October 2020
Volume 20, Issue 11
Open Access
Vision Sciences Society Annual Meeting Abstract  |   October 2020
Voxelwise modeling reveals selectivity for body part identity and location in BOLD fMRI responses to complex naturalistic stimuli
Author Affiliations & Notes
  • Matthew Shinkle
    University of Nevada, Reno
  • Mark Lescroart
    University of Nevada, Reno
  • Footnotes
    Acknowledgements  Pilot grant to ML through NIH COBRE Award P20GM103650
Journal of Vision October 2020, Vol.20, 1675. doi:https://doi.org/10.1167/jov.20.11.1675
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Matthew Shinkle, Mark Lescroart; Voxelwise modeling reveals selectivity for body part identity and location in BOLD fMRI responses to complex naturalistic stimuli. Journal of Vision 2020;20(11):1675. https://doi.org/10.1167/jov.20.11.1675.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Perception of bodies is central to social interaction, and large portions of human lateral occipitotemporal cortex (LOTC) purportedly represent body parts, poses, and actions. Prior research has found tuning for body parts and their locations throughout LOTC, but the spatial pattern of this tuning has varied. We used voxel-wise modeling to investigate the extent of tuning to both the position and identity of different body parts in BOLD fMRI data. We adapted OpenPose (a deep neural network for pose estimation) to parameterize locations of body parts in three stimulus sets used in previous fMRI studies: static natural images, natural movie clips, and computer-rendered scenes. These large and diverse stimulus sets enabled us to determine how brain responses are specifically related to body part identity and location among other sources of visual variation. We used regularized linear regression to estimate weights relating each voxel to each quantified feature. To validate our model, we correlated its predictions with observed responses in withheld data. Large contrasts between feature weights for well-predicted voxels indicate differential tuning for these features. This model yielded accurate predictions of BOLD responses throughout body- and face-selective areas including LO, FFA, OFA, EBA, and pSTS. Across subjects and stimulus sets, we reliably found a cluster of voxels displaying spatial bias toward the lower, central, contralateral visual field in bilateral caudal EBA. Though natural arrangements of body parts in stimuli can impair discrimination between tuning for different parts, we observed a bias toward hands and arms over faces in the same portion of EBA across all subjects, with inconsistent findings of a few other similarly-tuned regions. With the prior caveat, these results argue against broad selectivity for body parts across all of LOTC, and strengthen previous work suggesting a circumscribed model of body tuning in a few sub-regions near EBA.


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.