Indian Journal of Science and Technology

Abstract

Background/Objective: Exudates are the significant portions for the detection of Diabetic Retinopathy (DR). This paper demonstrates a complete framework for the detection of hard exudates in retinopathy images. Methods/Statistical Analysis: This paper presents two variants of Multiple Kernel induced Gaussian Spatial Fuzzy-C Means (MKGSFCM) algorithm for the segmentation of retinal fundus images. The algorithm is applied on different DR images and the performance of the algorithm is evaluated qualitatively and quantitatively. Findings: FCM and KFCM algorithms are commonly used clustering methods but are very sensitive to noise and other imaging artefacts. This paper presents a hybrid version of KFCM with induced Gaussian spatial information. Sensitivity and specificity values of the proposed work are observed to be high and also the possibility of exudate misclassification is significantly reduced by the proposed method as compared to existing algorithms. Improvement: The frame work presented can be developed further by the inclusion of adaptive weights for the multiple kernels.
Keywords: Diabetic Retinopathy, Exudates, FCM Clustering Algorithm, Multiple Kernel Induced Gaussian Spatial FCM Algorithm.