Asymmetric Weighting to Optimize Regional Sensitivity in Combined fMRI-MEG Maps

Functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) are neuroimaging techniques that measure inherently different physiological processes, resulting in complementary estimates of brain activity in different regions. Combining the maps generated by each technique could thus provide a richer understanding of brain activation. However, present approaches to integration rely on a priori assumptions, such as expected patterns of brain activation in a task, or use fMRI to bias localization of MEG sources, diminishing fMRI-invisible sources. We aimed to optimize sensitivity to neural activity by developing a novel method of integrating data from the two imaging techniques. We present a data-driven method of integration that weights fMRI and MEG imaging data by estimates of data quality for each technique and region. This method was applied to a verbal object recognition task. As predicted, the two imaging techniques demonstrated sensitivity to activation in different regions. Activity was seen using fMRI, but not MEG, throughout the medial temporal lobes. Conversely, activation was seen using MEG, but not fMRI, in more lateral and anterior temporal lobe regions. Both imaging techniques were sensitive to activation in the inferior frontal gyrus. Importantly, integration maps retained activation from individual activation maps, and showed an increase in the extent of activation, owing to greater sensitivity of the integration map than either fMRI or MEG alone.