Active discs for automated optic disc segmentation

Abstract

Glaucoma is currently one of the major causes of vision loss worldwide. Though the lost vision capability cannot be recovered, preventive steps such as clinical diagnosis and necessary treatment can be taken in order to minimize visual impairment due to glaucoma. The process of clinical diagnosis needs manual examination and outlining of the optic disc and cup, which is subjective and time consuming. In this paper, we propose a methodology for automatic segmentation and outlining of the optic disc using an active disc formulation. The method uses a disc template as a prototype. The initialization is automated using a matched filtering technique. We choose the disc to be isotropic, and allow for translation, isotropic scaling, and optimization of the corresponding parameters. The active disc is evolved towards the boundary of the optic disc by minimizing a local energy function. We report validations on three publicly available databases MES-SIDOR, DRIONS-DB, and Drishti-GS containing 1200, 110, and 101 retinal fundus images, respectively. The corresponding F-scores 0.8456, 0.8380, and 0.9077, respectively, for the databases show robustness, and high correlation of the proposed algorithm with the expert segmentation.