| ||||||||||||||||||||||||||||||
Overview
This project brings together a collaborative team consisting of two neurobiologists (David Senseman in San Antonio and Philip Ulinski in Chicago), a systems engineer (Bijoy Ghosh in St. Louis) and a computer scientist (Kay Robbins in San Antonio) who are working together to understand the mechanisms and structure that underlie the mapping of visual information into the complex internal representation of the turtle visual cortex.
Information about position in visual space is coded in topographic maps within the sensory cortices of mammals. By contrast, the visual cortex of freshwater turtles does not embody topographically-organized representations of the visual world. Single unit recording experiments in turtles reveal that neurons at any position in visual space respond to stimuli presented anywhere in binocular visual space. Interestingly, voltage sensitive dye (VSD) experiments with an in vitro preparation of the eyes and brain of turtles show that visual stimuli evoke waves of depolarizing activity that propagate across visual cortex. These waves can be conveniently represented as movies.
Analysis of the movies with a novel version of principal component analysis, a two-step Karhunen-Loeve (KL) decomposition, shows that the movies can be represented in low dimensional spaces. The first decomposition represents an entire movie as a trajectory in a phase space called A-space. Presentation of different stimuli result in different trajectories in A-space.
Individual trajectories in A-space can be then be represented as points in a second, low-dimensional phase space called B-space. Spots of light presented at different retinal loci result in discrete clusters of points in B-space. Simulations of spots of light along the horizontal meridian of the lateral geniculate complex using a large-scale model of turtle visual cortex suggest that the horizontal meridian of visual space (V) is mapped along an arc in B-space. This motivates the working hypothesis that there is an organized map of V-space to B-space (V -> B). Position in the visual cortex of turtles would, then, be coded in the spatiotemporal dynamics of the cortical waves.
This work was supported by NSF Grant 217884: CRCNS: How is Information Encoded in Turtle Cortex?