Purchase this article with an account.
Samuel Weinreb, Roksana Sadeghi, Liancheng Yang, Gayatri Kaskhedikar, Philip Troyk, Gislin Dagnelie; Optimizing measurement of targeted visuomotor responses to map phosphenes in intracortical visual prostheses. Invest. Ophthalmol. Vis. Sci. 2020;61(7):4274.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Intracortical visual prostheses (ICVPs) aim to restore sight by stimulating the visual cortex via penetrating electrode arrays. Ideally, these electrodes would access only primary visual cortex, including within the calcarine fissure, acting at a single level of the visual pathway; however, due to surgical and technical limitations, they will instead sit on the posterior occipital lobe in early clinical trials. As such, they will recruit phosphenes from three distinct retinotopic maps in V1, V2, and V3/Vp, which represent overlapping areas of visual space. Borders between these regions are variable and not anatomically defined; on a local scale, retinotopy is not strict; thus, locations and spatial relationships of phosphenes cannot be predicted a priori and must be mapped in each ICVP recipient.
Phosphenes were simulated by displaying dots of light to sighted subjects in a virtual reality headset, including 8 calibration points followed by a set 32 random dots repeated 3 times, all within the left hemifield ≤25° eccentricity. Subjects indicated the perceived location of each dot by pointing a finger in its direction while maintaining central fixation. The dot then disappeared and a saccade was made to its remembered location. These movements were measured using an external infrared camera and built-in eye-tracking, respectively. The calibration data were fitted to the corresponding points by fitting second order (X,Y) polynomials, which were then applied to raw data from the random dots to estimate their true locations. The accuracy of raw and corrected data was evaluated by calculating linear (X,Y) correlation coefficients with true dot locations.
Across 6 subjects, correlation between the raw data and actual random dot position showed mean R2 in (X,Y) of (0.79±0.15,0.96±0.03) for eyes and (0.49±0.32,0.83±0.17) for finger. With polynomial correction, mean R2 was (0.81±0.15,0.95±0.03) for eyes and (0.48±0.30,0.79±0.27) for finger.
Calibration, which would not be possible in blind subjects, did not improve the accuracy of estimated dot positions. Compared to pointing, eye movements are a more accurate proxy for the perceived location of a phosphene, but may not be reliable in patients after prolonged visual deprivation. We are exploring head movement recording, as well as paired presentation and relative location mapping as alternative modalities.
This is a 2020 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only