Spike-triggered average electrical stimuli as input filters for bionic vision—a perspective

1. Spatial summation within cat retinal receptive fields was studied by recording from optic-tract fibres the responses of ganglion cells to grating patterns whose luminance perpendicular to the bars varied sinusoidally about the mean level. 2. Summation over the receptive fields of some cells (X-cells) was found to be approximately linear, while for other cells (Y-cells) summation was very non-linear. 3. The mean discharge frequency of Y-cells (unlike that of X-cells) was greatly increased when grating patterns drifted across their receptive fields. 4. In twenty-one X-cells the relation between the contrast and spatial frequency of drifting sinusoidal gratings which evoked the same small response was measured. In every case it was found that the reciprocal of this relation, the contrast sensitivity function, could be satisfactorily described by the difference of two Gaussian functions. 5. This finding supports the hypothesis that the sensitivities of the antagonistic centre and surround summating regions of ganglion cell receptive fields fall off as Gaussian functions of the distance from the field centre. 6. The way in which the sensitivity of an X-cell for a contrast-edge pattern varied with the distance of the edge from the receptive field centre was determined and found to be consistent with the cell's measured contrast sensitivity function. 7. Reducing the retinal illumination produced changes in the contrast sensitivity function of an X-cell which suggested that the diameters of the summating regions of the receptive field increased while the surround region became relatively ineffective.

A light-sensitive, externally powered microchip was surgically implanted subretinally near the macular region of volunteers blind from hereditary retinal dystrophy. The implant contains an array of 1500 active microphotodiodes (‘chip’), each with its own amplifier and local stimulation electrode. At the implant's tip, another array of 16 wire-connected electrodes allows light-independent direct stimulation and testing of the neuron–electrode interface. Visual scenes are projected naturally through the eye's lens onto the chip under the transparent retina. The chip generates a corresponding pattern of 38 × 40 pixels, each releasing light-intensity-dependent electric stimulation pulses. Subsequently, three previously blind persons could locate bright objects on a dark table, two of whom could discern grating patterns. One of these patients was able to correctly describe and name objects like a fork or knife on a table, geometric patterns, different kinds of fruit and discern shades of grey with only 15 per cent contrast. Without a training period, the regained visual functions enabled him to localize and approach persons in a room freely and to read large letters as complete words after several years of blindness. These results demonstrate for the first time that subretinal micro-electrode arrays with 1500 photodiodes can create detailed meaningful visual perception in previously blind individuals.

Response properties of sensory neurons are commonly described using receptive fields. This description may be formalized in a model that operates with a small set of linear filters whose outputs are nonlinearly combined to determine the instantaneous firing rate. Spike-triggered average and covariance analyses can be used to estimate the filters and nonlinear combination rule from extracellular experimental data. We describe this methodology, demonstrating it with simulated model neuron examples that emphasize practical issues that arise in experimental situations.