Diagnosis of Diabetic Retinopathy Using Wavelet Based Contourlet Transform and k-NN Algorithm

Abstract

Diabetic retinopathy leads to vision loss. Moreover, it damages a petite blood vessel in the retina; however, it is due to poorly blood glucose levels in a diabetic patient. As diabetes is most frequently occurring disease due to sedimentary life style of people. It is very difficult to diagnose diabetic retinopathy without a retinal specialist. But the services of this specialist are available in very few cities and hence most of the population in India and abroad is prevented from early diagnosis of such disguised disease. Hence, a system is develop for diagnosis of diabetic retinopathy which helps a general physician. According to recent trends sharp rise in number of diabetic patient due to lifestyle. The use of contours (smooth shapes in retinal fundus image) and k-NN classification technique is one of the solution to the problem.A system has been developed for early diagnosis of diabetic retinopathy. The basic wavelet transform lags in reconstructing curved images perfectly and hence WBCT-Wavelet Based Contourlet Transform with pre-processing technique is used to detect abnormal blood vessel growth and other pathologies as a state-of-art image processing algorithm. The algorithm developed estimates the contour of query image and then image is classified using linear classification technique in concern class of diabetic retinopathy. Images from publically available database annotated by ophthalmologist particularly retinal specialist are used to train the algorithm.