Modeling extended sources using a maximum likelihood estimator

For the study of functional organization and reorganization of the human cortex by means of electromagnetic source imaging, changes in the location and spatial extent of neural sources is of interest. We present a probabilistic approach based on the maximum likelihood (ML) theory for modeling extended sources using anatomical and physiological knowledge to reduce the ambiguity of the ill-posed biomagnetic inverse problem. The number of sources, their location and extension is defined by a set of parameters which, due to measurement noise, is associated with a certain probability distribution defining the likelihood function (LF). The parameter set which maximizes the LF is determined by using the Metropolis algorithm. To investigate the influence of different levels of uncorrelated and correlated noise on the outcome of the proposed method, simulations with an activity located on both sides of an artificial sulcus were performed. The applicability of the proposed method is presented as an application on experimental somatosensory evoked magnetic fields. Our tests demonstrate the applicability of the presented method in locating extended sources and quantifying their extension.