Optimized KFCM Segmentation and RNN Based Classification System for Diabetic Retinopathy Detection

Abstract

Finding and diagnosis of various eye illnesses for ophthalmologist to helpful by using Human retinal image. Automated blood vessel segmentation diagnoses numerous eye infections like diabetic retinopathy, retinopathy of prematurity or glaucoma. In this work, we propose the Optimized Kernel-based Fuzzy C-Means (OKFCM) Segmentation and Recurrent Neural Network (RNN) based Classification system for Diabetic Retinopathy detection. In the proposed segmentation section consist of two main stages such as optic disc removal and Modified Ant Colony Optimization (ACO) based KFCM Segmentation. For the Diabetic Retinopathy classification, GLCM and moment built features are used. The proposed system is also named as an OKFCM-MACO-RNN. The OKFCM-MACO-RNN classification assessment process is complete on the diaretDB1 dataset by manipulative the value of features like accuracy, sensitivity, and specificity of the OKFCM-MACO-RNN method respectively 99.33, 81.65 and 99.42%. The OKFCM-MACO-RNN method is predictable to be able to notice exudates well. The OKFCM-MACO-RNN Segmentation performance is analyzed in terms of jacquard coefficient, dice coefficient and accuracy respectively 85.65, 72.84 and 93.15.