DETECTION AND GRADING OF DIABETIC RETINOPATHY IN RETINAL FUNDUS IMAGE USING SPIKING NEURAL NETWORK

DETECTION AND GRADING OF DIABETIC RETINOPATHY IN RETINAL FUNDUS IMAGE USING SPIKING NEURAL NETWORK

Neethu Krishna K A1, J. Mahil2

1P.G Scholar, Dept. of Biomedical Engineering, Udaya School of Engineering, Vellamodi, Tamil Nadu, India

2Professor, Dept. of Biomedical Engineering, Udaya School of Engineering, Vellamodi, Tamil Nadu, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Diabetic retinopathy (DR) is one of the world's most significant difficulties of diabetes identified with eye illness which happens when veins in the retina become swollen and release liquid which at last prompts vision misfortune. Early discovery of DR can anticipate the harm to the retina and vision misfortune or at least moderate its movement. The main objective of the proposed work is to detect the severity of Diabetic Retinopathy (DR) in the retinal fundus image using the spiking neural network. In this process, the retinal fundus images are collected from the STARE database. The data are pre-processed using the Wiener filter to extract the green channel and remove the noise. Then the Histogram equalization technique is used to enhance the contrast of the fundus image. DR is analyzed by separating blood vessels, veins, optic plates, and exudates from retinal fundus images by using mathematical morphology and finally a Prediction Spiking Neural Network(SNN) classifier and RG-based segmentation approach is implemented to segment the normal and DR from the fundus image.  The performance of this segmentation approach is executed for different DR datasets and the outcomes demonstrate that; the proposed segmentation approach produces 96% accuracy for the detection of DR when stood out from different existing strategies.

Key Words:  Diabetic retinopathy, retinal fundus image, Histogram equalization, Spiking Neural Network, RG-based segmentation.