Thickness measurements of ten intra-retinal layers from optical coherent tomography images using a super-pixels and manifold ranking approach

Abstract

The purposes of this paper are to calculate exactly the mean thickness, plot the thickness maps, and depict the early treatment diabetic retinopathy study (ETDRS) charts for ten intra-retinal layers by spectral domain optical coherence tomography(OCT). Using our previously the reported segmented method with a simple linear iterative clustering (SLIC) super-pixels and manifold ranking (SLIC_MR), the ten intra-retinal layers were fast and exactly segmented in 3-D OCT dataset, includes 55 B-scan images from 11 different healthy adult subjects. By our definitions of the sensitivity and specificity, we compared the segmented results with the recent graph-based method for the main layers in dataset. The experimental results demonstrated that the SLIC_MR method outperformed the graph-based method. The thickness maps were plotted in the ten intra-retinal layers and the overall layer, the ETDRS charts were depicted in the 9 sectors of each intra-retinal layer and the overall layer, and the bar graph displayed the mean and standard deviation of macular thickness in 9 sectors for ten retinal layers and the overall layer. The mean thickness of the central foveal area displayed the minimum thickness in layers 1, 2, 3, 4, 5 and the overall, and the maximum thickness in the central foveal area of the 6th layer. Both layers 4 and 5 have the similar mean thickness in each sector.