Abstract
Neurons within early visual cortex exhibit a well-characterized relationship between stimulus contrast and neural response: firing rates initially increase in a relatively linear fashion, but eventually saturate, evoking little-to-no change between higher contrast levels. While this neural contrast response function (CRF) has been demonstrated to have a monotonic relationship with behavioral performance, it remains controversial whether the same holds for population-based CRF measures obtained with human neuroimaging. In this fMRI study, we utilized an adaptation paradigm to capture the nonlinearity of the population contrast response function (pCRF). In a first experiment, we measured BOLD responses in early visual cortex (V1–V3) using a traditional fast event-related design, while participants viewed stimuli varying in contrast intensity throughout a scan (9 contrast levels, spaced between 3%–96% Michelson contrast). Stimuli were composed of equally-spaced apertures arranged in five concentric ring patterns. Each aperture contained a grating stimulus that scaled in size and spatial frequency with eccentricity, oriented along the radial axis out from fixation. Using conventional deconvolution analyses to estimate the pCRF, we were able to capture nonlinear pCRFs of individual voxels within each visual area. In a second experiment, we asked whether we could obtain similar pCRF estimates in a more flexible and timely manner by employing a model-based encoding approach. We measured BOLD responses while participants viewed the same stimuli, except that they were presented in a rapid manner (0.5s presentation duration). Utilizing this approach, we were able to predict BOLD responses for these rapidly presented contrast stimuli, and reliably reconstruct the pCRF. The results reveal that estimated pCRF parameters were highly comparable to those obtained using conventional methods. Overall, our results demonstrate that fMRI-based pCRF estimates do exhibit canonical non-linearities, and furthermore we propose a method that allows for a more flexible mapping of the pCRF in human visual cortex.
Acknowledgement: National Institutes of Health Grant EY028163