An Automatic Localization of Microaneurysms in Retinal Fundus Images

The retinal Microaneurysms (MA) are the first earlier and clinical sign of Diabetic Retinopathy (DR). It is one of the major symptom of DR that leads to blindness if left untreated. So it is necessary for regular screening of the eye. The ocular fundus image contains all the elements such as optic disk, blood vessels, exudates which are the distractors to detect MA accurately. The MA are small saccular pouches which are caused by local distensions of the capillary walls. They are circular in shape and occur near blood vessels. So it is being a challenge for the optometrists as well as the image processing scholars to detect it. Here, an automated system is generated to localize the MA among the several input retinal images. Our method starts with the pre-processing step which does the intensity variations adjustment of the image using adaptive histogram equalization. Then it is proceeded with MA segmentation process and thus it removes all the distractors, finally localize the MA. The proposed method was implemented in MATLAB and tested using public available datasets such as DRIVE, STARE and DIARETDB0. The performance of MA localization has analyzed using area under receiver operating characteristics.