Computational Visual Neuroscience Laboratory at CMRR
Home          People          Research          Publications          Resources          Contact
 

Modeling stimulus transformations in extrastriate cortex

What are the stimulus computations that are performed by extrastriate visual areas? To answer this question, we seek to develop quantitative computational models that predict physiological responses in human visual cortex to a wide range of stimuli. Our approach tightly integrates experimental and modeling techniques and exploits the wide coverage of visual areas provided by fMRI. The models we develop can be used to help understand the different types of information encoded in different physiological measures (e.g. single-unit responses, BOLD, EEG/MEG, ECoG).

Selected publications

  • Wandell, B.A., Winawer, J., & Kay, K.N. Computational modeling of responses in human visual cortex. In: Brain Mapping: An Encyclopedic Reference, edited by P. Thompson & K. Friston (2015). Journal link | PDF
  • Winawer, J., Kay, K.N., Foster, B.L., Rauschecker, A.M., Parvizi, J., & Wandell, B.A. Asynchronous Broadband Signals Are the Principal Source of the BOLD Response in Human Visual Cortex. Current Biology 23, 1–9 (2013). Pubmed | PDF
  • Kay, K.N., Winawer, J., Rokem, A., Mezer, A., & Wandell, B.A. A Two-Stage Cascade Model of BOLD Responses in Human Visual Cortex. PLoS Computational Biology 9(5), 1–16 (2013). Pubmed | PDF
  • Kay, K.N., Winawer, J., Mezer, A., & Wandell, B.A. Compressive spatial summation in human visual cortex. Journal of Neurophysiology 110(2), 481–493 (2013). Pubmed | PDF
  • Kay, K.N., Naselaris, T., Prenger, R.J., & Gallant, J.L. Identifying natural images from human brain activity. Nature 452, 352–355 (2008). Pubmed | PDF

Modeling bottom-up and top-down influences on high-level visual cortex

We have recently been exploring the use of a model-based approach to understand responses in high-level visual cortex (e.g. face-selective and word-selective regions of ventral temporal cortex). We seek to carefully characterize bottom-up stimulus effects and dissociate these from top-down cognitive influences. These investigations are aimed towards furthering our understanding of the relationship between early and late visual processing and clarifying the computational role that cognitive processes play in visual processing.

Selected publications

  • Kay, K.N. & Yeatman, J.D. Bottom-up and top-down computations in word- and face-selective cortex. eLife (2017). eLife | PDF
  • Kay, K.N., Weiner, K.S., & Grill-Spector, K. Attention reduces spatial uncertainty in human ventral temporal cortex. Current Biology (2015). Pubmed | PDF

Population receptive field mapping methods

   

Visual space is systematically represented in visual cortex (e.g. retinotopy). One of the major advances of fMRI has been the development of techniques that can be used to map the retinotopic organization of visual cortex. Newer techniques have been proposed to estimate population receptive field size. We are generally interested in developing increasingly accurate and efficient methods for mapping receptive fields (see analyzePRF).

Selected publications

  • Kay, K.N., Winawer, J., Mezer, A., & Wandell, B.A. Compressive spatial summation in human visual cortex. Journal of Neurophysiology 110(2), 481–493 (2013). Pubmed | PDF
  • Kay, K.N., Naselaris, T., Prenger, R.J., & Gallant, J.L. Identifying natural images from human brain activity. Nature 452, 352–355 (2008). Pubmed | PDF
  • Hansen, K.A., Kay, K.N., & Gallant, J.L. Topographic organization in and near human visual area V4. J. Neurosci. 27, 11896–11911 (2007). Pubmed | PDF
 

Encoding and decoding approaches to fMRI

Functional magnetic resonance imaging provides a rich source of information about ongoing physiological activity in the human brain. The data are highly multivariate (thousands of distinct spatial measurements) and therefore require effective and tractable experimental and analysis approaches. Encoding and decoding are useful tools for interpreting the distributed patterns of activity measured from the brain. We are generally interested in developing novel analysis techniques that can help us understand the information-processing operations performed by the brain.

Selected publications

  • [NEW] Kay, K.N. Principles for models of neural information processing. NeuroImage (2017). DOI
  • Naselaris, T. & Kay, K.N. Resolving ambiguities of MVPA using explicit models of representation. Trends in Cognitive Sciences (2015). Pubmed | PDF
  • Kay, K.N. Understanding visual representation by developing receptive-field models. In: Visual Population Codes, edited by N. Kriegeskorte & G. Kreiman (2011). Book Link | PDF
  • Naselaris, T., Kay, K.N., Nishimoto, S., & Gallant, J.L. Encoding and decoding in fMRI. NeuroImage 56, 400–410 (2011). Pubmed | PDF
  • Kay, K.N. & Gallant, J.L. I can see what you see. Nature Neuroscience 12, 245–246 (2009). Pubmed | PDF
  • Naselaris, T., Prenger, R.J., Kay, K.N., Oliver, M., & Gallant, J.L. Bayesian reconstruction of natural images from human brain activity. Neuron 63, 902–915 (2009). Pubmed | PDF
  • Kay, K.N., Naselaris, T., Prenger, R.J., & Gallant, J.L. Identifying natural images from human brain activity. Nature 452, 352–355 (2008). Pubmed | PDF

Denoising and fMRI analysis methods

Functional magnetic resonance imaging data suffer from numerous artifacts that reduce signal-to-noise ratio. Given that the experimental paradigms we use require a large number of conditions and high signal-to-noise ratio, we are interested in methods that can improve data quality. Whereas some optimizations occur at the data acquisition level, other benefits can come in data analysis. We develop analysis techniques that can be used to denoise fMRI data. We are especially interested in techniques that are practical for actual experimental studies, and we make our software code freely available (see GLMdenoise).

Selected publications

  • Henriksson, L., Khaligh-Razavi, S., Kay, K., & Kriegeskorte, N. Visual representations are dominated by intrinsic fluctuations correlated between areas. NeuroImage (2015). Pubmed | PDF
  • Kay, K.N., Rokem, A., Winawer, J., Dougherty, R.F., & Wandell, B.A. GLMdenoise: a fast, automated technique for denoising task-based fMRI data. Frontiers in Neuroscience 7 (2013). Pubmed | PDF
  • Kay, K.N., David, S.V., Prenger, R.J., Hansen, K.A., & Gallant, J.L. Modeling low-frequency fluctuation and hemodynamic response timecourse in event-related fMRI. Hum. Brain Mapp. 29, 142–156 (2008). Pubmed | PDF
 

University of Minnesota | Department of Radiology | Center for Magnetic Resonance Research