A Fully Automated 3D In-Vivo Delineation and Shape Parameterization of the Human Lamina Cribrosa in Optical Coherence Tomography.

Abstract

A fully automated method for delineation of the lamina cribrosa in optical coherence tomography (OCT) is proposed. It assesses the three-dimensional (3D) shape of the lamina cribrosa in-vivo, based on a series of OCT B-scans. The algorithm has several image processing steps and it is based on active contour detection performed along three orthogonal directions of the B-scan data cuboid. Further, the delineated 3D lamina cribrosa shape is parameterized with a fourth-order polynomial of two variables [Formula: see text] using the least-squares method. Datasets from a total of 255 subjects from three groups were analyzed: 92 primary open angle glaucoma patients, 77 glaucoma suspects, and 86 controls. Statistically significant differences were found in the coefficients of monomials xⁱy^j, with both i and j even, between patients and controls and between suspects and controls. From the data obtained, it can be concluded that the mean shape parameterization of the lamina cribrosa of glaucoma suspects has similar appearance to that of glaucoma patients but it is markedly different from that of healthy controls. The proposed algorithm enables automatically estimating, for the first time, the lamina cribrosa in 3D, further providing clinicians with a time-efficient discrimination tool supporting glaucoma diagnosis.