Abstract
Cortical dipole imaging has been proposed as a method to visualize electroencephalogram (EEG) in high spatial resolution. In the present study, we investigated the inverse estimation of the cortical dipole imaging. The truncated total least squares (TTLS) was applied to the inverse estimation to reduce the influence from both the measurement noise and the transfer matrix error caused by the head model distortion. The effect of the regularization parameter was evaluated by the relative error in a computer simulation. The estimation accuracy was improved by the TTLS compared with Tikhonov regularization. The proposed regularization technique was applied to human experimental data of visual evoked potentials. The TTLS improved the spatial resolution similar to the computer simulation.
