Abstract
Purpose::
Since 2002, 6 human patients with severe retinitis pigmentosa have been implanted with 4x4 epiretinal electrode arrays with the goal of restoring some useful vision (Humayun, 1999). However, for a retinal prosthetic device to be successful, electrical stimulation patterns must produce predictable brightness levels while minimizing charge density, absolute charge, and peak current so that device integrity and patient safety are maximized. We show here that a highly constrained linear-nonlinear model can predict visual sensitivity to a wide variety of temporal stimulation patterns.
Methods::
Trains of biphasic current pulses were presented on a single electrode and varied in pulse width, frequency, and the number of pulses. Sensitivity was measured using standard yes-no threshold and suprathreshold brightness matching techniques. Thresholds were measured for 35 temporal stimulation patterns, across 10 electrodes over 2 subjects. Suprathreshold sensitivity was measured for a subset of electrodes and stimulation patterns.
Results::
Data were well fit using a highly constrained linear-nonlinear model, similar to recent models of retinal processing. This model includes integration of current with time scales similar to those of ganglion cells, rectification, an accelerating nonlinearity showing adaptation effects similar to retinal contrast adaptation, and a slower second stage of integration. This model can predict visual sensitivity for a wide variety of temporal pulse train configurations during single electrode stimulation. Analogous calculations can be made for charge density per pulse or overall charge across the entire pulse train.
Conclusions::
A wide variety of stimulation patterns can be described with a remarkably simple model with a small number of free parameters. Models such as that described here will be crucial in constructing temporal patterns of electrical stimulation that produce the desired perceptual brightness while satisfying a complex set of engineering constraints.
Clinical Trial::
www.clinicaltrials.gov NCT00279500 345345
Keywords: computational modeling • clinical (human) or epidemiologic studies: treatment/prevention assessment/controlled clinical trials • perception