Estimation of average single-unit receptive field size by fMRI

Receptive fields (RF) are a fundamental property characterizing sensory neurons. In the visual domain, numerous electrophysiological and computational studies established the spatio-temporal characteristics of RFs and modeled early visual neurons in primates as Gabor filters with well-defined properties. Recently functional magnetic resonance imaging (fMRI) techniques have been introduced to estimate aggregate (voxel-based) "population" receptive field (pRF) sizes in humans (Dumoulin SO, Wandell BA, 2008). Population receptive field estimates are a function of: 1) the receptive field properties of single units belonging to a voxel, and 2) the scatter in the location of receptive field centers across units. In this study, we estimate RF sizes by exploiting the spatial-frequency selectivity of visual RFs modeled as Gabor functions. Blood oxygen level dependent (BOLD) measurements were collected from humans fixating in the center of band-limited white noise stimuli presented in a block-design (12 seconds ON, 20 seconds OFF). In different blocks, we modulated the spatial frequency content of the stimuli by changing the size of the pixels of the white noise. We found that the BOLD signal amplitude in retinotopic visual cortex changes dramatically with different stimulation conditions in a way consistent with the spatial frequency sensitivity of the Gabor RF models. We modeled the BOLD signal of each voxel as a sum of Gabors with homogeneous orientation distribution followed by a compressive non-linearity and fit this model to our data. The standard deviation of the Gaussian envelope of the Gabor function provides an estimate of RF size. Estimates obtained this way were compared to pRF estimates derived from additional experiments with moving bar stimuli. RF size estimates obtained with our method increased linearly with eccentricity as expected, but were significantly smaller in comparison to standard pRF measurements (Dumoulin SO, Wandell BA, 2008). The reason is that our method is not sensitive to the receptive field scatter that happens between units belonging to the same voxel. Similar experiments performed in anesthetized macaques provided RF size estimates comparable to electrophysiological measurements of single-unit RFs. We conjecture that our method estimates for the fist time average single-unit RF sizes in the human