Using multifocal VEPs to extract retinotopic sources of activity

Abstract

We used multifocal VEPs to investigate retinotopically organized visual areas by localizing the sources of activation elicited from a high density multifocal stimulus. The VEP stimuli consisted of a dartboard pattern containing 192 checkerboard patches. The patches were arranged in 8 cortically scaled rings of 24 patches each. The stimulus extended from .5 to 8.5 degrees of eccentricity. Each patch was modulated according to a binary m-sequence and was scaled to activate about 15 mm² of primary cortex. Each subject viewed the stimulus wearing a cap with 96 recording electrodes. The evoked scalp topographies were consistent with the known general configuration of early visual areas. We then decomposed the data using singular value decomposition. The first two temporal components accounted for 80% of the variance. However, the components extracted by singular value decomposition do not necessarily correspond to physiological sources of activity. In order to find plausible sources we explored the space spanned by these components by taking linear combinations. We then used various source localization algorithms on each of these combinations, including cortically constrained methods. We also devised a source localization procedure that constrained sources to follow the known retinotopic layout of our stimulus. We compared the sources from each algorithm with standard fMRI retinotopic mapping data from the same subjects. When the proper components were chosen we found an agreement between the EEG and the fMRI. This demonstrates that EEG in combination with fMRI can be used to find fast temporal information from early visual areas.

Footnotes

 NIH EY015825, NIH T32 EY07