Siddharth Mittal¹, Michael Woletz¹, David Linhardt¹, Christian Windischberger¹; ¹Medical University of Vienna

Population receptive field (pRF) modelling is a widely used technique for fMRI-based visual field mapping. While pRF mapping has been shown to be a reproducible method, inter-run variability is still present and not consistent across the visual field. This study aims to generate pRF estimations from simulated data and validate the variability observed by demonstrating its similarity to empirical data. Empirical retinotopy data were collected using a Siemens 3T Prisma Fit scanner from two subjects across five sessions, with six runs per session, resulting in 30 pRF estimations per subject using a 2D Gaussian model. The empirical pRF estimations exhibited systematic distributions dependent on visual field location (μ_x, μ_y, and σ). While these distributions showed a circular Gaussian-like form in foveal regions, distributions in more peripheral locations were ellipsoidal with increased width in eccentricity direction. To replicate this behaviour, we simulated visual receptive fields uniformly distributed across 51x51 spatial locations (x, y) with varying sizes (σ). Using the same stimulus as employed in the empirical data, we generated 5000 simulated BOLD responses per position, incorporating white noise to achieve a contrast-to-noise ratio (CNR) of 1 to 4. This large-scale dataset was processed using GPU-empowered pRF mapping software (GEM-pRF; Mittal et al., 2024) to estimate the probability distribution of pRF parameters for each simulated field. Maximum likelihood estimation was employed to find the best matching simulated distribution to each empirical voxel's estimated pRF results. The significance of these matches was validated using the multivariate Kolmogorov-Smirnov (Naaman, 2021) test, confirming that simulated variabilities reliably reflect empirical variability. This work demonstrates that simulations can accurately replicate the run-to-run variability of pRF parameter estimations. These findings represent a significant step forward in enabling the development of optimised visual stimuli to minimise variability.

Acknowledgements: This research was funded in whole or in part by the Austrian Science Fund (FWF) [https://www.doi.org/10.55776/P35583]