April 2010
Volume 51, Issue 13
Free
ARVO Annual Meeting Abstract  |   April 2010
Artificial Neural Network for Estimating the Functional Output of Retinal Ganglion Cells
Author Affiliations & Notes
  • I. Sliesoraityte
    Centre for Ophthalmology, Institute for Ophthalmic Research, University of Tuebingen, Tuebingen, Germany
  • E. Troeger
    Centre for Ophthalmology, Institute for Ophthalmic Research, University of Tuebingen, Tuebingen, Germany
  • A. Kurtenbach
    Centre for Ophthalmology, Institute for Ophthalmic Research, University of Tuebingen, Tuebingen, Germany
  • E. Zrenner
    Centre for Ophthalmology, Institute for Ophthalmic Research, University of Tuebingen, Tuebingen, Germany
  • Footnotes
    Commercial Relationships  I. Sliesoraityte, None; E. Troeger, None; A. Kurtenbach, None; E. Zrenner, None.
  • Footnotes
    Support  None.
Investigative Ophthalmology & Visual Science April 2010, Vol.51, 2485. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      I. Sliesoraityte, E. Troeger, A. Kurtenbach, E. Zrenner; Artificial Neural Network for Estimating the Functional Output of Retinal Ganglion Cells. Invest. Ophthalmol. Vis. Sci. 2010;51(13):2485.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: : To estimate the functional output of retinal ganglion cells in foveal, parafoveal and perifoveal regions using an artificial neural network.

Methods: : The data based on 120 eyes (60 retinitis pigmentosa patients (RP) and 60 healthy age-matched controls) was used to generate the 1200 inputs for an artificial neural network. The VERIS system was applied to record multifocal pattern electroretinograms (mfPERG). The normalized energy, obtained via mfPERG responses was used to evaluate the functional output of retinal ganglion cells within fovea, parafovea and perifovea in RP and healthy subjects’ eyes using semi-automated software developed by the authors. The retinal ganglion cells loss function, quantifying the amount of functioning ganglion cells was mathematically approximated using artificial neural network based on prognosis algorithm. The probability of the correct prognosis of the artificial neural network was determined.

Results: : In healthy subjects the normalized energy of the functioning retinal ganglion cells was 213.46 units, 41.85 units, and 4.9 units in foveal, parafoveal and perifoveal regions, respectively (p<0.001); whereas in RP patients it was 6.55 units, 2.4 units, and 0.68 units, respectively. The quantity of functioning retinal ganglion cells was 361 620 cells, 70 897 cells, and 11 9375 cells in healthy subjects; and 11 093 cells, 4 075 cells, 16 580 cells in RP within foveal, parafoveal and perifoveal regions, respectively (p<0.001). The probability of a correct prognosis of retinal ganglion cell loss using artificial neural network was 0.97 (95% Confidence interval 0.95, 0.99), 0.87 (95% Confidence interval 0.85, 0.91), and 0.95 (95% Confidence interval 0.93, 0.96) within foveal, parafoveal and perifoveal region, respectively (p<0.01).

Conclusions: : The functional output of the inner retina significantly decreases with a reduction in the quantity of functioning retinal ganglion cells within foveal, parafoveal and perifoveal regions. The developed artificial neural network proved to be reliable for estimating the spatial distribution of retinal ganglion cells loss in an individual subject. This provides with important clues for the optimal localization of sub- or epi-retinal implants in blind, intending to receive retinal prostheses patients.

Keywords: ganglion cells • electrophysiology: clinical • computational modeling 
×
×

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.

×