Purchase this article with an account.
Karl Zipser, Kendrick Kay; P-imaging: a technique for comparing visually evoked population responses across visual areas and subjects. Journal of Vision 2015;15(12):576. doi: https://doi.org/10.1167/15.12.576.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
A distinct advantage of fMRI is the ability to study activation in multiple visual areas simultaneously. However, the sheer number of units (voxels) from which measurements are obtained raises the question of how to devise analyses that preserve the native resolution of the data (i.e. single voxels) while still maintaining tractability. Furthermore, comparing response properties across visual areas is challenging due to the convoluted anatomical structure of cortex. These problems are exacerbated when attempting to compare and contrast brain responses from different individuals. To address these problems, we propose a technique called 'P-imaging' in which functional activity measurements are projected onto stimulus space. This projection is carried out based on an estimated receptive field for each unit. P-imaging allows visualization of activation across all units present in a given visual area, comparison of activation across visual areas within a subject, as well as comparison of results from different subjects. These capabilities facilitate interpretation of response activations, especially for complex visual stimuli. Moreover, P-imaging can be used to quickly and effectively compare measurements of activity for a set of voxels against predictions of activity for the same voxels based on a computational model. We apply P-imaging to a previously acquired fMRI dataset in which a large set of natural scenes were presented to three subjects. We find striking similarity of the population responses of V1 and V2 to individual scenes within and across subjects. We also show that a simple model of texture-energy integration accounts for much of the data in these early visual areas. However, we discover some discrepancies: human faces tend to evoke stronger responses than predicted by the model, and extended contours and extended periodic patterns evoke weaker responses than predicted by the model. These observations support revisions to standard models of early visual cortex.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only