Neural NoiseRetinal prostheses are neural-implants designed to restore visual perception in individuals with outer retinal degenerative diseases such as retinitis pigmentosa and age-related macular degeneration. These devices bypass degenerated photoreceptors by electrically stimulating the surviving inner retinal neurons — ganglion cells, bipolar cells, or both — to evoke visual percepts (phosphenes) that convey spatial information about the visual scene.
Three main approaches have been pursued: epiretinal devices stimulate from the ganglion cell side of the retina, subretinal devices are placed in the photoreceptor layer to stimulate bipolar cells, and suprachoroidal devices are implanted outside the retina for easier surgery at the cost of lower spatial resolution. The Argus II epiretinal prosthesis became the first commercially approved retinal implant, and multiple other systems have reached clinical trials, collectively demonstrating that electrical stimulation can produce stable visual percepts in blind individuals.
Key challenges include achieving spatial resolution sufficient for functional tasks like reading and navigation, managing the mismatch between the retinotopic organization of electrical stimulation and natural visual processing, ensuring long-term biocompatibility and device reliability, and developing image processing strategies that optimize the limited information channel. Advances in electrode technology, stimulation encoding strategies, and the integration of AI-based scene understanding are shaping the next generation of visual prostheses.