Visual Acuity of Simulated Thalamic Visual Prostheses in Normally Sighted Humans

Abstract

Simulation in normally sighted individuals is a crucial tool to evaluate the performance of potential visual prosthesis designs prior to human implantation of a device. Here, we investigated the effects of electrode count on visual acuity, learning rate and response time in 16 normally sighted subjects using a simulated thalamic visual prosthesis, providing the first performance reports for thalamic designs. A new letter recognition paradigm using a multiple-optotype two-alternative forced choice task was adapted from the Snellen eye chart, and specifically devised to be readily communicated to both human and non-human primate subjects. Validation of the method against a standard Snellen acuity test in 21 human subjects showed no significant differences between the two tests. The novel task was then used to address three questions about simulations of the center-weighted phosphene patterns typical of thalamic designs: What are the expected Snellen acuities for devices with varying numbers of contacts, do subjects display rapid adaptation to the new visual modality, and can response time in the task provide clues to the mechanisms of perception in low-resolution artificial vision? Population performance (hit rate) was significantly above chance when viewing Snellen 20/200 optotypes (Log MAR 1.0) with 370 phosphenes in the central 10 degrees of vision, ranging to Snellen 20/800 (Log MAR 1.6) with 25 central phosphenes. Furthermore, subjects demonstrated learning within the 1–2 hours of task experience indicating the potential for an effective rehabilitation and possibly better visual performance after a longer period of training. Response time differences suggest that direct letter perception occurred when hit rate was above 75%, whereas a slower strategy like feature-based pattern matching was used in conditions of lower relative resolution. As pattern matching can substantially boost effective acuity, these results suggest post-implant therapy should specifically address feature detection skills.

Simulation of thalamic prosthetic vision: reading accuracy, speed, and acuity in sighted humans

“…The simulation is described in detail in a previous publication (Bourkiza et al, 2013), and is summarized as follows. All current artificial prosthesis designs that include stimulation through sets of microwire contacts provide the recipient with a coarse visual experience made up of a set of isolated pixel-like percepts called phosphenes.…”

“…In preceding work, we used a simulation of thalamic vision to investigate the relationship between electrode number and visual acuity using a standardized test in sighted humans (Bourkiza et al, 2013). Here, we use a more advanced simulation to examine the same relationship for reading performance, again employing a standardized test in sighted humans.…”

Perceptual learning in a non-human primate model of artificial vision

“…While perceptual grouping has often been studied with simple stimuli such as gratings or lines910, an artificial vision based paradigm presents an excellent opportunity to understand learning of perceptual grouping, with potentially important translational benefits. We previously created and tested a gaze-contingent thalamic visual prosthesis simulation in humans for both individual letter recognition11 and reading12. In these two studies we assessed behavioral letter recognition and reading performance as a function of font size and phosphene count in tasks that required well-developed skill in perceptual grouping.…”

“…Interestingly, we also found that performance improvements can be seen within a single session, suggesting that learning is important for optimal visual prosthesis utility, and may have substantial impact on both device design and post-implant therapy. In order to study this learning in greater detail, and as preparation for electrophysiological work to be reported in a separate publication, we developed an animal model for visual prostheses where, after extensive training, animals are capable of performing similarly to humans on the letter recognition task11. Results from that training effort are reported here.…”

“…many months). We previously found that humans are capable of performing letter-based perceptual grouping tasks with no training1112 and exhibited learning within a single session11. Others have observed perceptual learning in humans over a small number of days with trial counts on the order of thousands2225.…”