The solution to the MEG inverse problem requires a forward model, which gives the mathematical relationship between the brain currents (sources) and the measured magnetic field. The forward model includes models for the sources, the geometry of the conducting medium (i.e., the head), and the MEG sensors. Besides the intrinsic non-uniqueness of the electromagnetic inverse problem solution and the correctness of a priori assumptions and constraints used to make the solution unique, the attainable accuracy in the solution highly depends on the accuracy of these models. The influence of the head model has been investigated in a number of studies (e.g., Güllmar et al., 2010; Stenroos et al., 2014; Vorwerk et al., 2014). The distinction between white and gray matter and the inclusion of the cerebrospinal fluid have been shown to considerably improve forward model accuracy (Vorwerk et al., 2014; Stenroos and Nummenmaa, 2016). To date, anatomically realistic and sufficiently detailed head models can be constructed from the segmentation of structural head images (usually MRI or computed tomography scans; for guidelines, see Vorwerk et al., 2014). Nonetheless, the forward model
