Relevance of 2D radiographic texture analysis for the assessment of 3D bone micro-architecture
Acquisition scheme for each bone sample (diameter: , height: ): three to four scans (, , , ) were necessary to cover the entire sample.
(A) Vertical cross section through the merging of the initial subvolumes. (B) Detail of the junction between two stacks showing misalignment and overlapping. (C) Vertical cross section through the sample after corrections. (D) 3D view of the half-carrot of a sample after corrections.
Illustration of 3D model-based and model independent methods: (A) MIL method: number of intercepts between the bone structure and set of test lines of different directions are calculated, (B) Direct method: the trabecular thickness is evaluated at each voxel as the diameter of the largest ball enclosed in the volume: the thickness is displayed in a frontal slice with a color scale graduated in .
Simulated radiographic images for sample of Fig. 2: (A) image, orientation: ; (B) rescaled ROI of image (A); (C) image, orientation: ; (D) rescaled ROI of image (C).
vs BV/TV (all bones pooled together): calcanei, femurs.
Multiple regressions: (on the left) and BS/BV (on the right) with three textural features. The features involved in the regression and the intermediate and final values of and value are given in the associated tables.
Simple linear regressions for the 25 projections of each sample performed between BV/TV and “Cooccurrence - 2 1 - Variance differences” (left), and Tb.N and “Power Spectrum Density - 0 2 15” (right).
Morphometric parameters of trabecular bone micro-architecture on calcanei and femurs samples. BV/TV refers to bone volume to tissue volume and BS/BV to bone surface to bone volume. Tb.Th, Tb.Sp, and Tb.N are, respectively, the plate model based trabecular thickness, spacing, and number. 2D and 3D DA are, respectively, the 2D and 3D degrees of anisotropy (see the text), and and are, respectively, the model independent trabecular thickness and spacing.
The most pertinent texture parameters for the first and second series of calcanei. The texture parameter name is organized as follows: the first part of the name describes the analysis method, the numbers set the conditions for calculation, and the last part identifies the calculated feature (see the text in Sec. ??? for details). The correlation is given between a texture parameter in column 1, and the corresponding morphometric parameter in column 2.
Stepwise linear regression results in the prediction of morphometric parameters from textural features.
Correlation coefficients of the actual values of morphometric parameters of the second series of calcanei with the predicted values obtained using the models defined with the first series.
Article metrics loading...
Full text loading...