Abstract
Purpose:
To study in-vitro the mechanisms of network-mediated electrical activation of ganglion cells in degenerate retina using a subretinal prosthesis.
Methods:
Photovoltaic arrays consisting of a hexagonal lattice of 70µm pixels separated by 5 µm trenches were placed subretinally on degenerate rat retinas (p100-p360, Royal College of Surgeons, RCS). Network-mediated responses were recorded from the retinal ganglion cells (RGCs) using a multielectrode array. We projected a sparse white noise, alternating square gratings and full-field contrast steps on the implant using 880nm light. Images were refreshed at 2Hz (1Hz grating contrast reversal). For the white noise, the implant was activated at 2Hz with 8ms, 5mW/mm2 pulses. Otherwise, it was illuminated at 20Hz with 4ms, 5mW/mm2 pulses. We displayed similar stimuli on the photoreceptor layer of a healthy rat retina using visible light for comparison.
Results:
In degenerate retina, half of the RGCs under electrical stimulation responded to the 2Hz image changes but not to every pulse of light. In healthy retina, all RGCs responded in the same way, demonstrating that flicker fusion and adaptation to static images take place with subretinal prosthetic stimulation.<br /> Prosthetic receptive field (RF) diameters in RCS rats (205 ± 69µm) were comparable to natural RF diameters in healthy animals (249 ± 44µm). RGCs responded to alternating gratings with stripe widths 65µm for prosthetic and 30µm for normal vision and for both, RGCs responses to grating alternation was frequency doubled (i.e., 2Hz for 1Hz alternation). These results are consistent with a subunit structure of the RFs with subunits connecting non-linearly to the RGCs, a feature observed across species and RGC types. Subunit diameter was 30µm in healthy retinas and ~65µm for prosthetic vision, matching the spacing between adjacent rows in the hexagonal array.
Conclusions:
Prosthetic and natural vision exhibited comparable RF diameters, adaptation to static images at high stimulation frequencies and evidence of non-linear summation in RF subunits, suggesting that with 70µm pixels, spatial performance is limited by the pixel pitch of the implant. Since stimulation thresholds are much lower than ocular safety limits, these results raise the possibility of decreasing the pixel size by a factor of 2. The implant would then be able to provide spatial resolution corresponding to a grating visual acuity of 20/120 in the human eye.