Detection of Microaneurysm using Machine Learning Techniques

Diabetic Retinopathy (DR) is a human eye disease which affects people with diabetics. DR affects retina of the eye and leads to blindness when it moves on to severe level. The early signs of diabetic retinopathy can be detected by screening of fundus image which is an effective method to prevent eye diseases. Detecting the disease at an earlier stage can prevent the patient from vision loss. This work aims to explore automatic methods for diabetic retinopathy (Microaneurysm) detection and eventually develop a successful system for detection and classification of diabetic retinopathy. In this work the proposed system consists of pre-processing, segmentation, feature extraction and classification of lesions. The retinal fundus images are taken from DIAbetic RETinopathy Database-calibration Level-1 (DIARETDB1). In pre-processing, the background pixels of the images are eliminated, resized, unwanted noise in the images are removed using average filter and median filter and then contrast enhancement is done using adaptive histogram equalization algorithm. The blood vessels of the image are detected using morphological operation (erosion and dilation). In segmentation phase the images are partitioned using k-means clustering algorithm for efficient image analysis. Statistical analysis is used for extracting the feature of the fundus image. The fundus image can be then classified with the help of support vector machine (SVM) classifier. The experiment results specificity, sensitivity, F-measure and accuracy are calculated based on the parameter of the proposed system. Based on the experiment result the accuracy achieved by SVM classifier is 95%.