Purpose: An electronic implant that can bypass the damaged photoreceptors and electrically stimulate the remaining retinal neurons to restore useful vision has been proposed. A number of key questions remain to make this approach feasible. The goal of this thesis is to address the following 2 specific null hypotheses: (1) Stimulus parameters make no difference in the electrically elicited retinal responses. (2) Just as we have millions of photoreceptors, so it will take a device that can generate millions of pixels/light points to create useful vision.
Methods: For electrophysiologic experiments, 2 different setups were used. In the first setup, charge-balanced pulses were delivered to the retinal surface via electrodes inserted through an open sky approach in normal or blind retinal degenerate (rd) mice. In the second setup, the rabbit retina was removed under red light conditions from an enucleated eye and then maintained in a chamber while being superfused with oxygenated, heated Ames media. In both setups, stimulating electrodes and recording electrodes were positioned on the retinal surface to evaluate the effect of varying stimulation parameters on the orthodromic retinal responses (i.e., recording electrode placed between stimulating electrodes and optic nerve head). For psychophysical experiments, visual images were divided into pixels of light that could be projected in a pattern on the retina in up to 8 sighted volunteers. Subjects were asked to perform various tasks ranging from reading and face recognition to various activities of daily living.
Results: Electrophysiologic experiments: In a normal mouse, a single cycle of a 1-kHz sine wave was significantly more efficient than a 1-kHz square wave (P < .05), but no such difference was noted in either of the 8- or 16-week-old rd mouse groups (8-week-old, P = .426; 16-week-old, P = .078). Charge threshold was significantly higher in 16-week-old rd mouse versus both 8-week-old rd and normal mouse for every stimulus duration (P < .05). In all groups, short duration pulses (40, 80, and 120 microseconds) were more efficient in terms of total charge (the product of pulse amplitude and pulse duration) than longer (500 and 1,000 microseconds) pulses (P < .05). In all groups, applying a pulse train did not lead to more efficient charge usage (P < .05). Psychophysical experiments: In high-contrast tests, facial recognition rates of over 75% were achieved for all subjects with dot sizes of up to 31.5 minutes of arc when using a 25 x 25 grid with 4.5 arc minute gaps, a 30% dropout rate, and 6 gray levels. Even with a 4 x 4 array of pixels, some subjects were able to accurately describe 2 of the objects. Subjects who were able to read the 4-pixel letter height sentences (on the 6 x 10 and 16 x 16 array) seemed to have a good scanning technique. Scanning at the proper velocity tends to bring out more contrast in the lettering. The reading speed for the 72-point font is a bit slower than for the next smaller font. This may be due to the limited number of letters (3) visible in the window with this large font.
Conclusions: Specific parameters needed to stimulate the retina were identified. Delineating the optimum parameters will decrease the current requirements. Psychophysical tests show that with limited pixels and image processing, useful vision is possible. Both these findings should greatly simplify the engineering of an electronic retinal prosthesis.