Computational Modelling of Functionally-Identified Retinal Ganglion Cells using a Multi-Objective Optimisation Approach

Retinal neuroprostheses aim to restore functional visual percepts to patients suffering from retinal degenerative diseases such as retinitis pigmentosa (RP) and age-related macular degeneration (AMD). In such patients, it is desirable to reconstruct a useful sense of artificial vision by selectively activating different neuron populations in a planned sequence and spatial pattern. However, current retinal neuroprostheses have limited ability in targeting different retinal neuron types. Improvements in the field of prosthetic vision are highly dependent on better understanding the fundamental mechanisms underlying retinal ganglion cell (RGC) electrical stimulation, and how these can be quantitatively controlled through artificial stimulation. The aim of this thesis is to develop accurate computational models of functionally-distinct RGCs to assist in the further understanding of biophysical mechanisms underlying RGC activation, so that more sophisticated stimulation schemes can be developed.Morphologically-realistic and functionally-accurate ON and OFF RGC models were developed by integrating multiple experimental information and biophysical principles, allowing the contribution of various morphological and intrinsic RGC properties in shaping RGC response patterns to be isolated. The multiple data used to optimise model parameters consisted of patch-clamp whole cell recordings of RGC spiking activity in the presence of multiple intracellular current injections, as well as associated action potential (AP) phase plots.In addition, the optimised RGC models were used to gain insights into the mechanisms underlying selective RGC responses to 2 kHz electrical stimulation. By adjusting the extracellular stimulus amplitude across a wide range of values, the models were able to reproduce the distinct patterns of excitation observed experimentally, suggesting the utility of this approach in developing stimulation strategies. The RGC modelling approach developed in this thesis will facilitate testing of a wide range of stimulus waveforms that aims for selective or differential activation of targeted RGC types, resulting in a dramatic improvement in the quality of prosthetic vision