An Optic Disc Detection Scheme on Retinal Images

For a computer aided detection system of retinal lesions, optic disc (OD) detection is the first and the most importance. Since the OD is once detected then other clinical importance regions such as fovea or macula can easily be determined, and the OD shape is often used to evaluate abnormal retinal features for diagnosis of abnormalities such as hypertensive retinopathy. In this paper, an approximate OD region is detected by an upper outlier detection scheme. And the final OD is extracted by the random walk scheme with seeds selected automatically. The presented OD detection scheme is assessed quantitatively in 89 images of the Standard Diabetic Retinopathy Database Calibration level 1 (DIARETDB1) database by comparing the detection results with ophthalmologist’s hand-drawn ground-truth images. The experimental results show that the OD detected by the presented algorithm approximately follows that extracted by an expert radiologist. Keywords— Optic disc, retinal, outlier, random walk, DIARETDB1.