Multiple Feature Sets and SVM Classifier for the Detection of Diabetic Retinopathy Using Retinal Images

Diabetes Mellitus is one of the growing vitally fatal diseases that can affect the patient's sight and its most severe effect is on blood vessels inside the eye called diabetic retinopathy. Due to its significance, a design of an efficient classifier for the detection of Diabetes disease is one of the challenging tasks. In this paper, we have proposed SVM classifier for diagnosing the diabetics from retinal images using two features like optic disc and blood vessel. Initially, the Gaussian filter is used for performing the pre-processing phase. Once the noise free image is generated, the segmentation processed is applied for detecting the both optic disc and blood vessel areas. Then, the relevant features are extracted from the optic disc and blood vessel such as mean, variance, perimeter, diameter, maximum intensity and minimum intensity. Then, the diabetic images are classified from the input images using the proposed SVM classifier. Finally, the experimentation results of the proposed classification technique is carried out using the Stare database, which shows that the SVM classifier can be successfully classifies the diabetic images with better classification accuracy of 96%.