December 2022
Volume 22, Issue 14
Open Access
Vision Sciences Society Annual Meeting Abstract  |   December 2022
Precise and generalizable cartography of functional topographies in individual brains
Author Affiliations & Notes
  • Ma Feilong
    Dartmouth College
  • Samuel A. Nastase
    Princeton University
  • Guo Jiahui
    Dartmouth College
  • Yaroslav O. Halchenko
    Dartmouth College
  • M. Ida Gobbini
    Dartmouth College
    Università di Bologna
  • James V. Haxby
    Dartmouth College
  • Footnotes
    Acknowledgements  This work was supported by NSF grants 1835200 (M.I.G) and 1607845 (J.V.H).
Journal of Vision December 2022, Vol.22, 3813. doi:https://doi.org/10.1167/jov.22.14.3813
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ma Feilong, Samuel A. Nastase, Guo Jiahui, Yaroslav O. Halchenko, M. Ida Gobbini, James V. Haxby; Precise and generalizable cartography of functional topographies in individual brains. Journal of Vision 2022;22(14):3813. https://doi.org/10.1167/jov.22.14.3813.

      Download citation file:


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

      ×
  • Supplements
Abstract

Each brain has unique functional topographies. The same functional region differs in size, shape, and topology across individual brains. The multivariate spatial patterns that encode neural representations also have distinct topographies across individuals. In this work, we present an individualized model of brain function that has a fine spatial resolution, which precisely captures these topographic idiosyncrasies of each brain and accurately predicts the brain's response patterns to new stimuli. This model, which we call "warp hyperalignment", first creates a functional brain template based on a group of participants, and the features (e.g., voxels) of this functional template comprise a high-dimensional feature space. The functional profiles of each individual brain is modeled as a linear transformation ("warping") of this template feature space. We applied warp hyperalignment to two fMRI datasets that comprised movie-watching, object category localizers, and retinotopic scans. First, we found that: (a) the modeled brain functional profiles based on independent movie data of the same individual were highly similar, and much more similar than those based on different individuals. Second, the model trained on movie data can accurately predict brain response patterns to object categories and retinotopic maps of each individual. The quality of these model-predicted maps sometimes exceeds the quality of maps based on localizer scans with typical durations. Third, the model accurately predicts fine-grained spatial patterns. The model trained on half of the movie data can accurately predict brain responses to the other half. Based on the similarity of measured and predicted response patterns to the movie, we were able to predict which individual time point (TR) the subject was watching with approximately 50% accuracy (chance accuracy < 0.1%). In summary, we present an individualized model of brain function that is precise, specific to the individual, and has a fine spatial resolution.

×
×

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.

×