Simulation of the dynamic range of retinal ganglion cells during microelectrode stimulation

Abstract

Electrical stimulation of retinal ganglion cells (RGC) plays an important role in the creation of retinal prostheses which aim to restore vision in people suffering from degenerative retinal diseases by stimulating the remaining healthy retinal neurons. In order to better investigate the influence of electrical stimulation on RGCs, in addition to experimental studies, many computational RGC models have been developed. Given that most of these models are based on the deterministic well-known Hodgkin and Huxley model, they also lack a stochastic component that would mimic the natural channel noise present in cells. When we calculate the stimulus threshold with a deterministic model, we lose information about the spiking probability, which increases from 0 to 1 as a function of stimulus intensity following thereby the shape of a sigmoidal curve. In this thesis we study the stochastic behavior of RGCs by calculating dynamic range (DR), defined as the stimulus intensity range in which spiking probability increases from 0.1 to 0.9. The main aim of the thesis is to analyze DR in dependency on the stimulated part of the RGC, since different cell sections show different geometrical and biophysical properties. For that, a multi-compartment RGC model based on Fohlmeister and Miller model from 1997 was implemented in NEURON and Python. A stochastic component was added to the model by injecting the maximum sodium conductance dependent noise current into each cell compartment. We observed that, regardless of the type of stimulation (intracellular or extracellular), DR values given as percentage seem to show a somewhat inverse relationship to threshold values. With intracellular stimulation, a steeper sigmoidal curve of spiking probability was observed for axonal than for somatic compartments, as reported by previous experimental studies. With extracellular stimulation this behavior was observed only after increasing channel noise at soma. While estimating DR with an upper threshold, similar DR values given in μA (not normalized to threshold) as with a lower threshold were observed, which could indicate from previous studies reported characteristic U-shape of spiking probability as a function of stimulus intensity.