Purchase this article with an account.
Nick Barnes, Adele F Scott, Ashley Stacey, Paulette Lieby, Chris McCarthy, Matt Petoe, Lauren N Ayton, Mohit Naresh Shivdasani, Nicholas C Sinclair, Janine Walker; Vision Processing with Lanczos2 Improves Low Vision Test Results in Implanted Visual Prosthetic Patients. Invest. Ophthalmol. Vis. Sci. 2014;55(13):1802. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The optics of the human eye can result in low-pass filtering of incoming light to the Nyquist frequency with respect to photoreceptor density, so that aliasing is not apparent and noise is suppressed. Vision processing using a Lanczos2 filter can perform the same function in prosthetic vision. Lanczos2 offers a better compromise than other practical filters in reducing aliasing and retaining sharpness. We evaluated the effectiveness for spatial localization of Lanczos2 Image Filtering at the Nyquist frequency (L2) compared to No Filtering (NF) and System Off (SO) in prosthetic vision.
Three participants with profound vision loss (bare light perception) due to retinitis pigmentosa were implanted with a retinal prosthesis comprising 20 stimulating electrodes (400 or 600 μm diameter) and two return electrodes (2000 μm) into the suprachoroidal space. Electrical stimulation occurred at 50-400 pulses/second, enveloped with a duty cycle of 50% over 2-5 seconds, with parameters optimal to each participant. Participants completed multiple trials (mean = 41.23/session) of the Basic Light and Motion Test Battery (BaLM): Localization of Light Test over five weekly experimental sessions. The test material consists of a black background with a white wedge extending from a centrally placed fixation disc on a computer monitor. Participants determine the location of the wedge that is placed randomly in four possible directions (4AFC). The study had a randomized design with participants masked to presentation order of the visual representation (i.e., L2, NF, and SO). Accuracy of localization was determined by the percentage of correct responses with a 25% chance rate and the 4AFC criterion of 62.50%.
Participants achieved significantly higher rates of accuracy in L2 trials with 72.5% (n = 160) correct responses compared to 55.0% for NF (n = 120, p = 0.001) and 24.3% for SO (n = 144, p < 0.0001). NF failed to achieve the success threshold and, as expected, SO achieved approximately the chance percentage of correct responses.
Lanczos2 filtering with cut-off at the Nyquist frequency is an effective approach for prosthetic vision under conditions of limited resolution and dynamic range. The superior accuracy rates achieved indicate that visual function performance can be enhanced using the Lanczos2 Filter, especially with the current hardware limitations of visual prostheses.
This PDF is available to Subscribers Only