EARLY DETECTION AND CLASSIFICATION OF DIABETIC RETINOPATHY USING EMPIRICAL TRANSFORMS AND SVM by Sumandeep Kaur
Abstract
Diabetic Retinopathy is the name given to „disease of retina‟. This is the disease in which the blood vessels in light-sensitive tissue, known as retina, are affected due to diabetes. It comprises of various impacts on eyes such as swelling, leakage and blockage of the blood vessels in an eye. During the initial stages, patients remain unaware of this disease due to lack of easily detectable symptoms in them. But at the later stages, diabetic retinopathy may lead to visual impairment or even blindness.The objective of this work is for timely diagnosis and classification of diabetic retinopathy using curvelet transform and support vector machine is developed. Diabetic retinopathy is detected by discovering haemorrhages and exudates in the fundus images. Firstly, retinal images are enhanced using empirical transform. Canny edge detection is applies for extracting eyeball from retinal fundus image. Then morphological operations are applied for locating the imperfections in the images. At the end, images are classified into normal, proliferative or nonproliferative by using support vector machine. Both accuracy and sensitivity of the images is improved when compared with previous technique in which only k-means and fuzzy classifier is used. So, this approach could be employed for medical purposes. Hence, enhancement of images using curvelets helps to improve the accuracy and sensitivity of detection and classification. The number of exudates detected in present work is more than that of the process without enhancement. Also, the accuracy of detection system depends upon the peak signal to noise ratio in the image. If the value of the peak signal to noise ratio is less, number of exudates detected by noisy image are far less than when detected with the curvelet denoised images. The sensitivity, specificity and accuracy of system are calculated as 96.77%, 100% and 97.78% respectively.
