Abstract
Purpose: :
To determine if compressing an image has an effect on retinal vessel caliber.Background: Compression of digital images is an important consideration while evaluating photographic data for clinical and epidemiologic trials. Advantages of compression are reduced file size which facilitates transmission and storage while compromising quality of the image.IVAN is semiautomated software that measures retinal vascular caliber from digital images of the fundus. The software identifies vessel edges and measures the width of arterioles and venules. The process is supervised by an ocular disease evaluator (ODE), who edits the software drawn measurements as necessary.
Methods: :
Twenty Images, with good photo quality and no previous image compression, were selected from the reading center database. Original size of each image was approximately 14 MB. Using Adobe Photoshop 3, the images were compressed progressively, and graded using IVAN v1.3. A screenshot was taken after the first evaluation and subsequent grading was based off of this image, to ensure measurement of same sections. Intra-grader reproducibility was evaluated on 2 image sets. Data was analyzed by comparing the calculated Central Retinal Arteriole Equivalent (CRAE) and the Central Retinal Venular Equivalent (CRVE) of the original image and image at each compression level.
Results: :
Based on the Wilcoxon signed rank test, there was no statistical difference in measuring the CRAE up to a compression ratio of 29:1 and 130:1 for CRVE (p<0.001). It was noted that as the compression was increased further, the retinal vessel caliber increased significantly.
Conclusions: :
Image compression can cause loss in the sharpness of vessel borders and hence lead to an artifactual broadening of the vessel, causing a faulty vascular caliber measurement. Compression of a fundus image to a ratio of 29:1, approximate file size 14MB to 0.5MB, is a reasonable option, as it combines the advantages of reduced file size and continue to yield accurate vascular caliber measurements.
Keywords: image processing • clinical (human) or epidemiologic studies: systems/equipment/techniques • retina