in MEG source imaging.
• Using averaged rank pseudo MRI preserves the generality of functional group source response findings.
• This approach enhances the applicability of MEG/EEG source localization when real MRI data is inaccessible.