Detection of Non-proliferative Diabetic Retinopathy using GUI

We looked at a strategy for identifying and categorizing exudates in colored retinal images in this study. Filter banks are used in an innovative approach for extracting possible exudate candidate sites. The bogus exudate areas are removed when the optic disc region is deleted. A Bayesian classifier, which is a set of Gaussian functions, is then used to distinguish between exudate and nonexudate areas. Using publicly available retinal image datasets and performance criteria, the proposed system is reviewed and tested. We compare the proposed system to previously presented techniques in order to establish its validity. The exudate identification algorithm was used on 30 retinal photographs, 21 of which had exudates and nine of which did not. When compared to an ophthalmologist, the sensitivity and specificity for exudate identification were 88.5 percent and 99.7%, respectively. In 14 retinal pictures, HMA (Haemorrhages and Microaneurysms) was found. For HMA identification, the algorithm has a sensitivity of 77.5 percent and a specificity of 88.7%. This research offers promising results in the automatic detection of key NPDR (Non-proliferative Diabetic Retinopathy) traits.