Purchase this article with an account.
Ted Maddess, Marconi Barbosa, Samyoul Ahn, Tailoi Chan-Ling; Novel morphometric analysis of higher order structure of human radial peri-papillary capillaries. Invest. Ophthalmol. Vis. Sci. 2019;60(9):5617.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We apply novel analyses to images of the radial peri-papillary capillaries (RPCs). The RPCs are significant for eye disease due to their unique perfusion of the peri-papillary nerve fibre layer. We examined which new morphometric features of the RPCs are related to the age of human retinas.
The inputs to the analysis were z-stacks from 3D confocal fluorescence microscopy from 62 human retinas aged 9 to 84 years. The retinas had no ophthalmic history. The analysis comprised two stages: 1) segmenting the z-stacks into 3D tubular networks of vessels, and 2) characterizing the tubular networks using features derived from the 3D Minkowski functionals (MFs). The 3D MFs measure: the capillary volume, surface area, mean breadth, and Euler number, which quantify progressively higher-order correlations between voxels. The mean breadth is related to tortuosity, wall shear stress and resistance to flow, and the Euler number is related to the density of loops (collaterals). For each retina 4 signature functions were derived by remeasuring the MFs over a range of dilations of scale. Ten features such as critical points and moments of each the signature functions were extracted. The resulting 40 features/retina were analysed by principal component analysis (PCA) and cross-validated step-wise regression.
Features derived from the surface area, mean breadth and Euler number were most correlated with the first PCA component (67% of the variance), which was in turn correlated with age. For the step-wise regressions against age the 25th, 50th, and 75th percentiles of the number of features selected per regression were 2, 3, and 4. These 2 to 4 feature models had r2-values of 0.626 ± 0.089 (mean ± SE), and median F-statistics of 28.7 (p < 10-9). Surface area maximum slope was the most commonly selected feature in 81.4% ± 0.46% (mean ± SE) of the simple 2 to 4 feature models. Three Euler number-based features were selected in between 10.7 ± 0.28% and 22.7 ± 0.50% of the models. Higher-order features such as mean volume or mean surface area are were not selected.
Age-related capillary structure was best characterised by higher order correlations between voxels. The results indicate the importance of pressure-equalizing loops and tortuosity as quantitative measures related to perfusion efficiency.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only