Abstract
Methods: Human subjects fixate the centre of a 60 region cortically-scaled dartboard stimulus. Checkerboard patterns within each region are simultaneously but separately modulated in contrast according to novel custom-designed temporal waveforms (not m-sequences). The visual evoked potentials (VEPs) are recorded with 32 channel arrays of electrodes (Synamps). Kernel sets are derived modelling the mapping from the 60 region spatiotemporal input to the multichannel output. Singular value decompositions are made of the kernel sets, to provide least-squares reduced-dimension fits of the kernel sets as sums of components of space-time outer products, over sets of visual field spatial locations, and over scalp spatial locations. Current sources within the 3D head-space are then inferred by dipole fitting procedures. Results: Kernel sets from 12 health subjects, with some repeat sessions, indicate excellent within-subject replicability, with great diversity of spatiotemporal kernels between subjects, as has been found previously. The precisions of the kernels obtained, with analysis of the number of significant components in the singular value decompositions, indicates that these kernel sets cannot be treated as having homogeneous time courses around spatial annuli of constant eccentricity, as has been elsewhere suggested. The spatiotemporal kernel over scalp-space can for some regions be fitted by a single dipole, but for many regions the kernels cannot be adequately explained by a single dipole, as has been previously attempted. Sectors of the visual field are found for which regions at increasing eccentricity are fitted by a sequence of single dipoles moving anteriorly. Triphasic kernel waveforms generally require pairs of dipoles, spatially close but at different orientations. Conclusion: The results advance the use of the multichannel multifocal VEP as a measure of spatiotemporal cortical activity with a temporal resolution well beyond that of brain imaging techniques.