Automated layer segmentation of macular OCT images via graph-based SLIC superpixels and manifold ranking approach

Intra-retinal Layer Segmentation of Macular OCT Images using Edge Superpixels and Manifold Ranking Method

“…By testing a normal macular 3D OCT dataset and a macular 3D OCT dataset with CSR, We evaluated the proposed algorithm on main boundaries error. The experiments demonstrated promising results with comparisons to the manual tracings of two independent observers and the method of the [11,17]. The proposed algorithm can not only overcome the organic texture artifacts and speckle noise, but also be computationally efficient, and be not relatively susceptible to the low contrast of macular fovea and the bad structure of the CSR.…”

“…For the full ten layers segmentation of each 2D slice, the proposed algorithm took about 5.28s much faster than 9.02s [11] and 9.6s [17] in dataset1, and the proposed algorithm also took about 2.9s much faster in dataset2. For the mean signed and unsigned border positioning differences of the two datasets, Tabels 1, 2, 3 and 4 respectively summarize the main boundaries as follows.…”

“…The proposed algorithm mainly compared with the two independent observers, the SLIC superpixels method [17], and the Iowa's algorithm [11] by the mean signed and unsigned border positioning differences of the ILM, NFL/GCL, IPL/INL, INL/OPL, OPL/ONL, ELM, IS/CL and BM/Choroid boundaries, since the two independent observers did not attempt to trace the GCL/IPL, CL/OS and OS/BM boundaries that are considered invisible. In addition, for the purpose of the clinical and medical analysis, the thickness map of each layer was respectively plotted by the proposed algorithm in the dataset 2.…”

“…In comparision with the single methods and SLIC superpixels method [17], contributions in this paper mainly include as follows: 1) A simple adaptive filter was designed, and could well replace the superpixel and the Dijkstra's algorithm to effectively remove some disturbances from vitreous artifacts.…”

Nondestructive Early Detection of Bruising in Pear Fruit Using Optical Coherence Tomography

“…This bright, thick line represented the thin peel surface layer of the pear fruit, which does not contain any bruise information. Using the graphsegment approach (Gao et al, 2017), another boundary between the peel and flesh tissue was found. Taking the pixels on the tissue boundary as a starting point, the intensity/gray-scale profile of each A-scan was normalized, and the average normalized A-scans taken from the ROI in each OCT B-scan were used for the subsequent quantitative analysis.…”