Non-parametric modeling of the optical nerve response by trans-corneal stimulation using differential neural networks

Nowadays, the field of biomedical intelligent stimulators has received more and more attention. Those devices have been applied for the treatment of several pathologies. Among others, the visual diseases have attracted special attention due the difficulties associated to obtain desired responses in the optical nerve. The trans-corneal stimulation is strongly dependent on many factors. One of the most important aspects relies on how to produce the required stimulation signal to produce the desired response. However, this is not an easy task, due to the relationship between the stimulation signals and the response is almost unknown. Within the modeling theory, it can be a good choice to select an adaptive technique to achieve a good approximation of the uncertain function relating the stimulation and response signals. Neural networks seem to be a good option to obtain such uncertain nonlinear functions. The differential neural network (DNN) is a class of neural networks used to reproduce continuous signals. Therefore, the DNN technique can be applied to generate the relation between the stimulation and response signals. In this paper, we have explored the possibility to use a set of several DNNs working in parallel to produce the aforementioned relationships. The DNN produces a set of models that can be used with the stimulated signals as inputs and to produce a similar signal to that monitored in the optical nerve. The set of DNN was successfully applied to reproduce the optical nerve response. A technological platform was produced to test the adaptive model suggested in this study. The device proposed in this paper was used to simulate the response in the optical nerve, to acquire the image that regulates the amplitude of these stimulation signals. The numerical simulations showed the closeness between the simulated signal and the trajectories produced by the DNN.