Vismodel

From VISTA LAB WIKI

Jump to: navigation, search

We are implementing software to simulate different stages along the early visual pathways. The simulation combines physically accurate stimulus generation, cone encoding, retinal ganglion cell spiking responses, and SVM/linear classification algorithms. The spectral images (photon spectral power distributions) are transformed into an irradiance image and cone photoreceptor absorptions using ISET. The retinal ganglion cell spikes are predicted by specifying a receptive field and using the basic computing model for retinal spikes. Finally, we apply classifiers (SVM) to the spiking output data to measure the information contained in the spikes.

This page is an overview. There is a separate Vismodel programming page with a description of code, and a variety of other related pages about portions of the analyses.

Developers: Jon Winawer, David Janssen, Guillaume Leseur, Estelle Comment, Brian Wandell

To see other ongoing projects visit Local projects.

Contents

[edit] Installation

The RGC code depends on both ISET and the vismodel directory in the vistaproj repository. In the past, we set out to integrate this code with the DARPA Synapse project, but that is no longer our mission.

To checkout the code with SVN

  • On a Windows machine place the repository link https://white.stanford.edu/srepos/vistaproj/vismodel link in the tortoisesvn slot.
  • On a Linux machine use the command line svn co https://white.stanford.edu/srepos/vistaproj/vismodel

See the checkout instructions for ISET.

Running the vismodel code requires a version later than Matlab 2008a, because we use the new 'class' implementation.

You must add the vistaproj/vismodel directories to your path, as well as ISET.

[edit] Code overview

This is a very general overview. More detailed comments and examples of Vismodel programming are provided elsewhere.

Physically accurate modeling of signals along the visual pathways. Includes ISET modeling along with biological modeling of cones, retinal ganglion cells, and other components. The goal is to build these models as a foundation for integrating cellular phenomena (spikes), mass action (LFP), and BOLD signaling.

[edit] Scene

[edit] Cones

[edit] RGC

Vismodel spike modeling uses a Poisson process, connection, and feedback.

[edit] V1

[edit] V4

[edit] Classifiers

[edit] Applications

[edit] Perception

We are using vismodel to understand performance in psychophysical tasks (e.g., contrast sensitivity to gratings).

The idea is to create models of the optics and cone photon catches using ISET. The Vismodel RGC code converts cone absorptions into retinal ganglion cell spikes. We then use SVM spike classifiers to evaluate the information available in the retinal ganglion cell spike trains. We can compare this with, say, the information available in the cone mosaic.

[edit] BOLD Imaging

We plan to use these simulations to model the relationship between the spiking activity and the BOLD response in V1. The rgc BOLD methods are not yet developed, just planned.

Stick in the Mayhew reference on modeling BOLD from spikes. Look-up table thing.

[edit] TODO

  • Add cone volts field. Probably keep absorptions and cone volts. Or at least have volts and the conversion factor back to absorptions.
  • Insert eye movement routines back into the optical image.
Personal tools