Search Papers | Poster Sessions | All Posters
Poster B157 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
An open dataset of functional MRI responses to egocentric navigation through natural scenes
Emily M. Chen1 (), Frederik S. Kamps1, Rebecca R. Saxe1; 1Massachusetts Institute of Technology
Successful visually-guided navigation is crucial for everyday life, yet how the human visual cortex responds during dynamic, egocentric navigation is understudied. Here, we created a dataset of functional MRI responses to the egocentric visual experience of visually-guided navigation. This dataset includes voxelwise fMRI responses estimated for 172 unique dynamic scene videos and 18 control videos of faces, objects, or abstract patterns. Scene videos depicted unfamiliar places with no visible people and varied in ego-motion direction, affordances, and scene content. Participants also completed a separate set of functional localizer scans. Univariate analyses of subject-specific scene-selective regions revealed clear scene selectivity at the individual stimulus level, as well as considerable reliable variation in the strength of response across different scene videos. The top-performing videos, which could be used to optimize the efficiency of localizer tasks, tended to depict dynamic ego-motion through doorways and tight spaces with numerous viewpoints, whereas bottom-performing videos tended to depict limited ego-motion through open spaces with little change in viewpoint. Next, multivariate analyses showed evidence of region-specific scene representations, consistent with previous work finding functional distinctions between scene regions. These features set the stage for future experiments using this dataset to test representations of naturalistic variation in navigational information, as well as comparing computational models to human brains.
Keywords: scene perception fMRI occipital place area natural images