Abstract
Background: We used functional magnetic resonance imaging adaptation (fMRIa) and radial frequency (RF) stimuli to investigate shape coding in the lateral occipital complex (LOC) an important cortical area that mediates integration of local contour information. An RF pattern consists of a closed contour whose radius sinusoidally deviates from circularity as determined by frequency (i.e., the periodicity of bumps around the circle) and amplitude (i.e., bump size). As a Fourier basis set, the family of RF patterns defines a large space of unique closed-contour shapes. Methods: BOLD responses in LOC were collected for 7 participants on a 3T GE scanner according to standard fMRI techniques. LOC was localized by contrasting BOLD responses to scrambled and non-scrambled objects. Results: BOLD responses were lowest for blocks of purely circular stimuli than for all other blocks consisting of single RF contours with fixed shape and amplitude. Blocks in which contours varied in amplitude and/or shape exhibited varying degrees of adaptation release with respect to fixed-shape-and-amplitude blocks. Conclusions: Results are consistent with a population code in which closed-contour shapes are represented as deviations from a circularity prototype. In particular, we interpret weaker BOLD responses to pure circles as an instance of efficient predictive coding that reduces redundancy between neural outputs. We are further characterizing the topology of BOLD response in RF space and the corresponding shape selectivity of population subunits in LOC.
CIHR Training Grant in Vision Health Research.