Abstract
Purpose :
Patients with geographic atrophy implanted with the photovoltaic subretinal prosthesis (PRIMA, Pixium Vision) demonstrated form perception with acuity closely matching the 100μm pixel size. This success might be partly attributed to significant preservation of the retinal signal processing during network-mediated stimulation, as opposed to direct activation of the retinal ganglion cells (RGC). Here, we explore the role and limits of stimulation selectivity in prosthetic vision using measurements and modeling of retinal activation.
Methods :
An anatomically accurate model of the implanted area, including the PRIMA array under five retinal layers, was created in the Sim4Life software. The layers' thickness and the cells' length were adjusted to match the patients' optical coherence tomography data. The electric potential generated by the implant was used as the source of neural simulation, while the response to various patterns was assessed using 33 sustained-ON RGCs tiled above the implant. Illumination settings varied up to the maximum used in clinical trials: 9.8ms pulses at 3 mW/mm2 irradiance. Retinal stimulation was also assessed via electrophysiology in rats, with and without synaptic blockers.
Results :
Modeling shows that the stimulation threshold of RGCs is about 3 times higher than that of bipolar cells. Sufficiently large patterns, such as 4-pixel wide bars, can elicit direct dendritic and somatic activation of RGCs, but narrow patterns, such as Landolt C, even at highest illumination settings do not. Axonal stimulation was not observed with any pattern. Electrophysiology in rats demonstrated that the RGCs threshold was about 3 times higher than that of bipolar cells. Response to alternating gratings without blockers extended down to 30μm bars, while with synaptic blockers (direct activation of RGCs), there was no response to alternating gratings even for 160μm bars.
Conclusions :
Since patients' perception of wide and narrow patterns was accurate and visually similar, direct activation of RGCs does not seem to alter the network-mediated prosthetic vision. Rat experiments with blockers indicate that bipolar cell stimulation is crucial for high spatial resolution, while low-resolution output from direct activation of RGCs does not obscure the network-mediated response. These findings can provide essential guidance for designing future generations of subretinal implants.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.