Purchase this article with an account.
Michael Beyeler, Devyani Nanduri, James Weiland, Ariel Rokem, Geoffrey Boynton, Ione Fine; Optimizing stimulation protocols for prosthetic vision based on retinal anatomy. Journal of Vision 2018;18(10):205. doi: https://doi.org/10.1167/18.10.205.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Introduction: The field of electronic retinal prostheses is developing rapidly, with three varieties of retinal prostheses approved for commercial use and several others in development. However, many of these devices stimulate retinal axon fibers as well as cell bodies, leading to elongated and poorly localized percepts that severely limit the quality of the generated visual experience (Nanduri et al. 2011). We previously developed a computational model that describes these distortions and predicts a patient's perceptual experience for any electrical stimulation pattern (Horsager et al. 2009, Nanduri et al. 2012, Beyeler et al. 2017). However, improving the design of neuroprosthetic devices requires a solution to the inverse problem: What is the optimal stimulation protocol to elicit a desired visual percept? Methods: A simulated Argus II epiretinal prosthesis (Second Sight Medical Products Inc.) implant was placed on top of a map of ganglion axon pathways (Jansonius et al. 2009), designed to mimic known retinal anatomy, and used to generate predictions about the shape and location of visual percepts. The location and orientation of the implant with respect to the fovea and the optic nerve head was estimated using fundus images. The resulting predictions closely matched reported percepts when subjects were asked to trace the phosphenes generated by single electrodes on a touch screen. These synthetic percepts were then used as features in a regularized regression optimized to find stimulation protocols that would minimize perceptual distortions of Snellen letters. Results and Conclusions: Percepts produced with the optimized stimulation protocols partially compensated for the perceptual distortions caused by axonal stimulation: letters were much more recognizable than those generated using conventional protocols. Future work will include validating these results in patients and developing more sophisticated machine learning methods that can compensate for spatiotemporal distortions across a wider range of implants.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only