Modeling motion detection in striate visual cortex using massive excitatory feedback

Citation

Suarez, Humbert H. (1995) Modeling motion detection in striate visual cortex using massive excitatory feedback. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/78a8-3m56. https://resolver.caltech.edu/CaltechETD:etd-10252006-092222

Abstract

This thesis is a detailed description and analysis of a model of direction-selective simple cells in cat striate visual cortex. There are three main defining features of our modeling effort compared to previous ones. (1) Local excitatory intracortical connections, known to be very numerous, are taken into account. (2) The model is very detailed: compartmental models of neurons are used and spiking is modeled using Hodgkin and Huxley-like active ionic currents. (3) Model responses are analyzed through standard electrophysiological methods and are compared in detail to physiology. Two separate operating modes are described. When the model acts as a proportional amplifier, contrast-response curves are relatively linear. In the hysteretic amplifier mode, contrast-response curves are much steeper initially, including an early portion with expansivity nonlinearity, but saturate abruptly at high contrasts. These features of the second mode are very similar to cortical contrast-response curves, but very different from the thalamus'. The second mode also predicts that hysteresis is latent in cortex, but that because of resetting through inhibition, cortical neurons do not fire in the absence of stimulation. In both modes, the model achieves strong amplification of the input through the excitatory cortical feedback. Amplification results in small changes in conductance for stimuli moving in the null direction, long a puzzling experimental finding; direction selectivity also persists during blockade of all inhibition in a single cell, as observed in recent experiments. Due to the nonlinearity of this amplification, bandpass velocity-response curves of thalamic neurons can be transformed into velocity low-pass cortical curves. Direction selectivity is invariant over a wide range of contrasts and velocities, a prominent feature of direction-selective cells in cortex. The model also makes specific predictions concerning the effects of selective blockade of cortical inhibition on direction selectivity at different velocities. Finally, we address the important issue of testing experimentally the linearity of cortical neurons. The same intracellular linearity test that has been used for cortical neurons is performed on the model. Although the model has substantial nonlinearities, it appears quite linear according to the linearity test. We explain these surprising observations in detail, and conclude that such tests are much more limited in usefulness than apparent at first.

Thesis Files