Localization of the Optic Disc in Retinal Fundus Image using Appearance Based Method andVasculature Convergence

  • Suha Dh. Athab Department of Computer Science, University of Baghdad, Baghdad, Iraq
  • Nassir H. Selman Department of Computer Science, University of Baghdad, Baghdad, Iraq
Keywords: Optic disc, vasculature convergence, intensity thresholding

Abstract

Optic Disc (OD) localization is a basic step for the screening, identification and appreciation of the risk of diverse ophthalmic pathologies such as glaucoma and diabetic retinopathy.In fact, the fundamental step towards an exact OD segmentation process is the success of OD localization. This paper proposes a fully automatic procedure for OD localization based on two of the OD most relevant features  of high-intensity value and vasculature convergence. Merging ofthese two features renders the proposed method capable of localizing the OD within the variously complicated environments such as the faint disc boundary, unbalanced shading, and the existence of retinal pathologies like cotton wall and exudates,which usually share the same color and structure with the OD. To demonstrate the robustness, reliability and broad applicability of the proposed approach,we tested 1614 images from publically available datasets, including Messidor (1200 images), TheStandard Diabetic,Retinopathy Database (DIARETDB0 ,130 images), Digital Retinal,Images for Optic Nerve,Segmentation (DRIONS ,110 images), TheStandard Diabetic,Retinopathy Database (DIARETDB1,89 images),High,Resolution Fundus (HRF,45 images),and Digital,Retinal Image for Vessels,Extraction (DRIVE,40 images). The method successfully localized 1599 images and failed in 15 images, with an average success rate of 99.07% and an average computation time of 0.5 second per image.

Published
2020-01-27
How to Cite
Athab, S. D., & Selman, N. H. (2020). Localization of the Optic Disc in Retinal Fundus Image using Appearance Based Method andVasculature Convergence. Iraqi Journal of Science, 61(1), 164-170. https://doi.org/10.24996/ijs.2020.61.1.18
Section
Computer Science