Poster No:
2560
Submission Type:
Abstract Submission
Authors:
Edan Daniel-Hertz1, Jewelia Yao1, Sidney Gregorek1, Patricia Hoyos1, Anna Lyn Williams1, Jesse Gomez1
Institutions:
1Princeton Neuroscience Institute, Princeton, NJ
First Author:
Co-Author(s):
Introduction:
The current theory explaining the organization of category selective regions in the ventral visual stream posits that regions located laterally (i.e. fusiform face areas) or medially (i.e. place area) emerge there as a result of connections to foveal or peripheral retinotopic representations from earlier visual cortex (Hasson et al., 2002). This is also consistent with how faces and scenes are viewed, with foveal and peripheral vision respectively. If indeed the eccentricity selectivity follows a linear gradient from medial (peripheral) to lateral (foveal), a limb-selective region located laterally to the face region should have receptive fields (RFs) sampling foveal visual space. This, however, is inconsistent with how limbs are likely viewed statistically, with peripheral vision. Does this suggest that the single linear foveal-peripheral representation historically ascribed to the ventral stream is an inadequate description of its functional topography?
Methods:
To answer this question, 27 participants (17 females, 18.6±0.75y) underwent functional MRI while completing a visual category localizer (Stigliani et al., 2015) and a population RF mapping experiment (Finzi et al., 2021). Data was preprocessed based on the HCP minimal preprocessing pipeline, and analyses were performed in each individual's native brain space. Category-selective regions of interest (ROIs) were defined for each participant using data from the visual category localizer experiment. A separate group of 25 individuals (13 females, 23±8.6y) completed an eye-tracking experiment during naturalistic movie-watching, to characterize how bodies and limbs are sampled in retinal space.
Results:
In the eye-tracking experiment we found that faces were more than three times more likely to be the focus of fixation compared to bodies (T[26]=19.6, P<.001), or to background/objects (T[26]=23.2, P<.001), indicating that limbs are most commonly processed using peripheral vision. Comparing pRF properties across visual categories, we found that pRFs in the previously uncharacterized occipitotemporal sulcus (OTS) limb-selective region were significantly more eccentric than in the neighboring Fusiform face-selective region (Limb: 5.8°, Face: 3.6°, T[13]=3.36, P=.005). We found no significant difference in the size of the pRFs across those visual regions (Limb: 7.01°, Face: 6.49°, T[13]=1.01, P=.33). These findings are surprisingly inconsistent with the linear gradient of eccentricity, and suggest that our visual experience drives the biases in spatial computations of high-level vision.
Seeking to produce a parsimonious model of visual cortex organization, we quantified the graded changes in pRF eccentricity beyond the ventral surface by parcellating visual cortex into seven anatomically-defined ROIs according to cortical folds, from the medial collateral sulcus (CoS) to the lateral mid-temporal gyrus (MTG). We extracted average pRF properties from each bin in our dataset and the Human Connectome Project's 7T retinotopy dataset (Benson et al., 2018). In both datasets, we observe a non-linear gradient of pRF eccentricity, with pRFs being most foveal near the lateral fusiform gyrus and becoming increasingly peripheral as one travels laterally or medially.
Conclusions:
We demonstrate that the spatial computations performed by the previously uncharacterized limb-selective region are inconsistent with predictions from a single linear gradient of eccentricity model. Limb-selective pRFs sample peripheral visual space, consistent with how they are experienced visually. We make the novel observation that there may in fact be two eccentricity gradients, forming a diverging non-linear topography across ventral visual cortex. This primary description supports a novel organization principle in which the ventral and lateral streams of visual processing may not be as separate as once described, but instead linked by a shared, parabolic-like gradient of RF properties.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 2
Perception, Attention and Motor Behavior:
Perception: Visual 1
Perception and Attention Other
Keywords:
Computational Neuroscience
FUNCTIONAL MRI
Perception
Vision
1|2Indicates the priority used for review
Provide references using author date format
Benson, N., Jamison, K., Vu, A., Winawer, J. & Kay, K. The HCP 7T Retinotopy Dataset: A new resource for investigating the organization of human visual cortex. J. Vis. 18, 215–215 (2018).
Finzi, D. et al. Differential spatial computations in ventral and lateral face-selective regions are scaffolded by structural connections. Nat. Commun. 12, 2278 (2021).
Hasson, U., Levy, I., Behrmann, M., Hendler, T. & Malach, R. Eccentricity bias as an organizing principle for human high-order object areas. Neuron 34, 479–490 (2002).
Stigliani, A., Weiner, K. S. & Grill-Spector, K. Temporal Processing Capacity in High-Level Visual Cortex Is Domain Specific. J. Neurosci. 35, 12412–12424 (2015).