Abstract
The contrast sensitivity function (CSF) is widely used in visual perception research and to characterize visual disorders. The CSF defines the lowest contrast level that participants can detect as a function of spatial frequency. Here, we translated this psychophysical model to a neural model by presenting a new method that estimates the CSF in the human visual cortex using ultra-high field (7 Tesla) functional magnetic resonance imaging (fMRI). During the fMRI experiment, we presented a full-field stimulus consisting of different gratings that vary systematically in contrast and spatial frequency. We modeled the CSF using an exponential function, whose parameters include maximum contrast sensitivity and its corresponding spatial frequency and width of the left side and right side of the CSF (Chung & Legge, 2015). Next, we predicted the fMRI response by a multiplication of the CSF model with the stimulus sequence and convolved this model time course with the hemodynamic response function (method analogous to the population receptive field method by Dumoulin & Wandell, 2008). We show that the CSF model explains a significant amount of the variance in the fMRI time series. Moreover, the properties of the CSF model differ between locations in the visual cortex. For example, the preferred maximum spatial frequency and maximum contrast sensitivity are highest for voxels in the foveal region and decrease as a function of eccentricity. These results are similar across visual field maps V1, V2, and V3. Thus, the cortical CSFs vary systematically across eccentricity. This method can be applied to clinical conditions where a reduction in the CSF is present, e.g. amblyopia, glaucoma or macular degeneration. In these cases, the presented method can provide novel insights into the neural processes underlying the perceptual abnormalities.