Abstract
Abstract: :
Purpose: To develop an algorithm for generation of visual pattern–specific temporal stimulation codes for an array of epiretinal electrodes. To simulate the central visual system function of mapping an array of single ganglion cell codes onto the corresponding visual percept. Methods: A: In a retina encoder (RE) simulation for part of the central retina, a hexagonal array of 64 x 64 photosensors / pattern pixels formed the input to an evenly interlaced distribution of partly superimposed 34 x 34 receptive field (RF) filters of P–On, P–Off, and M ganglion cells each with one RF–center pixel and six RF–periphery pixels. B: Each RF–center pixel belonged exclusively to one ganglion cell and each RF–periphery pixel belonged to exactly two neighboring ganglion cells. The P–On, P–Off, and M cell filters generated distinct temporal output signal codes in response to visual stimulation of the RF–center pixels, and the 12 stimulus combinations of RF–periphery pixels with or without additional center pixels. The different temporal codes had a range of + / – 12 discrete amplitude values and 4 time values. C: Assuming that all epiretinal electrodes with exclusive inputs from one of the RF–filter outputs stimulate single ganglion cells, the central visual system, which is interpreted here as visual decoder (VD), had to resolve the high redundancy of the data stream of ganglion cell codes. This decoding process of VD was achieved in a two–step process. Results: (1) The VD mapping function from the highly redundant single ganglion cell codes onto visual percepts was successfully simulated as demonstrated by perfectly decoding / retrieving arbitrary visual input patterns from the redundant, parallel RF–filter data stream. (2) VD processed pattern shifts by exactly one pixel between a given visual pattern and the photosensor array to partly reduce the code redundancy. (3) In order to remove the remaining code redundancy, VD subsequently applied a logical decision sequence to the codes of neighboring ganglion cells. (4) Individual ganglion cell codes were permitted to contribute to a given visual percept only, if its code redundancy had been completely resolved. (5) This 2–step decoding procedure yielded perfect reconstruction of visual patterns. Conclusions: A. One essential task of the human central visual system for visual perception, namely the removal of afferent visual signal redundancy, can be significantly simplified by eye movements. B. Incorporation of relative movements between visual pattern and sensor array into the generation of RE output codes is proposed here to improve the visual percept quality of patients with tunable retina implants.
Keywords: electrophysiology: non–clinical • shape, form, contour, object perception • computational modeling