Wireless trans-corneal stimulus for the optical nerve based on adaptive modeling using continuous neural networks

Retinal prosthesis design has become a hot field of researching around the world. Restoring partial vision to the blind patients that suffer from degenerative disease has become an important medical and scientific task. However, there are some doubts on how to propose the stimulation signals. The same question arises when the stimulation may be done by trans-corneal or transdermic pathways. One method that could be used is to apply a no-parametric algorithm to obtain a nonlinear model representing the relationship between the optical nerve response signal and the stimulus inputs. Then, it can be applied an inverse model methodology to identify the unknown inputs required to obtain the desired optical nerve response. In this study, we proposed an adaptive modeling based in continuous neural networks (CNN) to obtain an artificial model of the relationship between the optical nerve response and the selective stimulation. This model tries to determine the adequate stimulation signals that will be applied on the trans-corneal or transepidermic part of the eye. Indeed, the input signal effectiveness will be measured as the degree of accuracy obtaining the desired response in the optical nerve. A set of CNN working as a parallel identifier provides the adaptive model of the aforementioned relation. An artificial optical nerve response was developed as well as the electrical stimulator for the trans-corneal area. These both designs were applied into the CNN identifier to test the methodology suggested in this paper. The numerical results demonstrate the accuracy achieved by the modeling algorithm.