A GMM-based feature extraction technique for the automated diagnosis of Retinopathy of Prematurity

摘要: Retinopathy of Prematurity (ROP) is an ophthalmic disease that a leading cause childhood blindness throughout the world. Accurate diagnosis ROP vital to identify infants who require treatment, which can prevent blindness. Arterial tortuosity and venous dilation in retina are important signs ROP, so it necessary extract these features from points on vessels or vessel segments. Then, image represented with statistics such as minimum, maximum mean values. However, provide biased estimates contains both healthy abnormal vessels. In this work, we present novel feature extraction technique represents each parameters two-component Gaussian Mixture Model (GMM). Using features, performed classification experiments manually segmented retinal dataset consisting 77 images. The results show GMM-based outperform other based classical statistics, accuracy over 90%. Moreover, if extracted whole without distinguishing veins arteries, proposed better performance compared using traditional statistics.