Intra-retinal layers segmentation of macular OCT images based on the graph optimal approach

Based on the graph optimal technology, a novel automated intra-retinal layers segmentation method is proposed in this paper. Eleven boundaries of ten retinal layers in optical coherence tomography (OCT) images are exactly, fast and reliably quantified. Instead of considering the intensity or gradient features of the single-pixel, the proposed method focuses on the whole edge-based image cues. The image is represented as a complete weighted graph with connected components as nodes. Each node is ranked based on the gradient and spatial distance information of the connected components, and the affinity matrices. Segmentation is efficiently implemented in a three-stage project to extract eleven boundaries. The segmentation algorithm is evaluated on 64 OCT images from two different databases, and is compared with the manual tracings of two independent observers. It demonstrates encouraging results in term of the mean unsigned boundaries errors and the mean signed boundaries errors.