PROBABILISTIC RECONSTRUCTION OF MULTIPLE SOURCES IN THE BIOELECTROMAGNETIC INVERSE PROBLEM

A probabilistic multiple source solution for the bioelcctromagnetic inverse problem is described. The model-dependent solution assumes a finite number of discrete primary sources at fixed locations within a bounded conductor. Covariance statistics derived from a set of detectors outside the conducting region are used to determine a metric on the space of possible sources. This metric function is used to construct a weighted pseudo-inverse matrix, which, in turn, may be used to estimate the spatio-temporal distribution of source activity. The results are embodied in the form of the PROMS (probabilistic reconstruction of multiple sources) algorithm. Computer simulations using the algorithm are described. These methods are compared with other algorithms, including minimum norm estimation, and the MUSIC and spatial filtering algorithms.