Maximizing the imaginary part of coherency to detect cortico-cortical interactions

Abstract

Event Abstract Back to Event Maximizing the imaginary part of coherency to detect cortico-cortical interactions Laura Marzetti1*, Stefania D. Penna1, Vittorio Pizzella1, Gian L. Romani1 and Guido Nolte2 1 G. D`Annunzio University Foundation, ITAB, Italy 2 Fraunhofer FIRST, Athena project team, Germany In order to understand brain function, it is fundamental to describe interactions between functionally specialized brain regions. To this end, MEG is an ideal technique thanks to its whole head coverage and temporal resolution that allow reconstruction of source activity at millisecond scale. Thus, neuronal interactions can be studied on the estimated brain source rather than at sensor level where severe limitations arise by volume conduction or field spread self-interaction effects[1]. Nevertheless, since this effect can never be completely abolished because unmixing of the sources bis never perfect, we estimate seed based cortico-cortical interaction by the imaginary part of coherency (IPC) which cannot be generated by independent sources [1]. When estimating brain sources without introducing anatomical constraints on dipole orientation, e.g. limiting sources to be normal to the cortex surface, brain signals along x, y and z spatial directions are obtained at each voxel using weighted minimum norm solutions. Principal component analysis (PCA) can be used to reduce dimensionality, but since PC is related to power an alternate approach can be used when analysing interaction. We propose an approach to find for each pair of voxels the respective dipole orientations which maximize the IPC between those voxels. The problem can be solved analytically in closed form. In this way we find for a given seed the largest possible interaction with any other voxel as measured by IPC. The method is tested on subjects performing a force controlled motor task (squeeze and release a water filled pump) with the left hand. As a seed region the contralateral primary motor area M1 is chosen and the IPC is estimated for various frequencies. Frequency specific preferential directions are found that correspond to patterns showing e.g. a somato-motor circuit involved in low alpha and a visuo-spatial attention network in high alpha.