[Back]

T. Esler:
"Simulating and optimizing electrical current flow and neuronal activation in retinal implants";
Supervisor, Reviewer: F. Rattay, A. Burkitt; University of Melbourne, Department of Electrical & Electronic Engineering, 2018; oral examination: 2018.

A number of challenges facing electrical stimulation of the retina necessitate more sophisticated methods
of stimulation in order to achieve targeted and meaningful perception. Computational models of
current
ow and neural activation in complex retinal tissue provide a means for the exploration of
novel stimulation strategies and the development of stimulus optimization techniques.
The rst component of this project develops multi-layered biophysical models of the retina that
capture the characteristic anatomical properties of each retinal layer. Models used in this research build
on single-layered models of electrical stimulation that include the microscopic structure and physiology
of neural tissue. A novel four-layer model is presented that models current
ow and passive membrane
activation due to simultaneous multi-electrode stimulation, taking into account the organized structure
of parallel bers in the nerve ber layer. Using this model, simulations are used to elucidate strategies
for the preferential activation of the target ganglion cell layer over the nerve ber layer in epiretinal
stimulation. These strategies indicate that preferential activation of retinal ganglion cells over passing
axons is achievable given carefully chosen electrode locations and stimulation waveforms.
The project then focuses on the validation of simpler linear-nonlinear models of retinal ganglion
cell activity for the purpose of optimizing multi-electrode array stimulation parameters. Models of
spiking probability are compared and validated against the developed multi-layered biophysical model
of retinal stimulation and current
ow in order to establish a biophysical basis for the simpli ed model's
structure. Results demonstrate that the linear electrical receptive eld of the linear-nonlinear model
matches the transmembrane currents induced by electrodes, known as the activating function, at the
site of action potential initiation. Importantly, this result allows for the typically experimentallydetermined
electrical receptive eld of the retinal ganglion cell to be estimated using a biophysical
model, yielding a biophysical version of linear-nonlinear model for biphasic stimulation.