Computer based Technique to Examine Diabetic Retinopathy in Fundus Retinal Images

• Published in: 2020 5th International Conference on Computing, Communication and Security (ICCCS) pp. 1 - 8
• Main Authors: R, Remya K, Giriprasad, M N
• Format: Conference Proceeding
• Language: English
• Published: IEEE 14-10-2020
• Subjects: Biomedical imaging; Blood-vessel extraction; Classification; Data mining; Diabetes; Diabetic retinopathy; Feature extraction; Performance validation; Retina; Retinal images; Retinopathy; Sensitivity
• DOI: 10.1109/ICCCS49678.2020.9277279

Eye is a necessary organ in the human sensory system which is responsible for receiving the light based information. Untreated eye diseases will lead to vision loss. Diabetic Retinopathy (DR) is one of the age related eye disease which arise due to diabetes and the early detection and treatment will help in regulating the DR. The proposed work aims to develop a system for automated detection of the DR with the help of Fundus Retinal (FR) images. This work aims to develop and implement a simple and straight forward technique to detect the DR class images with greater accuracy. This work implements a set of procedures, such as pre-processing (blood-vessel mining), feature extraction, feature selection and classification. The proposed work is tested using the benchmark FRI databases such as; Diaretdb0 and Diaretdb1. During the examination, only the Green-Channel-Image (GCI) is considered and the proposed method offered a classification accuracy greater than 96% on the considered image datasets with the Support-Vector-Machine (SVM-linear) classifier. The experimental result of the proposed work on 150 numbers of RGB class FR images confirms that the proposed technique with the SVM-L classifier helps to attain a better result compared to DT, NB and KNN classifiers.