Semi-automated Clinical Staging of Retinopathy of Prematurity Images

In this paper we propose feature extraction and classification of Retinopathy of prematurity (ROP) images with semi-automated approach. Retinopathy of prematurity is a condition present in premature neonates affecting the vision which can lead to blindness. The early detection can help to prevent loss of vision in new born. Diagnosis of ROP requires continuous monitoring of fundus images in the first four weeks. Because of advanced science the survival of premature neonates is on rise which leads to more number of high risk neonates for ROP. The fundus image analysis by human readers may lack accuracy because of excessive workload as the number of qualified doctors is less compared to the number of patients. The proposed semi-automation method is based on thresholding technique. As the images vary in size & illumination preprocessing of data is done. Based on thresholding technique, we can find the different stages of ROP. High accuracy is achieved to find the different stages.