A Symmetrical Bayesian Model for
fMRI and EEG/MEG Neuroimage Fusion.

Figure 1. Basis for the simulations. Time evolution of the ideal PCD () and NMD () used for generating the simulated EEG and fMRI data, at the point of maximum neural activity.

Figure 2. Localization error (in millimeters) as a function of the eccentricity of the source, for LORETA and the EEG↔ fMRI method (fusion). Here the dotted lines represent the standard error for both curves.

Figure 3. FWHM of the PSF (in millimeters) as a function of the eccentricity of the source for LORETA and the EEG↔ fMRI method (fusion).

Figure 4. Visibility (on a lograithmic scale) as a function of the eccentricity of the source, for LORETA and the EEG↔ fMRI method (fusion)

Figure 5. Three orthogonal planes map of the ideal  used for simulating the EEG data for a left temporal dipole source, at the point of maximum neural activity.

Figure  6. Temporal behavior of the estimated activation variable (A), at the point of maximum neural activity, for a simulated left temporal dipole source.

Figure 7.  Reconstruction of the temporal dynamics for the estimated NMD () for a simulated left temporal dipole source, at the point of maximum neural activity.

Figure 8. Temporal evolution of the absolute value of the estimated PCD, for a simulated left temporal dipole source, at the point of maximum neural activity.

Figure 9.Three orthogonal planes map of the LORETA solution for a simulated left temporal dipole source. This map corresponds to the time instant of best signal to noise ratio.

Figure 10. Three orthogonal plpanes map of the PCD obtained with fMRIàEEG  method, for a simulated left temporal dipole source. The time instant of best signal to noise ratio is represented here.

Figure 11. Three orthogonal planes map of the estimated PCD using EEG↔ fMRI method for a simulated left temporal dipole source. Here, the time instant of best sigal to noise ratio was choused.

Figure 12. Three Orthogonal planes map of the ideal PCD used for generating te simulated data corresponding to a distributed left occipital source, at the oint the maximum neural activity.

Figure 13.  Tthree orthogonal planes map of the LORETA solution for a simulated distributed left occipital source, at the point of maximum neural activity. Here the time instant of best signal to noise ratio was chosen.

Figure 14. Three orthogonal planes map of the PCD obtained with fMRIàEEG method for a simulated distributed left occipital source. Here the time instant of best signal to noise ratio is represented.

Figure 15. Three orthogonal planes map of the estimated PCD using the EEG↔ fMRI method for a simulated distributed left occipital source. Here the time instant of best signal to noise ratio is choused.

Figure 16. Hotelling’s T2 statistic over the MEG sensors evaluated at each time point. TheWorsley  gnificance level (Worsley threshold) was used for thresholding the statistic. For comparative purposes the univariate threshold was also ploted.

Figure 17. Somatosensory MEG and fMRI data for the 41ms time instant. a) The SPM of the fMRI data for a somatosensory experiment. b) Inverse Solution for the MEG data of the somatosensory experiment using LORETA constrained to the cortical surface. c) SPM of the symmetric fusion method, constrained to the cortical surface.