Block diagram of the proposed methodology for segmenting 3D μCT of trabecular bone.
Voxel array with 26-connectivity.
Mutual information between the filtered image (from 3D μCT data), f, and its morphological opening, γ, and top-hat transform, WTH, for different structuring element sizes.
Algorithm for selecting automatically the size of the structuring element for the top-hat operator.
Single slice from 3D μCT data: (a) Ray aliasing (white arrow) and ring artifacts (black arrow). (b) Image integration. (c) Anisotropic diffusion filtering. (d) Contrast enhancement. The outcomes of the thresholding techniques (e) clustering, (f) maximum entropy, (g) moment preservation, and (h) concavity-based applied to the preprocessed image in (d).
Block diagram of the experimental setup.
Setup of 3D x-ray microtomography imaging.
Example of a single slice from 3D μCT data acquired from rat vertebra. The marked region-of-interest (ROI) contains a notably concentration of trabeculae bone.
Distribution of NPRI values reached by segmentation approaches. The asterisk “*” indicates statistically difference (p < 0.05) between two algorithms.
3D reconstructions from μCT data VOIs (regarding 100 slices) segmented by (a) clustering, (b) moments, (c) concavity, (d) entropy approaches, and (e)−(g) the corresponding ground-truth segmentations.
(a) Bone volume segmented with moment preservation method. (b) Volumetric tetrahedral mesh generated with the Marching Cubes method from (a).
Parameters of the μCT image system acquisition.
NPRI analysis of thresholding algorithms obtained from 38 samples.
Multiple comparisons using Games-Howell test (α = 0.05) for determining differences in NPRI values between thresholding algorithms.
3D morphometric indices of 38 bone samples.
Multiple comparisons for determining differences in BV/TV values.
Article metrics loading...
Full text loading...