Fast Optic Disc Segmentation In Fundus Images

The identification of the optic disc (OD) in fudus images is a useful landmark for a large number of ophthalmic diseases. The method proposed herein detects and segments the OD in a rather simple and fast manner, compared with those methods recently published. The procedure involves first: segment the blood vessel, then locate the OD using the highest intensity values found near the vessels. Finally, OD segmentation is performed using a modification of the elliptical Hough transform. The results are evaluated using five different metrics already used in the literature on the subject, and applied to a public database of 1200 fundus images, obtaining 89% in the Dice coefficient and 82% in Jaccard coefficient, and taking a processing time of 4.1 seconds per image. Compared with other published methodologies, the results presented with the current technique, improve speed performance while maintaining similar precision metrics.