Shape classification techniques for discrete 3D porous media. Application to trabecular bone

Abstract

The clinical process used to screen osteoporosis is the Bone Mineral Density (BMD). Since this density measurement does not cover the entire diagnosis range, work is being carried out on the segmentation of the bone and other complex porous media to provide quantitative information about their microarchitecture. Two shape classification techniques have been recently proposed in the literature. In this paper we compare these different methods and propose a new original rod/plate classification technique. The efficiency of the 3 processes is then studied on test vectors composed of both rods and plates, then applied on real trabecular bone samples. Results of this study emphasize the pros and cons of the 2 published techniques, and discuss the improvements of the new region-growth-based method. Finally, the interest of such a tool in osteoporosis screening is discussed.

3D Image Analysis and Artificial Intelligence for Bone Disease Classification

“…To switch between the 2 skeleton variants, the HSGA improves a recent algorithm [/***compléter bonnassie] which classifies the voxels of an object according to their topological predisposition to belong to a plate or to a rod zone [36]. Then, surface and curve thinning algorithms are applied depending on whether each primitive belongs to the set of rods or plates.…”

“…Plate zones are badly described in the case of beam-only models [15,36], which lead to a nonnegligible bias for morphological results [13]. It is suspected that this lack of geometrical accuracy also alters mechanical results.…”