A New and Efficient Method for Automatic Optic Disc Detection Using Geometrical Features

Automatic optic disc (OD) detection is a key processing step in algorithms aiming for automatic extraction of retinal vascular abnormalities for diagnosing cardiovascular diseases and cup-to-disc ratio assessment for diagnosing glaucoma. In this paper, we propose a new method for automatic OD segmentation and center computation. The proposed method automatically determines the threshold intensity value by utilizing OD area information in each image. The Region Growing Technique is applied in the thresholded image to find the potential OD regions. Following this, we produce the image gradient and apply the Hough circle detection algorithm in each of the potential OD regions to identify the OD and compute its center. For evaluation, we apply our method on STARE and DRIVE dataset and Singapore Malay Eye Study dataset, and demonstrate that our method is highly accurate and efficient.