Purchase this article with an account.
A. C. Wainwright, J. J. Wang, G. Liew, Z. Y. Ping, W. Hsu, L. M. Li, T. Y. Wong, P. Mitchell, Sydney Childhood Eye Study; Comparison of Color, Red-Free and Enhanced Images in Measuring Fractal Dimension of the Retinal Vasculature. Invest. Ophthalmol. Vis. Sci. 2009;50(13):98.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Fractals can be used to describe naturally occurring branching patterns, and their utility is now being explored to determine whether they enhance our understanding of the microvasculature in health and disease. We developed a program (IRIS-Fractal) to calculate the fractal dimension (Df) of retinal vascular structures from digital images, and aimed to assess whether use of color, red-free or enhanced images affects Df as measured by IRIS.
We selected same-eye color and red-free retinal images of the right eyes of 100 children aged 12 years, participants of the Sydney Childhood Eye Study. Original color and red-free images were graded following standard protocols, and regraded after an enhancement process using Adobe Photoshop CS2. Enhancement consisted of applying the Auto Levels function followed by the Shadow/Highlight function, with default settings for each. To process 100 images took 5 minutes on a Dell Precision PWS380 P4 2.8GHz.
Mean pixel brightness values of the original color and red-free images was 54.65 and 33.16 respectively. Before image enhancement 1% of color and 100% of red-free images were ungradeable for Df using IRIS. Following enhancement mean brightness values improved to 96.14 for color and 97.67 for red-free, and all images were gradeable. The enhancement process increased image brightness, improved contrast and largely corrected for uneven illumination without unduly burning out lighter areas of an image. Mean Df of color images was 1.4648 before and 1.4706 after enhancement (paired t-test p <0.0001). Mean Df of enhanced red-free images was 1.4802, significantly higher than the mean Df of enhanced color images (paired t-test p <0.0001), but the two measures correlated well (r=0.80).
We show that applying a standard image enhancement procedure to color images results in higher Df measures than unenhanced images, and this enhancement is needed before red-free images can be graded. The highest Df measurements from enhanced red-free images suggests that image enhancement may improve IRIS’s ability to detect vascular structures and that enhanced red-free images may be ideal for Df measurement. Further validation is needed.
This PDF is available to Subscribers Only