A cascade model of information processing and encoding for retinal prosthesis

Abstract

Retinal prosthesis offers a potential treatment for individuals suffering from photoreceptor degeneration diseases. Establishing biological retinal models and simulating how the biological retina convert incoming light signal into spike trains that can be properly decoded by the brain is a key issue. Some retinal models have been presented, ranking from structural models inspired by the layered architecture to functional models originated from a set of specific physiological phenomena. However, Most of these focus on stimulus image compression, edge detection and reconstruction, but do not generate spike trains corresponding to visual image. In this study, based on state-of-the-art retinal physiological mechanism, including effective visual information extraction, static nonlinear rectification of biological systems and neurons Poisson coding, a cascade model of the retina including the out plexiform layer for information processing and the inner plexiform layer for information encoding was brought forward, which integrates both anatomic connections and functional computations of retina. Using MATLAB software, spike trains corresponding to stimulus image were numerically computed by four steps: linear spatiotemporal filtering, static nonlinear rectification, radial sampling and then Poisson spike generation. The simulated results suggested that such a cascade model could recreate visual information processing and encoding functionalities of the retina, which is helpful in developing artificial retina for the retinally blind.

Keywords: NSFC grants; Poisson spike generation; contrast gain control; firing rate; linear spatiotemporal filter; nerve regeneration; neural regeneration; photoreceptor degeneration; retinal prosthesis; spike trains; static non-linear rectification; synaptic transmission.

Figure 1

Schematic diagram of the retinal cascade model. Layers of the cells are represented by boxes. Two distinct synaptic layers, the inner plexiform layer and outer plexiform layer, define two successive information processing and encoding stages.

Figure 2

Spatiotemporal filters used for information processing in the out plexiform layer (OPL). (A) Spatial, low-pass Gaussian kernel S_σ(x, y), (x, y) are the Cartesian coordinates of pixels in the stimulus image. (B) Temporal, low-pass exponential cascade kernel T_α,τ(t), t is the time variable, and α is the total number of exponential kernel.

Figure 3

Spatiotemporal filtering images in the outer plexiform layer (OPL) with different relative surround weights. (A) Original visual stimulus image. (B) Response to the stimulus image with ω = 0.8. (C) Response to the stimulus image with ω = 0.85. (D) Response to the stimulus image with ω = 0.9. (E) Response to the stimulus image with ω = 0.95. (F) Response to the stimulus image with ω = 1. When relative weight was set as 0.95, this choice results in a predominance of edges and contrast in uniform zones. The OPL reacts to a mixture of spatiotemporal band pass properties and extracts feature information of visual stimulus image.

Figure 4

Static nonlinear rectification function in the inner plexiform layer. The S-type simplified synaptic transmission function means that the gain of biological visual systems is high at low light levels and low at high light levels. This processing step ensures that the spike rate of the ganglion cells is maintained in an appropriate range. V_Bip means normalized gray value at pixel.

Figure 5

Radial sampling pattern of the Static nonlinear rectification image in the inner plexiform layer. (A) Static nonlinear rectification output in the inner plexiform layer. (B) The radial sampling pattern of all spike rates for 1 circular level and 25 samples per circle.

Figure 6

Spike trains generated by one of 25 sampled retinal ganglion cells. Each dark vertical line indicates a retinal ganglion cell fires spike at certain time. Visual information is encoded in the form of interspike times.