Abstract
Purpose: :
En face adaptive optics scanning laser ophthalmoscope (AOSLO) images of the anterior lamina cribrosa surface (ALCS) represent a 2D projected view of a 3D surface. Laminar pore parameters calculated from these 2D images do not best represent the native 3D geometry. Using images acquired in living eyes, we modeled the ALCS in 3D and transformed 2D AOSLO images onto the 3D surface to better characterize laminar pores.
Methods: :
AOSLO images of the ALCS were acquired in vivo in the control and glaucomatous eyes of a rhesus monkey with unilateral experimental glaucoma and were used to calculate laminar pore areas. Spectral domain optical coherence tomography (SDOCT) [Spectralis HRA+OCT] cross-sectional radial scans (20° field, 48 B-scans) of the optic nerve head were acquired in both eyes. The ALCS was manually marked in SDOCT B-scans using a custom MATLAB program. A thin plate spline was fit to the marked points to model the ALCS in 3D. After registering and projecting the 2D AOSLO images onto the 3D surface, the surface was tessellated into a 3D triangular mesh. The 3D surface area of a given pore was calculated by adding the area of the individual triangles that formed the surface within the pore’s boundary. The differences in the area of corresponding pores in the 2D and 3D images from living eyes were compared.
Results: :
This modeling method was validated using synthetic data (surfaces and objects of known geometry). For example, a circular pore projected onto a hemisphere resulted in a transformed surface area within 0.7% of the theoretical value. In the monkey with experimental glaucoma, the mean ALCS heights in control and glaucomatous eyes were 193.1 μm and 361.7 μm, respectively, indicating a more severely sloped ALCS in the glaucoma eye. The root mean square (RMS) errors in fitting a thin plate spline to the marked ALCS points in the control and glaucomatous eyes were 10 μm and 9.4 μm, respectively. Mean pore areas in the 2D and corresponding 3D images were 1,158 ± 976 μm2 vs. 1,205 ± 1004 μm2 in the control eye and 1,488 ± 1123 μm2 vs. 1,645 ± 1243 μm2 in the glaucomatous eye, respectively. The mean increase in pore area after 3D transformation was 4% (46.8 ± 32.1 μm2) in the control eye and 10.5% (156.9 ± 133.9 μm2) in the glaucoma eye.
Conclusions: :
The thin plate spline can accurately fit laminar surfaces with different slopes (as indicated by the equally low RMS fitting errors measured in the control and glaucoma eyes). The larger increase in mean laminar pore area in the glaucomatous eye after 3D transformation is reflective of its more steeply curved ALCS (compared to the control eye). This 3D transformation and tessellation method can potentially be used to better understand 3D changes in laminar pore geometry in glaucoma.
Keywords: image processing • lamina cribrosa • optic disc