Use of prior knowledge in brain electromagnetic source analysis

Using multichannel measurements of EEG and/or MEG, macroscopic source activities can be estimated in the human brain using brain electric source analysis (BESA, Scherg 1990). If a discrete number of brain areas is active, functional brain images which depict the locations and orientations of equivalent dipole sources as well as the dynamics of the local macroscopic currents can be obtained from such data--in principle--without external knowledge. However, given a certain number of sources or 'neural masses' (Freeman 1975) which contribute to an event related response (ERP), it can be difficult to find the correct solution due to background noise in the data and distortions from the head model. Prior knowledge based on anatomy and physiology can be useful to constrain spatial or temporal parameters of the model and to define better cost functions for fitting locations and orientations. An analysis of the auditory evoked N100 complex and of the auditory mismatch negativity (MMN) is presented which illustrates the use of spatial constraints. Also, the use of a modified cost function is demonstrated which limits source currents in certain time intervals.