Image Preprocessing And Analysis On Eye Fundus Images Segmentation By Using Density Clustering Methods

Abstract

Abstract: In order to do an automated evaluation of various retinal illnesses such as Diabetic retinopathy, Glaucoma, and Macular Edema, fundus images must be pre-processed first. For many reasons, it's difficult to accurately detect the optic disc. Many blood vessels cross the optic disc, making it difficult to discern the disc's boundaries in fundus images. Lesion regions in diabetic retinopathy look very much like an optic disc's colour and texture, so an automated retinal image analysis system must identify and remove these areas. DR is diagnosed early in this study using machine learning (ML) approaches. For example: Bayesian Classification; K-Means Clustering; PNN; SVM; and Bayesian Classification In order to determine the most effective strategy, these options will be weighed against one another and evaluated. For training and testing, a total of 300 fundus images are processed. Using image processing techniques, these raw photos are processed to extract the features. The results of an experiment show that PNN, Bayes Classifications, SVM, and K-Means Clustering are all more accurate than 94% of the time. It appears that SVM is the best method for detecting early signs of degenerative disease.