Purchase this article with an account.
Ian Pennock, Chris Racey, Kendrick Kay, Thomas Naselaris, Emily Allen, Yihan Wu, Anna Franklin, Jenny Bosten; Color-selective brain responses and hue representations from ultra-high-field fMRI of natural scenes. Journal of Vision 2021;21(9):2009. doi: https://doi.org/10.1167/jov.21.9.2009.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Color selectivity in early visual cortical regions has been relatively well studied using fMRI, but less is known about responsiveness to chromatic stimuli in cortical regions beyond V4. One study by Lafer-Sousa et al. (J. Neurosci., 2016) found three color-biased regions (posterior, central and anterior) along the ventral visual pathway located between face- and place-selective cortical areas. We have investigated color selectivity in visual cortical areas using the Natural Scenes Dataset (NSD; Allen et al., J. Vision, 2020), a large-scale fMRI dataset of BOLD responses in 8 participants who each viewed 9,000-10,000 color natural scenes (22,500-30,000 trials). In a whole-brain analysis we correlated the average color saturation of the images (relative to background grey) with the responses of each voxel. Our results confirm the existence of posterior, central and anterior color-biased regions in the ventral stream. Furthermore, given the high signal-to-noise ratio provided by NSD, we observe color-selective regions extending further anteriorly than those identified by Lafer-Sousa et al. Importantly, our results reveal that color selectivity diverges into two distinct streams in the ventral visual pathway. We created regions of interest (ROIs) within these two streams and have identified the images within each ROI that elicit the largest responses, allowing us to characterize the properties of color selectivity in the two streams. We also characterized hue-specific representations in visual cortical areas. For 8 hue bins, we conducted a representational similarity analysis, first grouping images based on the relative proportion of pixels in each hue bin and selecting the highest 500 images per hue bin. We then constructed representational dissimilarity matrices from the average activity across images in each hue bin and used multidimensional scaling to reconstruct ‘neural’ color spaces. Our results show distinct and reliable patterns of hue representation for different visual cortical areas.
This PDF is available to Subscribers Only