An actor-model framework for visual sensory encoding

A fundamental challenge in neuroengineering is determining a proper input to a sensory system that yields the desired functional output. In neuroprosthetics, this process is known as sensory encoding, and it holds a crucial role in prosthetic devices restoring sensory perception in individuals with disabilities. For example, in visual prostheses, one key aspect of image encoding is to down-sample the images captured by a camera to a size matching the number of inputs and resolution of the prosthesis. Here, we show that down-sampling an image using the inherent computation of the retinal network yields better performance compared to a learning-free down-sampling encoding. We validated a learning-based approach (actor-model framework) that exploits the signal transformation from photoreceptors to retinal ganglion cells measured in explanted retinas. The actor-model framework generates down-sampled images eliciting a neuronal response in-silico and ex-vivo with higher neuronal reliability to the one produced by original images compared to a learning-free approach (i.e. pixel averaging). In addition, the actor-model learned that contrast is a crucial feature for effective down-sampling. This methodological approach could serve as a template for future image encoding strategies. Ultimately, it can be exploited to improve encoding strategies in visual prostheses or other sensory prostheses such as cochlear or limb.