May 2007
Volume 48, Issue 13
ARVO Annual Meeting Abstract  |   May 2007
Automated Detection of Venous Beading in Severe Non-Proliferative Diabetic Retinopathy
Author Affiliations & Notes
  • K. Belkacem-Boussaid
    Biomedical, Kestrel Corporation, Albuquerque, New Mexico
  • G. Zamora
    Biomedical, Kestrel Corporation, Albuquerque, New Mexico
  • S. Nemeth
    Biomedical, Kestrel Corporation, Albuquerque, New Mexico
  • A. Das
    Ophthalmology, University of New Mexico Health Sciences Center, Albuquerque, New Mexico
  • Footnotes
    Commercial Relationships K. Belkacem-Boussaid, None; G. Zamora, None; S. Nemeth, None; A. Das, None.
  • Footnotes
    Support NEI grant 1R43EY017472-01.
Investigative Ophthalmology & Visual Science May 2007, Vol.48, 5023. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      K. Belkacem-Boussaid, G. Zamora, S. Nemeth, A. Das; Automated Detection of Venous Beading in Severe Non-Proliferative Diabetic Retinopathy. Invest. Ophthalmol. Vis. Sci. 2007;48(13):5023.

      Download citation file:

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

  • Supplements

Purpose:: To demonstrate the feasibility of an automatic system for the screening of severe non-proliferative diabetic retinopathy (NPDR) subjects in order to monitor their retinal characteristics for quantitative detection of venous beading, a high risk feature for advancing into the proliferative stage of retinopathy.

Methods:: Using a sequence of image processing and analysis algorithms, a combination of the difference of two Gaussian matching filters and a thresholding based entropy method, we extracted the veins from the retinal vasculature of patient with severe NPDR. A set of ten, 8 bit low resolution, images has been used to test the algorithms. We divided each vein into 72 pixel segments. The diameters along each segment are measured. The kurtosis, which is a statistical measure known as the measure of peakedness of a probability distribution, is then calculated on the set of extracted diameters. The notion of the kurtosis is used here to objectively quantify the venous beading feature in the retinal image. We are proposing a new definition of this quantification; Venous Beading Index (VBI).

Results:: The preliminary results show high correlation between the VBI and the manual assessment of the venous beading by a highly trained grader. The higher the value of the VBI the more likely there is presence of venous beading. A value of the VBI equal to 3.35 was a clear indication of venous beading whereas a value of the VBI equal to 0.95 was not indicative of venous beading.

Conclusions:: The preliminary results show the feasibility of an automatic screening tool to detect and quantify one of the high risk features, venous beading, for conversion to proliferative diabetic retinopathy. This computer aided tool could help the eye practitioner identify subjects with high risk venous characteristics for consideration of imminent treatment or close follow-up.

Keywords: diabetic retinopathy • image processing • retina 

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.