To analyze neural activity using magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI), we developed a method for fixing equivalent current dipoles of MEG in activation areas of fMRI. It includes a procedure for dividing large fMRI activation volumes into subvolumes in each of which a dipole is placed and another procedure for grouping neighboring dipoles whose temporal changes are inseparable based on MEG data. To optimize the procedures' parameters, we carried out simulations and found that (1) any single dipole within 10 mm from a true source can explain MEG data with a correlation of 94% on average for the low signal-to-noise ratio of 3 and (2) a neighboring dipole within a few tens of millimeters from the dipole nearest to the true source tends to be highly incorporated in explaining MEG data. We applied the method to data measured in a language experiment and detected 13 significant sources. The results show that the present method is promising for detecting neural activity originating from a number of separate neural sources. © 2002 Elsevier Science (USA)

No abstract