Improving the neuromagnetic source-imaging workflow

Magnetoencephalography (MEG) is a non-invasive method for direct measurement of brain activity. Due to its excellent temporal resolution, good spatial resolution, and whole-head coverage, MEG has become an important tool for studying brain function. MEG provides multichannel time-series data, which is often transformed into an estimate of the active neural sources; this process is referred to as source imaging, and it is often supported by the structural magnetic resonance image (MRI) of the subject's head. This Thesis contributes to exploring and improving some of the factors affecting MEG source imaging. Although the Thesis focuses on MEG, the findings apply also to electroencephalography (EEG). Using structural MRIs in MEG source imaging requires their accurate co-registration with MEG, which involves digitizing the head surface and anatomical landmarks. Publication-I of the Thesis evaluates the performance of the state-of-the-art digitization system, explores the usability of two alternative systems, and suggests guidelines for accurate digitization in MEG and EEG. Further, Publication-II of the Thesis explores beamformer-based MEG source-imaging workflows in four major open-source software packages and investigates how their results differ when applied to identical datasets. Publication-III describes a tool for generating a pseudo-MRI based on the digitized head shape, thus omitting the need for MRI acquisition, and validates the use of such a pseudo-MRI in the MEG source-imaging workflow. In conclusion, this Thesis enhances the MEG source-imaging workflow by addressing critical aspects such as digitization accuracy and software-based discrepancies and enabling high-accuracy MRI-free source localization.