A robust algorithm for optic disc segmentation and fovea detection in retinal fundus images

Abstract

Accurate optic disc (OD) segmentation and fovea detection in retinal fundus images are crucial for diagnosis in ophthalmology. We propose a robust and broadly applicable algorithm for automated, robust, reliable and consistent fovea detection based on OD segmentation. The OD segmentation is performed with morphological operations and Fuzzy C Means Clustering combined with iterative thresholding on a foreground segmentation. The fovea detection is based on a vessel segmentation via morphological operations and uses the resulting OD segmentation to determine multiple regions of interest. The fovea is determined from the largest, vesselfree candidate region. We have tested the novel method on a total of 190 images from three publicly available databases DRIONS, Drive and HRF. Compared to results of two human experts for DRIONS database, our OD segmentation yielded a dice coefficient of 0.83. Note that missing ground truth and expert variability is an issue. The new scheme achieved an overall success rate of 99.44% for OD detection and an overall success rate of 96.25% for fovea detection, which is superior to state-of-the-art approaches.

An innovate approach for retinal blood vessel segmentation using mixture of supervised and unsupervised methods

“…For that purpose, we first compute the centres of the optic disc and macula relying on the method proposed by Rust et al. in [24]. Let, and are the coordinates of the optic disk and macula center respective, then the reproducible orientation of the image is computed as, …”

“…Prior to computing Haar wavelet responses, we identify a reproducible orientation of the image, which is then used to rotate the image. For that purpose, we first compute the centres of the optic disc and macula relying on the method proposed by Rust et al in [24]. Let, (x M , y M ) and (x OD , y OD ) are the coordinates of the optic disk and macula center respective, then the reproducible orientation of the image 𝜃 is computed as,…”