Purchase this article with an account.
Deepa Prabhu, Clare MacMahon, Lisa Wise, Marten De Man, Chris McCarthy; Effect of ‘demagnification’ on localisation with Simulated Prosthetic Vision. Invest. Ophthalmol. Vis. Sci. 2020;61(7):925.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The integration of a digital ‘zoom-out’ feature into prosthetic vision devices is being actively explored to help overcome current limitations of decreased field of view (FoV) in these devices. The present simulated prosthetic vision study aimed to understand the immediate effects of demagnified viewing on localisation performance and its implications for training. Accuracy, time-to-locate the stimulus and head movements during a pointing task were compared for two FoV conditions: (i) implant matching (12° FoV) and (ii) camera matching/demagnified (58° FoV).
Ten sighted participants performed 2-blocks of a pointing task involving randomly presented circular stimuli. A 44-channel retinal implant with 12° FoV was simulated on a head-mounted display. For Block-1, a central patch of the camera image matching the simulated implant’s FoV (12°) was extracted and presented to the simulator. For Block–2, a demagnified view was created by compressing the full image taken from a 58° FoV camera into 12° FoV and then presenting it to the display via the simulator. A touchscreen monitor was used to present the targets and record pointing responses. Head and finger movements were tracked using the Qualisys Motion Capture System. This study was approved by the institution’s ethics committee.
Absolute errors with marginally higher mean and lower variance were observed with58° FoV (328±93.69, σ2 = 8778.53) compared with 12° FoV (277±96.19, σ2 = 9253.79). Although none of the performance metrics were found to be significantly different in individual 1-way ANOVAs, the identified location of stimulus positions on the touchscreen were found to be distributed over a larger area for 58° FoV compared with 12° FoV.
Our results suggest that presenting a larger coverage of scene to the narrower FoV of a typical retinal implant may help to encode a larger visual space while maintaining near equivalent visuo-motor performance to that achieved with a 1:1 image-to-implant FoV mapping. Relatively higher precision (less variance) with demagnified viewing indicates consistent repeatable biases in pointing error that can be effectively corrected with training. Therefore, demagnification can have potential advantages for inferring spatial relationships in the scene during visuo-motor tasks and provide improved outcomes with training.
This is a 2020 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only