Edge Detection via Lateral Inhibition Anastasio Section 3.1 Learning Objectives: After mastering the material in this lesson, you should be able to: explain how lateral inhibition generates center-surround receptive fields write code to model processing of visual stimuli by a neuronal network with lateral inhibition

Receptive fields of retinal ganglion cells --recall: retinal ganglion cells do NOT function like pixels in a digital camera

Hierarchical processing of visual information --retinal ganglion cells respond to “center-surround” patterns of light --simple cells in primary visual cortex detect edges --neurons in much higher processing centers such as the inferior temporal cortex respond to entire faces Principles of Neural Science, Kandel, pp. 588, 604, 626

Lateral inhibition Tutorial on Neural Systems Modeling, Anastasio, p. 67

Horseshoe crab (Limulus) eye

Edge detection via lateral inhibition What happens to firing rate of a neuron that is stimulated with light, and whose neighbors are also stimulated with light? Tutorial on Neural Systems Modeling, Anastasio, p. 68

Simulating later inhibition with matrix multiplication Output activity Input (light intensity) Connectivity matrix

An example x (input) y (output)

Lateral inhibition crudely calculates the second spatial derivative (inv) Tutorial on Neural Systems Modeling, Anastasio, p. 75

What determines the receptive fields of retinal ganglion cells? Firing rate --note how flooding a receptive field either entirely with light or entirely with darkness has very little effect upon the firing rate of the corresponding retinal ganglion neuron --if you shine a spot of light the size of the center of the receptive field, it makes a big difference where you place that spot of light -- Ganglion cells in the fovea have small receptive fields of 0.5 minutes of arc (viewing a quarter from 500 feet), while ganglion cells which contribute to peripheral vision have receptive fields of one degree or more --two main kinds of retinal ganglion cells: on-center and off-center. --on-center respond when the center of the receptive field is lit up, and the surrounding region is dark --off-center respond to the inverted stimulus --it is interesting to think about how such a receptive field is produced based on the wiring of the retinal ganglion cells to rods, cones, horizontal cells, etc. --it is a result of a method of wiring called later inhibition, which we will explore later in the class Time (s) Principles of Neural Science, Kandel, p. 588

Retinal ganglion cell receptive fields are produced by lateral inhibition --stimulation by light actually causes the membrane potential of a rod to decrease Bipolar cells link receptor cells to retinal ganglion cells Horizontal cells and amacrine cells provide lateral inhibition Tutorial on Neural Systems Modeling, Anastasio, p. 76

This figure shows a computer simulation of the retinal ganglion cell neural image in response to Prof. Heeger's daughter and her bunny. The image on the left is the input stimulus presented to the eye. The image on the right is a simulation of the retinal ganglion cell responses. The intensity at each position is proportional to the response of a retinal ganglion cell with a receptive field centered at that position. Mid-gray means no response. White means a large response by the ON-center ganglion cells. Black means a large response by the OFF-center cells. What's the point of center-surround receptive fields? They emphasize edges. http://www.cns.nyu.edu/~david/courses/perception/lecturenotes/ganglion/ganglion.html