Computer Assisted 'Top-Down' Assessment of Diabetic Retinopathy

摘要

Diabetic retinopathy (DR) is a major cause of blindness in the develope d world, causing vision problems that are preventable if adequate screening and treatment systems are available. In the last ten years automated computer-based assessment for the presence of DR and its progression has improved, given high image quality. Current automated assessment is based on a bottom -up approach, which aims to detect lesions defined by ophthalmologists such as microaneu-rysms or new retinal vessel growth. An alternate perspective is to find new feature variables that may be able to identify not only certain stages of disease progression but may also be useful in identifying the risk of disease progression: a top-down approach. Our research proposes a novel feature variable, the colourisation index, and this is the first publication to consider the top -down approach. Ninety images were captured from non -dilated eyes using a non -mydriatic camera and imported into the coloriz ation index programme. A significant difference between no DR and retinopathy present (P < 0.05) was found. Further, usin g ANOVA and Fischer's LSD post hoc test, significant differences were found between no DR versus moderate DR (p = 0.003); no DR versus severe DR (p = 0.016) and mild DR versus moderate DR (p = 0.022). The index provides a quantitative result for the optic disc colour mix, suggesting that the colourisation index may be a useful tool in classifying disease progression.