An Efficient Ridge Detection Method for Retinopathy of Prematurity Severity Analysis

Retinopathy of Prematurity (ROP) is a blinding disease affecting the retina of low birth weight preterm infants. Accurate diagnosis is essential to identify the severity and stage of ROP to prevent childhood blindness. Progression of ROP into various stages is identified through the properties of ridge present in the anterior pole of retina. Ridge detection is very challenging in infant fundus images due to their poor contrast and other artefacts. This work proposes an efficient segmentation algorithm to detect the ridge in the infant fundus images. An image dataset with 140 retinal fundus images of preterm infants was used for this work. The proposed algorithm gave a detection accuracy of 95.71 percent and sensitivity and specificity of 91.89 percent and 100 percent respectively, when tested on real images having wide variations in quality.