A maximum entropy method for MEG source imaging

The estimation of three-dimensional dipole current sources on the cortical surface from the measured magnetoencephalogram (MEG) is a highly underdetermined inverse problem as there are many “feasible” images which are consistent with the MEG data. Previous approaches to this problem have concentrated on the use of weighted minimum norm inverse methods. While these methods ensure a unique solution, they often produce overly smoothed solutions and exhibit severe sensitivity to noise. Here, the authors explore the maximum entropy approach to obtain better solutions to the problem. This estimation technique selects that image from the possible set of feasible images which has the maximum entropy permitted by the information available. In order to account for the presence of noise in the data, the authors have also incorporated a noise rejection or likelihood term into their maximum entropy method. This makes their approach mirror a Bayesian maximum a posteriori (MAP) formulation. Additional information from other functional techniques like functional magnetic resonance imaging (fMRI) can be incorporated in the proposed method in the form of a prior bias function to improve solutions. The authors demonstrate the method with experimental phantom data from a clinical 122 channel MEG system