Reconstructing human population receptive field properties

Vision is the dominant sense in humans; about 25% of the human cerebral cortex with roughly 5 billion neurons process the visual information. Especially in early visual cortex, these neurons process information from a small part of the visual field known as their receptive field. Recent advances in functional magnetic resonance imaging (fMRI) data-analysis techniques have revealed these receptive field properties. These fMRI data-analysis techniques are non-invasive and readily applied in humans, both in healthy and clinical subject groups. But these fMRI receptive field properties are not the same as single neuron receptive field properties. Given typical neural packing densities and standard fMRI resolutions (!2.5 mm isotropic), about a million neurons contribute—though not necessarily to the same extent—to each fMRI recording site. Therefore, the region of visual space that stimulates the recording site is referred to as the population receptive field (pRF). There are several methods to reconstruct the pRF properties. Here, we focus on approaches that fit an explicit neural model to the fMRI data. The advantages of neural model-based approaches are that they (a) explicitly model the underlying neural properties, (b) are flexible in both the nature of the model and are independent of the stimulus layout, and (c) can be used to explain and generate predictions for any stimulus condition. We will focus on the pRF properties.