A prerequisite to reliably assess functional connectivity is to rely on source space data, i.e., the activity of each neuronal pool. In fact, the estimation of functional connectivity from sensor space data, i.e., from the activities measured by the sensors outside the head, is not only prone to the above described fake connectivity profile but is also very difficult to be interpreted in terms of the underlying functional connectivity between brain regions given that the same connectivity profile can be generated by different configurations of interacting sources. Unfortunately, the estimation of the activities of brain regions requires the solution of an ill-posed electromagnetic inverse problem (Ilmoniemi and Sarvas, 2019). In practice, there exist reasonable approximations incorporating prior knowledge (e.g., from fMRI) and constraints (e.g., minimum norm) that make this issue less severe and allow the use of MEG in brain imaging and in the study of brain networks.
