Magnetoencephalographic data indicates separate component modeling is necessary in evaluating evoked responses

A double Gaussian model was used to estimate dipole source contributions to magnetocephalograms. The method was tested using four 2-mm-long dipoles fixed in conductive material enclosed within the human skull. Evoked responses were simulated by Gaussian activation of one dipole with a mean latency of 100 ms and a second dipole at 200 ms. Simulations were performed over the left temporal bone and the occipital bone. Source contributions to each resultant waveform were estimated as the amplitude parameters of a best-fit double Gaussian function. Comparing dipole inverse solutions from waveform peaks with those from estimated contributions revealed only a small difference in localization accuracy. Source strength is slightly better estimated using the component-model method. The peak picking method overestimated strength when dipoles were aligned and underestimated when they were opposed