July 2018
Volume 59, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2018
Discrete wavelet transform of the electroretinogram for glaucoma classification; choice of mother wavelet by variable ranking
Author Affiliations & Notes
  • Marc Sarossy
    CERA, University of Melbourne, Melbourne, Victoria, Australia
    Engineering, RMIT University, Melbourne, Victoria, Australia
  • Behzad Aliahmad
    Engineering, RMIT University, Melbourne, Victoria, Australia
  • Dinesh Kant Kumar
    Engineering, RMIT University, Melbourne, Victoria, Australia
  • Footnotes
    Commercial Relationships   Marc Sarossy, None; Behzad Aliahmad, None; Dinesh Kumar, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science July 2018, Vol.59, 5098. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Marc Sarossy, Behzad Aliahmad, Dinesh Kant Kumar; Discrete wavelet transform of the electroretinogram for glaucoma classification; choice of mother wavelet by variable ranking. Invest. Ophthalmol. Vis. Sci. 2018;59(9):5098.

      Download citation file:

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

  • Supplements

Purpose : The discrete wavelet transform (DWT) is a fast and efficient technique for analysing the electroretinogram(ERG) revealing both time and frequency information about the signal. Choice of mother wavelet is important to best extract the desired components. In this work we optimize selection of the Daubechie’s wavelet for ERGs collected with a Photopic Negative Response (PhNR) stimulus by variable ranking and feature selection in the classification of glaucomatous and non-glaucomatous eyes.

Methods : Left eyes of 21 glaucoma patients (mean age 66.4) and 18 normal aged matched controls (60.7) were tested. ERGs were measured with the LKC RetEval using a orange stimulus on a blue background. 1.9Khz sample rate was used. 200 sweeps were collected for each recording session. Raw data was centred and scaled to a mean RMS amplitude of 1mV and truncated to 256 samples - DWT requires signal length to be a power of 2. 8 level (full) DWT was performed on the average of the raw sweeps using the WMTSA package in R with all available wavelets from daublet, symlet, best localized and coiflet families. Variable ranking was performed with the Caret package in R on the detail coefficients. PhNR amplitude was measured from the processed waveform generated by the device.

Results : Good quality tracings were obtained from all subjects. There was a significant difference in mean PhNR amplitudes between glaucoma (-3.47 ± 2.46 mV) and control groups (-5.49 ±1.98 mV) (p=0.023). Of mother wavelets used for the DWT, Coiflet 30 resulted in the highest variable importance in the time-frequency domain of the PhNR. The 3rd 64 sample width wavelet (33ms –the ‘d6(2)’ coefficient) – the location of the PhNR - showed a small difference in amplitude between the glaucomas and controls but did not reach statistical significance (p=0.21)

Conclusions : The DWT transforms a time domain signal into the wavelet domain encoding both time and frequency. Theoretically, it can reveal the timing of changes of the frequency components of the signal. The DWT is non redundant and the dimensionality remains that of the original signal. The timing of the individual wavelets within the decomposition tree are fixed. In this work we show that even with best choice of mother wavelet, use of a single coefficient of the DWT is inferior to traditional calculation of the PhNR in classifying responses into glaucoma and non glaucoma

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.


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.