Abstract
Purpose :
Intracortical visual prostheses (ICVPs) are a novel class of devices that transmit visual stimuli from an external camera to an electrode array implanted in the visual cortex, thereby eliciting visual percepts (phosphenes) and restoring limited functional vision to the blind. The visual world must be encoded into patterns of electrode stimulation that produce intelligible images, requiring knowledge of where in the visual field each electrode produces a phosphene. This project aims to develop a multi-modality phosphene mapping method for calibration of ICVPs.
Methods :
We simulated the perceptual experience of electrode stimulation in an ICVP recipient by displaying phosphenes (points of light) to sighted subjects using a FOVE virtual reality headset. Subjects indicated the position of each phosphene within their visual field by making a saccade or pointing a finger towards the perceived target, and these movements were measured using a built-in eye-tracking feature and a LeapMotion camera, respectively. A set of known points was presented in order to determine calibration parameters for each subject, which were used to apply linear adjustments to the measurements to reconstruct the initial set of points.
Results :
Average R2 between presented and reconstructed coordinates were 0.97 and 0.96 for eye tracking and 0.92 and 0.97 for finger tracking in the x and y directions, respectively. By applying linear adjustments, a set of calibration points was recreated with an average root-mean-square error (RMSE) of 0.131 normalized units (fig. 1a). Similar methods applied to finger tracking produced maps with an average RMSE of 0.226 (fig 1b).
Conclusions :
Phosphene maps reconstructed by tracking saccades had, on average, less random error and systemic distortion than those produced by tracking finger position. Eye tracking thus appears more promising as a primary method for ICVP calibration, whereas finger tracking may be useful as an adjunct or as a standalone phosphene mapping method in individuals without intact eye movements. Next steps include applying nonlinear adjustments to the measurements and integrating data from both modalities with relative mapping of adjacent phosphene pairs.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.