Gradient Ridge Enhancer: A Robust Approach for Ridge Detection in Infant’ Retinal Images

Retinopathy of Prematurity is a blindness disease found in premature infants having a weight of about 1250 gm and born before 31 weeks of gestation, causing abnormal growth of blood vessels leading to retinal detachment which results in permanent blindness. The ROP consists of various stages, the first stage consists of a demarcation line that separates the vascularized retina from the peripheral retina. In the second stage, the demarcation line increases in height and width known as a ridge. Identification of demarcation line or ridge in the retinal images of premature infants is an intricate task because of the low-contrast retina images. This research work proposes a novel approach where we extract the intensity channel from HSI Color space, applying gamma correction for contrast enhancement and median filter for noise reduction. For further enhancement of the ridge, we propose a Gradient Ridge Enhancer (GRE) filter which improves the visibility of the ridge structure by combining the Sobel magnitude filter with median filtered image. Contrast enhancement and color correction are done using CLAHE and gamma correction to handle the low contrast and uneven illumination in retinal images. Lastly, image segmentation is performed using the OTSU method, followed by a post-processing algorithm to obtain the ridge-segmented mask. The proposed method achieved an accuracy of 0.9854 for ridge detection when tested on a dataset of 1103 retina images from premature infants with varying image quality, demonstrating the potential of the proposed GRE filter for efficient diagnosis of ROP disease.