Visualizing and Interacting with Large Volumetric Datasets Containing the Optic Nerve Head (ONH) Connective Tissues

Abstract

Abstract: : Purpose: To develop software to visualize and dynamically interact with large volumetric reconstructions of the ONH tissues to define the underlying connective tissue architecture. Background: Quantitative characterization of the onset and progression of glaucomatous connective tissue damage through finite element modeling requires an accurate coordinate representation of the connective tissue structures to be measured and modeled. Our 3–D reconstructions (4.0x10⁸ to 1.1x10⁹ voxels) of serial high–resolution (2.5x2.5 µm/pixel) images of the stained face of an ONH specimen embedded in paraffin and sectioned at 3 µm intervals on a microtome. No commercial software for visualizing the entire volumetric datasets is currently available for a workstation environment, nor to dynamically interact with the dataset to identify, isolate and extract the individual connective tissue architectural components (we currently do this by manually segmenting the structures in static 2–D sections, requiring 100–200 hours per ONH). Methods: The open–source Visualization Toolkit (VTK) has been extended in C++ to create software that allows the user to explore sub–sampled views of the entire volume and selectively render full–scale radial and orthogonal voxel data planes. Within a given data plane, we can select and define discrete sets of points and are implementing the ability to edit and categorize them within any section through the dataset. These point cloud delineations identify the boundaries of the principle architectural components, producing segmented regions of interest for further processing or qualitative assessment and generating geometric surface descriptions for quantitative hypothesis testing and finite element model construction. Results: Discrete point sets that quantitatively represent the following structures are now being generated in 3–8 hours per ONH: internal limiting membrane, Bruch's membrane opening (anterior scleral canal opening), Bruch's membrane proper, Border Tissue of Elschnig, anterior and posterior aspects of the peripapillary sclera and lamina cribrosa, scleral canal wall, pial and dural sheaths, and the central retinal vessels. Conclusions: Our new software supports interactive visualization and point cloud delineation of ocular structures contained within large volumetric datasets.

Keywords: imaging/image analysis: non–clinical • optic disc • computational modeling