May 2005
Volume 46, Issue 13
ARVO Annual Meeting Abstract  |   May 2005
Simulating Prosthetic Vision
Author Affiliations & Notes
  • L.E. Hallum
    Biomedical Engineering, Univ of New South Wales, Sydney, Australia
  • S.C. Chen
    Biomedical Engineering, Univ of New South Wales, Sydney, Australia
  • P.J. Preston
    Biomedical Engineering, Univ of New South Wales, Sydney, Australia
  • G.J. Suaning
    School of Engineering, University of Newcastle, Newcastle, Australia
  • N.H. Lovell
    Biomedical Engineering, Univ of New South Wales, Sydney, Australia
  • Footnotes
    Commercial Relationships  L.E. Hallum, None; S.C. Chen, None; P.J. Preston, None; G.J. Suaning, None; N.H. Lovell, None.
  • Footnotes
    Support  None.
Investigative Ophthalmology & Visual Science May 2005, Vol.46, 1522. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      L.E. Hallum, S.C. Chen, P.J. Preston, G.J. Suaning, N.H. Lovell; Simulating Prosthetic Vision . Invest. Ophthalmol. Vis. Sci. 2005;46(13):1522.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

We report the psychophysical equivalence of modulating phosphene (P) size and intensity for a simulated face recognition task. Further, we propose a more realistic means of simulating P arrays, based on subjective reports from acute surgical trials (Rizzo et al., IOVS 44, 5362).


Subjects (n=38) were afforded simulated prosthetic vision via a computer monitor and mouse. For each subject, two simulations were employed each involving the recognition of 20 faces (2AFC paradigm): (1) the P array comprised 400 circular Ps arranged in a regular hexagonal mosaic (RHM) with min. P–to–P spacing 0.45 arcdeg; (2) as per (1) but 100 Ps and 0.90 arcdeg. At random, for 16 subjects, the sizes of Ps were modulated according the the "underlying" stimulus (face). For 22 subjects, the intensities were modulated. Five seconds' scanning with the mouse was allowed prior to 2AFC.


For the 400/size simulation, faces were recognized at an average rate of 90.7+6.8% as compared with 87.8+7.4% for the 400/intensity task. For the 100/size simulation, faces were recognized at an averaged rate of 63.8+11.5% as compared with 64.1+16.7% for 100/intensity.


The figure depicts what is possibly a more realistic means of simulating prosthetic vision. P locations were derived via uniform jitter applied to a RHM. Average P spacing corresponds to approx. 3.5 arcdeg. This is a rough approximation to disordered P locations reported in clinical trials (Humayun et al., Vis Res 43, 2573–). Rizzo et al. provided subjective reports of P profiles from a surgical trial, one of which (parent P) was used to generate Ps (children) in the figure as follows: (1) the Fourier transform (FT) of the radial signature of the parent was derived; (2) phases of Fourier terms were perturbed and moduli of terms scaled according to desired size of the child; (3) the modified FT was then inversed, forming the child's radial signature, which is used to form the child profile.


Further human trials, plus modeling retinal charge injection, will serve to inform the jitter, perturbation, and scaling steps above, making for more realistic simulations in the sighted.



Keywords: low vision • face perception • image processing 

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.