Purchase this article with an account.
Mukhit Kulmaganbetov, Nantheera Anantrasirichai, Alin Achim, Julie Albon, Nick White, James Edwards Morgan; Texture analysis of OCT phantoms. Invest. Ophthalmol. Vis. Sci. 2020;61(7):2038.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Texture analysis has shown value in optical coherence tomography (OCT) imaging for the detection of early degenerative retinal changes. The calibration of texture-based techniques requires gold standard targets. We, therefore, evaluated a range of retinal imaging phantoms to evaluate techniques for texture-based analysis with OCT.
Four groups of phantom types were prepared and imaged (Table 1): different concentrations of a medium matrix (gelatin solution); different sized polystyrene beads (PBs); different volume of PBs and different refractive indices of scatterers (PBs and SiO2). The phantom particles were selected to lie in the size range of subcellular organelles that would be expected to change during retinal neuronal degeneration. OCT at 1040nm (bandwidth 70nm) was performed to acquire 3D image datasets (512×512×1024 pixels) of each phantom, using a custom research OCT with a microscope attachment. Five grey-level co-occurrence matrix features (energy, contrast, entropy, correlation and homogeneity) in four directions (0o, 90o, 180o, 270o) were extracted from each random volumes of interest (VOI, n=10, 30×30×30 pixels).
Textural parameters were analysed in 20-dimensional feature space using principal component analysis (PCA) and support vector machine (SVM) written in Matlab. PCA was used to reduce the feature dimensions by eliminating redundant data, using SVM served as a classifier against the known state of the phantom. The classification approaches were tested on the 160 OCT images of our four groups of 16 types of phantoms (Table 2).
We presented a semi-automated method for the classification of OCT images. Even though the OCT scans of the various phantoms are optically similar for unaided eye, they can be discriminated using OCT-based texture analysis.
This is a 2020 ARVO Annual Meeting abstract.
Table 1. Groups and types of phantoms
Table 2. Classification accuracy of different phantom groups
This PDF is available to Subscribers Only