Purchase this article with an account.
A. C. Wainwright, G. Liew, B. Taylor, Z. Y. Ping, W. Hsu, M. L. Lee, T. Y. Wong, P. Mitchell, J. J. Wang, Blue Mountains Eye Study; Effect of Image Brightness, Focus and Contrast on Measurement of Multifractal Dimension of Retinal Vasculature. Invest. Ophthalmol. Vis. Sci. 2008;49(13):2114.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Multifractal dimension (Dref) of the retinal microvasculature is a new global measure of vascular branching architecture and can be measured from retinal photographs. Dref may be useful for assessing overall health of retinal circulation. In this report we investigate whether image quality (image brightness, focus and contrast) affects measures of Dref.
From the Blue Mountains Eye Study population we selected 30 right-eye, disc-centered, greyscale images of participants free from hypertension or diabetes which were judged to have good image clarity. We cropped each image with a circular mask with a diameter of 3.5x optic disc diameter to provide a consistent image field. We altered the cropped images with Adobe Photoshop CS2 (Adobe Systems) to give image sets which varied in parameters including brightness (-40, -30, -20, -10, +20, +30, +40), focus (increasing blur using ‘lens blur’ filter: hexagonal iris, radius 10, 20, 30), and an expanded-contrast set (‘auto levels’ operation: default setting). We graded both the unaltered and altered image sets using the IRIS-Fractal program following a standard protocol. We compared mean Dref of the altered image sets with that of the unaltered set using SAS.
The mean Dref of the unaltered image set was 1.4542. Decreasing image brightness (b) was associated with decreasing Dref measures (b -20: Dref = 1.4479, p = 0.039; b -30: Dref = 1.4411, p = 0.0052; b -40: Dref = 1.4113, p < 0.0001). Increasing brightness, however, did not change Dref (b +10: Dref = 1.4552, p = 0.116; b +20: Dref = 1.4544, p = 0.27; b +30: Dref = 1.4548, p = 0.33; b +40: Dref = 1.4541, p = 0.60). Minimal blurring of the images did not significantly alter Dref (lens blur r10: Dref = 1.4553, p = 0.12) while blurring with lens blur (lb) r20 or more decreased the Dref (lb r20: Dref = 1.4374, p < 0.0001; lb r30: Dref = 1.4117, p < 0.0001). Applying a maximal contrast stretch operation to the image set increased Dref (Dref = 1.4570, p = 0.0054).
Variations in image quality (image brightness, focus and contrast) can significantly affect the Dref measurement of the retinal microvasculature. While the absolute differences reported are small, this potential source of measurement error should be further investigated.
This PDF is available to Subscribers Only