Estimation of human head tissue conductivities by using in vivo E/MEG data and comparison of three different estimation algorithm results

Knowledge of tissue conductivities is needed to construct reliable volume conductor models of the human body and the head in solving forward and inverse bioelectric field problems. In this study three different estimation algorithms are applied to in vivo human head tissue resistivity estimation by using EEG and MEG data. The applied algorithms are conventional least-squared error algorithm (LSEE), Bayesian MAP algorithm and statistically constrained minimum mean squared error estimator (MiMSEE). The algorithms intake a priori information on body geometry (realistic boundary element model), statistical properties of regional conductivities (assumed to be uniformly distributed between upper and lower bounds), linearization error and instrumentation noise. The EEG-MEG data set has been obtained from a source localization experiment in which the median nerve has been stimulated. The anatomical boundary information has been extracted from 256 T1-weighted MRI images. The MEG data have been obtained by using a 31-channel magnetometer over the somatosensory cortex. By using the data, scalp, skull and brain conductivities have been estimated and estimation variances are calculated by using the algorithms. It is shown that MiMSEE algorithm gives lower error rates than the other two algorithms. The calculated error rates are 90 % for the LSEE, 20.5 % for the Bayesian MAP estimator and 12.5 % for the MiMSEE.