Bayesian population receptive field modelling.

We introduce a probabilistic (Bayesian) framework and associated software toolbox for mapping population receptive fields (pRFs) based on fMRI data. This generic approach is intended to work with stimuli of any dimension and is demonstrated and validated in the context of 2D retinotopic mapping. The framework enables the experimenter to specify generative (encoding) models of fMRI timeseries, in which experimental stimuli enter a pRF model of neural activity, which in turns drives a nonlinear model of neurovascular coupling and Blood Oxygenation Level Dependent (BOLD) response. The neuronal and haemodynamic parameters are estimated together on a voxel-by-voxel or region-of-interest basis using a Bayesian estimation algorithm (variational Laplace). This offers several novel contributions to receptive field modelling. The variance/covariance of parameters are estimated, enabling receptive fields to be plotted while properly representing uncertainty about pRF size and location. Variability in the haemodynamic response across the brain is accounted for. Furthermore, the framework introduces formal hypothesis testing to pRF analysis, enabling competing models to be evaluated based on their log model evidence (approximated by the variational free energy), which represents the optimal tradeoff between accuracy and complexity. Using simulations and empirical data, we found that parameters typically used to represent pRF size and neuronal scaling are strongly correlated, which is taken into account by the Bayesian methods we describe when making inferences. We used the framework to compare the evidence for six variants of pRF model using 7 T functional MRI data and we found a circular Difference of Gaussians (DoG) model to be the best explanation for our data overall. We hope this framework will prove useful for mapping stimulus spaces with any number of dimensions onto the anatomy of the brain.

A Retinotopic Basis for the Division of High-Level Scene Processing between Lateral and Ventral Human Occipitotemporal Cortex.

In humans, there is a repeated category-selective organization across the lateral and ventral surfaces of the occipitotemporal cortex. This apparent redundancy is often explained as a feedforward hierarchy, with processing within lateral areas preceding the processing within ventral areas. Here, we tested the alternative hypothesis that this structure better reflects distinct high-level representations of the upper (ventral surface) and lower (lateral surface) contralateral quadrants of the visual field, consistent with anatomical projections from early visual areas to these surfaces in monkey. Using complex natural scenes, we provide converging evidence from three independent functional imaging and behavioral studies. First, population receptive field mapping revealed strong biases for the contralateral upper and lower quadrant within the ventral and lateral scene-selective regions, respectively. Second, these same biases were observed in the position information available both in the magnitude and multivoxel response across these areas. Third, behavioral judgments of a scene property strongly represented within the ventral scene-selective area (open/closed), but not another equally salient property (manmade/natural), were more accurate in the upper than the lower field. Such differential representation of visual space poses a substantial challenge to the idea of a strictly hierarchical organization between lateral and ventral scene-selective regions. Moreover, such retinotopic biases seem to extend beyond these regions throughout both surfaces. Thus, the large-scale organization of high-level extrastriate cortex likely reflects the need for both specialized representations of particular categories and constraints from the structure of early vision. SIGNIFICANCE STATEMENT: One of the most striking findings in fMRI has been the presence of matched category-selective regions on the lateral and ventral surfaces of human occipitotemporal cortex. Here, we focus on scene-selective regions and provide converging evidence for a retinotopic explanation of this organization. Specifically, we demonstrate that scene-selective regions exhibit strong biases for different portions of the visual field, with the lateral region representing the contralateral lower visual field and the ventral region the contralateral upper visual field. These biases are consistent with the retinotopy found in the early visual areas that lie directly antecedent to category-selective areas on both surfaces. Furthermore, these biases extend beyond scene-selective cortex and provide a retinotopic basis for the large-scale organization of occipitotemporal cortex.