Efficient Diabetic Retinopathy Detection using Machine Learning Techniques

Abstract

The medical technology has seen a tremendous growth in this century. Innovative high- end technologies that are created for health care benefits the patients as well as the medical professional in a wider perspective. Diabetes mellitus is a medical complaint among all age groups which occurs due to the increase in the blood sugar level. Diabetic retinopathy is said to be a symptomless diabetic eye illness which affects the retina of human eye and leads to blindness. It affects the retinal blood vessels. There is a growth of abnormal blood vessels in the retinal surface. Diabetic retinopathy can be detected using Ridge based vessel segmentation, Computer Driven Tracing of Vessel Network, Adaptive Local Thresholding it does not have uniform illuminations. Latest technological advancements in image processing provide a more efficient diagnosis of diabetic retinopathy with the help of feature extraction. The retinal scanned image is first pre-processed and feature extraction is done using HAAR wavelet Transform for the quantitative measure of the accuracy of the disease. The image is segmented and classified based on the training sets of data using SVM classifier. This process tends to provides more accuracy and about 98% sensitivity in+ the retinal classification.