August 2023
Volume 23, Issue 9
Open Access
Vision Sciences Society Annual Meeting Abstract  |   August 2023
Probing the role of bypass connections in core object recognition by chemogenetic suppression of macaque V4 neurons
Author Affiliations & Notes
  • Kohitij Kar
    Department of Biology, Centre for Vision Research, York University, Toronto, Canada
  • Footnotes
    Acknowledgements  KK was supported by the Canada Research Chair Program. This research was undertaken thanks in part to funding from the Canada First Research Excellence Fund. We also thank Jim DiCarlo, Carolyn (Wan-hsun) Wu, and Kailyn Schmidt for their support.
Journal of Vision August 2023, Vol.23, 5736. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kohitij Kar; Probing the role of bypass connections in core object recognition by chemogenetic suppression of macaque V4 neurons. Journal of Vision 2023;23(9):5736.

      Download citation file:

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

  • Supplements

The macaque ventral visual pathway is typically modeled as a series of hierarchically organized cortical areas that successively transform the retinal input into visual object-based linearly separable neural representations. The most behaviorally explicit form of this representation has been discovered in the macaque inferior temporal (IT) cortex. However, each stage of the ventral stream (e.g., areas V1, V2) projects to multiple other areas in the brain. In addition, many projections from early visual areas (e.g., V1, V2) bypass area V4 and connect directly to the IT cortex. Therefore, the underlying neural circuitry is far more complex than a feedforward architecture. Here I provide evidence for the functional relevance of such bypass connections during object recognition. I hypothesized that images that require fewer transformations to generate linearly separable object representations most benefit from bypass connections. This would allow fast object detection under time-sensitive decision-making like assessing predatory threats. Using deep convolutional neural networks (DCNNs), I categorized 10,000 images (10 objects, 1000 images/object) into two categories (with equal categorization performance at the final layer). The “slow-evolved” images required more transformations to reach their final object classification accuracy than the “fast-evolved” images. To test this hypothesis in the primate ventral stream, I expressed inhibitory DREADDs within a 5x5 mm subregion of the V4 cortex via multiple viral injections (AAV8-hSyn-hM4Di-mCherry; two macaques). I recorded from multi-electrode arrays implanted over the transfected V4 and downstream IT cortex while monkeys’ performed object discrimination tasks. Successful V4 neural suppression (~20%) ensured I could produce a partial lesion within the ventral stream hierarchy. Interestingly, I observed that monkeys’ accuracies were significantly higher for “fast-evolved” compared to “slow-evolved” images (only when the objects overlapped with the transfected V4 receptive field). These results suggest that bypass connections allow "fast-evolved" images to retain high object recognition accuracy despite a V4 lesion.


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.