Frequency Domain Distributed Inverse Solutions

Abstract

Much interest has been generated in the past few years by Distributed Solutions that estimate the primarycurrent distributions in the whole brain volume. This paper presents a LORETA based distributed method in theFrequency Domain. The method assumes a stochastic stationary time series model for generators that is masked byinstrumental noise. Tikhonov regularization is used for dealing with the noise component. A commonregularization parameter is computed by the generalized crossvalidation procedure for the complex valued FFTcoefficients. It is shown by simulations that the method is effective for localising generators from noisy data andthat a simple LORETA without regularization may fail to recover the generator spectra.