Optic Disc Localization in Complicated Environment of Retinal Image Using Circular-Like Estimation

Abstract

Optic disc (OD) localization is an important prerequisite task for developing systems for automated diagnosis of various serious ophthalmic pathologies. New hybrid algorithms for disc localization are presented in this research. For classical disc localization, the data sources are complete and typically have normal ocular structure. Our investigation concerns the conditions in some complex environments such as when unbalanced shade in an image occurs, or when a low contrast OD disturbed shape OD exists, and in the case where various bright lesions are found. In this paper, fully automatic disc localization in retinal fundus images within various complicated environments is proposed. This technique is designed for identifying disc candidate locations in the case of unclear disc boundaries and when unbalanced shading may be present. A combination of modified smoothed gradient edge generation and a Hough transform is deployed. All candidate results are combined to locate the final location using a simple voting method. The proposed approach was evaluated on seven publicly available datasets: DIARETDB0, DIARETDB1, MESSIDOR, ROC, E-OPHTHA, HRF, and ARIA databases. The accuracy of the proposed method is 99.23%, 100.00%, 98.58%, 99.00%, 96.34%, 100.00%, and 91.61%, respectively.