September 2015
Volume 15, Issue 12
Vision Sciences Society Annual Meeting Abstract  |   September 2015
Assessing the upper bound on performance of multi-voxel pattern analysis in peripheral V1
Author Affiliations
  • Rachel Millin
    Neuroscience Graduate Program, University of Southern California
  • Bosco Tjan
    Neuroscience Graduate Program, University of Southern California Department of Psychology, University of Southern California
Journal of Vision September 2015, Vol.15, 995. doi:10.1167/15.12.995
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Rachel Millin, Bosco Tjan; Assessing the upper bound on performance of multi-voxel pattern analysis in peripheral V1. Journal of Vision 2015;15(12):995. doi: 10.1167/15.12.995.

      Download citation file:

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

  • Supplements

Multi-voxel pattern analysis (MVPA) of fMRI data is a powerful technique for assessing the amount of information present in a cortical area for a given task. However, the utility of this method is limited by the areal size of the cortical region of interest, the spatial scale of the neural response, and the resolution of BOLD fMRI. These limitations are particularly relevant to the study of early peripheral visual processing, due to the small area of primary visual cortex (V1) devoted to processing information from the peripheral visual field. We used a forward model of BOLD fMRI to determine the bounds on the stimulus information available to MVPA for peripheral stimuli on the V1 cortical surface. The model captures the main spatial and temporal properties of BOLD fMRI in V1, regardless of the specific mechanisms of the neural response. It includes the V1 cortical magnification factor, the spatial and temporal correlation in BOLD signal and noise, the signal to noise ratio across voxels, and a biophysical model of the BOLD signal. The model also incorporates other factors that influence BOLD data, such as head and eye movements. After determining model parameters from an independent dataset, we simulated multiple BOLD data acquisitions from V1 when letter stimuli were presented in the visual periphery. A cross-validation scheme was used to assess the predictability of the stimuli as a function of letter size. The spatial and temporal factors of BOLD fMRI captured by the model place a lower limit on the stimulus size for which information is available through MVPA. This limitation was validated with empirical BOLD data. The model can thus be used to test the feasibility of using MVPA in a given experiment, and to optimize experiment design to maximize the probability of an informative outcome.

Meeting abstract presented at VSS 2015


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.