Fractal analysis of the retinal vascular network in fundus images

Annu Int Conf IEEE Eng Med Biol Soc. 2007:2007:6456-9. doi: 10.1109/IEMBS.2007.4353837.

Abstract

Complexity of the retinal vascular network is quantified through the measurement of fractal dimension. A computerized approach enhances and segments the retinal vasculature in digital fundus images with an accuracy of 94% in comparison to the gold standard of manual tracing. Fractal analysis was performed on skeletonized versions of the network in 40 images from a study of stroke. Mean fractal dimension was found to be 1.398 (with standard deviation 0.024) from 20 images of the hypertensives sub-group and 1.408 (with standard deviation 0.025) from 18 images of the non-hypertensives subgroup. No evidence of a significant difference in the results was found for this sample size. However, statistical analysis showed that to detect a significant difference at the level seen in the data would require a larger sample size of 88 per group.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Fractals*
  • Humans
  • Hypertension / pathology
  • Image Interpretation, Computer-Assisted / methods*
  • Retinal Vessels / pathology*
  • Stroke / pathology