Purchase this article with an account.
Ian A. Sigal, Jonathan L. Grimm, John G. Flanagan, C R. Ethier, Inka Tertinegg, Paul Sanfilippo; Statistical Shape Analysis of the Lamina Cribrosa. Invest. Ophthalmol. Vis. Sci. 2011;52(14):2436.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop statistical models of lamina cribrosa shape (LCS).
Lamina outlines were described with 2D elliptic Fourier analysis. The first Fourier harmonic was used to align the outlines and eliminate variation unrelated to shape. Principal component analysis of all harmonics was then employed to identify the main modes of shape variation. To test that the technique adequately captures LCS, we first analyzed a set of 143 laminas with simple and known shape variation produced from virtual sections of a generic simplified parameterized model of the optic nerve head. We then analyzed a set of 20 lamina outlines obtained from manually segmented superior-inferior histological sections of normal human eyes.
Analysis of the simplified laminas successfully identified the mean and the imposed variations in LCS (Figure). Curvature was the main mode of shape variation, accounting for 66% and 46% of shape variance in the simplified and realistic laminas, respectively. The top 5 modes accounted for 96% and 87% of shape variance in simplified and realistic laminas, respectively. The sharp corners at the lamina insertion into the sclera were only approximated by 60 harmonics.
Statistical models can capture LCS and its variation. These models are more flexible than current techniques because they incorporate covariations, have minimal redundancy, and do not require defining a priori the features to extract. In the set of real laminas our analysis identified substantial superior-inferior shape variation asymmetry. To characterize lamina insertions precisely more than 60 harmonics are needed. An efficient characterization of shape may increase the sensitivity to detect changes and abnormalities.
This PDF is available to Subscribers Only