Retinal Microaneurysm Detection and Post Rocessing for True Vessel Extraction

Diabetic Retinopathy (DR) is the leading ophthalmic pathological reason of blindness among people of working age in developed countries. The main cause of DR is abnormal blood glucose level rise, which damages vessel endothelium, causing an increase in the vessel permeability. The initial manifestations of DR are tiny capillary dilations known as microaneurysms. A method for automatic detection for Microaneurysm and post processing for true vessel extraction in the retinal image is implemented. The proposed system uses the images obtained from diaretdb0 and diaretdb1 database. The preprocessed images are converted to gray and green channel to which the Canny edge detection operator is implemented for the retinal image segmentation based on a threshold value. Further, from the segmented image the combined features for the microaneurysm and true vessels are extracted by using Receiver Operating Characteristics. Finally, the extracted features such as area of blood vessel, area of microaneurysm, entropy and homogeneity are given to the classifiers such as BPNN, KNN and NB classifiers. Thus a combined approach for microaneurysm detection and post processing for true vessel extraction has been implemented and the accurate results are obtained. The proposed system is a novel approach which overcomes all the disadvantages of the existing system. The performance analysis of three different classifiers such as KNN, NB and BPNN has been analyzed and their comparative results proved that KNN classifier outperforms other two classifiers. This kind of systemic processes may be a new perspective in detection, classification and identification of retinal blood vessels and microaneurysm detection. The experimented effectiveness and robustness, together with its simplicity and fast implementation, make this proposed automated blood vessel segmentation and microaneurysm detection method a suitable tool for being integrated into a complete prescreening system for early detection of DR.