Novel image analysis for automated quantification of dynamics of 3D lamina cribrosa microstructure under controlled IOP modulation imaged by in-vivo OCT

Abstract

Purpose : Quantitative analysis of structural changes of the lamina cribrosa (LC) under varying intraocular pressures (IOP) levels may lead to new insights into the pathophysiology of glaucoma. We present a novel automatic image analysis method for quantitative assessment of dynamic changes of the LC microstructure imaged in-vivo by OCT at different IOP levels.

Methods : Seven healthy non-human primates were imaged with spectral-domain OCT (Leica, Chicago, IL). 3D images of the LC region were acquired at baseline and two levels of elevated IOP. Image processing includes rigid registration to account for ocular motion and multi-scale filtering by Gaussian derivatives. Quantitative characterization of 3D pore and beam segmentations is using Euclidean distance transformation and skeletonization, resulting in maps of centers of tube-shaped pores and mesh-shaped beams with associated local thickness. Pore and beam sizes across regions were modeled as distributions of width. Expert identification of a 3D region-of-interest (ROIs) identical for all co-registered images per pressure series represents the only manual interaction, analysis is otherwise automatic and thus fully reproducible. The pipeline was tested on seven pressure series as part of our non-human primate research study. Statistical significance of pressure-induced changes to thickness of pores and beams was assessed on thickness distributions for baseline, IOP1 and IOP2 via pair-wise non-parametric Kolmogorov-Smirnov testing (Figure 2).

Results : Automated 3D LC segmentation was successfully accomplished in all eyes and pressure settings (Figure 1 illustrates processing steps for one case). Despite strong variation of LC appearance and size of ROIs across subjects, we observe a trend of decrease of pore thicknesses (overall average change -4.04%) and increase of beam thicknesses (overall average change 4.63%) from baseline to elevated IOP (Table in Figure 2).

Conclusions : We demonstrate automatic and efficient quantitative analysis to support OCT-based 3D analysis of the dynamics of LC microstructure as response to elevated IOP or changes occurring over time. Preliminary testing on seven cases demonstrates feasibility.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.