Optic Disc (OD) segmentation is a key preprocessing component in many algorithms designed for the automatic extraction of retinal anatomical structures and lesions, thus, an associated module of most retinopathy screening systems. Most of the OD segmentation algorithms processes exclusively one colour plane of the retinography, thus discarding colour information. In this paper, a new algorithm for the segmentation of the OD is proposed. After a pre-processing step to remove vessels, an edgebased level-set algorithm is applied. The main contribution of the paper consists in the utilization of the colour information during the segmentation. Specifically, vector gradients in L*a*b* are applied in the edge-based level-set algorithm and, instead of the Euclidean norm, CIE94 colour difference formula is applied to those vector gradients. In addition, in the vessel removal step, the three-plane image is modified in L*a*b* colour space. The proposed algorithm was tested with 22 retinographies where physicians had manually detected OD edges. The algorithm was able to automatically detect the OD in all the cases, with a 92,35% of intersection between the OD area marked by physicians and the OD detected with the algorithm. The Mean Distance to the Closest Point was also calculated and it was under 5 pixels for 100% of the images.
Aurora Sáez, Irene Fondón, Begoña Acha, Soledad Jiménez, Pedro Alemany, Qaisar Abbas, Carmen Serrano, "Optic Disc segmentation based on level-set and colour gradients" in Proc. IS&T CGIV 2012 6th European Conf. on Colour in Graphics, Imaging, and Vision, 2012, pp 121 - 125, https://doi.org/10.2352/CGIV.2012.6.1.art00022