Comparing RBF and BP neural networks in dipole localization

We propose to use a Radial Basis Function (RBF) network for source localization in the brain, and systematically compare its performance with that of the backpropagation neural network (BPNN). Also, these two neural network techniques are compared with a conventional technique, the Levenberg-Marquardt (LM) technique. We conclude that the RBF and BPNN techniques are complementary to each other in that each one excels in estimating different source parameters. Both network techniques are superior to the LM technique in the presence of noisy data typical of clinical EEG measurements.