Purchase this article with an account.
P. Gupta, G. Quellec, K. Lee, M. Dolejsi, M. K. Garvin, M. Sonka, M. D. Abramoff, Sr.; 3-D Textural Analysis of Intraretinal Layers in Macula-Centered SD-OCT Volumes From Neovascular AMD Patients. Invest. Ophthalmol. Vis. Sci. 2010;51(13):1785.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine the diagnostic potential of texture and thickness properties of intraretinal layers outside of fluid-filled regions by Spectral-Domain Optical Coherence Tomography (SD-OCT) volumes in neovascular Age-Related Macular Degeneration (AMD).
91 macula-centered SD-OCT volumes (200x200x1024 voxels, 6x6x2 mm3) were obtained from 13 normal (13 scans) and 26 neovascular AMD eyes (78 scans) using Cirrus HD-OCT machines (Carl Zeiss Meditec, Inc., Dublin, CA). 10 intraretinal layers were automatically segmented in each volume using a 3-D multi-layer graph search-based approach. 23 features were extracted locally in each layer, within vertical columns of 0.3x0.3 mm2 square base: the average and standard deviation of layer thickness, intensity level distribution features, run length features, co-occurrence matrix features and wavelet analysis features. The distribution of these features across the 9 AREDS macular subfields (center circle, inner/outer nasal, temporal, superior and inferior subfields), were compared in each layer using a Kolmogoroff-Smirnov (K-S) test. The footprint of fluid-filled regions was manually annotated in AMD eyes, then removed from the analysis.
A significant difference was observed between normal and AMD in each subfield and in most layers for 22 of these 23 features; no difference was observed for the kurtosis of low-frequency wavelet coefficients. An average K-S statistic (D) greater than 0.4 was measured for 4 features across all layers and subfields: the average intensity (D=0.507), the standard deviation of the highest frequency wavelet coefficients (D=0.447), as well as the 2 co-occurrence analysis features of inertia (D=0.441) and correlation (D=0.424). The thickness of the layers ranked amongst the least discriminant features (average: D=0.318, standard deviation: D=0.263). Across all features, the largest differences between AMD and normal were detected in the 4 inner subfields for the connecting cilia, inner plexiform, and outer segment layers.
3-D texture analysis of intraretinal layers in macula-centered SD-OCT volumes provides clinically useful metrics, and may be more discriminant than the parameter that is easiest to assess clinically: layer thickness. Texture analysis has the potential to improve management of patients with neovascular AMD undergoing treat-and-extend anti-VEGF injections.
This PDF is available to Subscribers Only