A Curvature Based Approach for the Automated Screening of Retinopathy of Prematurity in Preterm Infants

摘要

Retinopathy of Prematurity (ROP) is a curable retinal disorder which mainly affects premature infants. The early treatment of the disease starts with the diagnosis of Plus disease which is characterised by excessive tortuosity and dilation of the retinal blood vessels. The method used for the disease diagnosis involves a retinal specialist, to visually compare the fundus vasculature with a gold standard image. This technique often lead to errors due to inter expert variability in the final diagnostic decision. So there arises the need for a computer aided system for the early detection of the disease. In the presented work, the presence of ROP is investigated by evaluating the retinal blood vessel tortuosity in the fundus images using a curvature based method. The method includes vessel delineation using COSFIRE filters followed by quantification of its tortuosity. The developed system is tested on a private dataset consisting of 35 images. The result obtained shows that the reported system achieves a sensitivity and specificity of 0.88 and 0.94 respectively.