Background: Neuroimaging techniques with high spatio-temporal resolution would are crucial for the advancement in brain research, improvement of clinical diagnosis and management of neuropsychiatric disorders. Functional MRI (fMRI) is characterized by its high spatial resolution. On the other hand, the techniques measuring electromagnetic features of neurons, such as electroencephalography (EEG), provide millisecond order temporal resolution. Therefore, integrative analyses of the fMRI and EEG are expected to provide information with high spatio-temporal resolution enabling to clarify dynamic multiple cortical activities
Objective: We propose a novel fMRI-EEG integrative reconstruction method for multiple cortical activities using EEG data, and we validate the accuracy of our method by comparing it with other popular reconstruction approaches that are assumed to have obtained prior information from fMRI.
Methods: We determined the first model via fMRI data, and we obtained the final model which contained the source that the fMRI could not capture through iterative model selection procedures based on the Akaike information criterion (AIC). We then used a linearly constrained generalized least-squares (LCGLS) filter to suppress unconscious activities. We carried out numerical simulations to validate the proposed method and compared it to two commonly used representative reconstructions method, sLORETA and the LCMV beamformer methods, using the residual sum of the squares.
Results: The proposed method gave a good estimation of the multiple cortical activities by suppressing other fMRI-visible and fMRI-invisible sources.
Conclusion: These results demonstrate that the proposed method can reconstruct cortical activities more accurately than either sLORETA or the LCMV beamformer methods.