Feasibility of functional magnetic resonance imaging of ocular dominance and orientation preference in primary visual cortex

A recent hemodynamic model is extended and applied to simulate and explore the feasibility of detecting ocular dominance (OD) and orientation preference (OP) columns in primary visual cortex by means of functional magnetic resonance imaging (fMRI). The stimulation entails a short oriented bar stimulus being presented to one eye and mapped to cortical neurons with corresponding OD and OP selectivity. Activated neurons project via patchy connectivity to excite other neurons with similar OP in nearby visual fields located preferentially along the direction of stimulus orientation. The resulting blood oxygen level dependent (BOLD) response is estimated numerically via the model’s spatiotemporal hemodynamic response function. The results are then used to explore the feasibility of detecting spatial OD-OP modulation, either directly measuring BOLD or by using Wiener deconvolution to filter the image and estimate the underlying neural activity. The effect of noise is also considered and it is estimated that direct detection can be robust for fMRI resolution of around 0.5 mm, whereas detection with Wiener deconvolution is possible at a broader range from 0.125 mm to 1 mm resolution. The detection of OD-OP features is strongly dependent on hemodynamic parameters, such as low velocity and high damping reduce response spreads and result in less blurring. The short-bar stimulus that gives the most detectable response is found to occur when neural projections are at 45 relative to the edge of local OD boundaries, which provides a constraint on the OD-OP architecture even when it is not fully resolved.Author summary: Ocular dominance (OD) and orientation preference (OP) cells of the visual cortex are numerically simulated to investigate the feasibility of their in vivo determination via functional magnetic resonance imaging (fMRI) in humans. A short oriented bar with OD-OP features is mapped to the primary visual cortex and the blood oxygen level dependent (BOLD) response is numerically estimated via a spatiotemporal hemodynamic response function model. The model predicts the BOLD response from a given neural activity and vice versa. Our results show that direct OD-OP detection from BOLD is feasible for a resolution of 0.5 mm, while detection via Wiener deconvolution can be attained to resolutions from Δx ≈ 0.125 − 1.0 mm.