An investigation of segmentation methods and texture analysis applied to tomographic images of human vertebral cancellous bone.

The goal of this study is to determine architectural and textural parameters on computed tomographic (CT) images, allowing us to explain the mechanical compressive properties of bone. Although the resolution (150 microm) is of the same order of magnitude as the trabecular thickness, this method enables the possibility of perfecting an in vivo peripheral CT system with an acceptable radiation dose for the patient. This study was performed on L2 vertebrae cancellous bone specimens taken after necropsy in 22 subjects aged 47-95 years (mean: 79 years). The segmentation process is a crucial point in the determination of accurate architectural parameters. In this paper the use of two different segmentation methods is investigated, based on an edge enhancement and a region growing approach. The images are compared and the architectural parameters extracted from the images segmented by both methods lead to a quantitative evaluation. The parameters are found to be globally robust towards the segmentation process, although some of them are much more sensitive to the approach used. Highly significant correlations (P < 0.0005) have been obtained between the two segmentation methods for all the parameters, with rho ranging from 0.70 to 0.93. In order to improve the assessment of bone architecture, texture analysis (run length method) was investigated. New features are obtained from an image reduced to 16 grey-levels. Textural parameters in addition to architectural parameters in a multivariate regression model increase significantly (P = 0.01) the prediction of the maximum compressive strength (variation of r2 from 0.75 up to 0.89).

High-resolution computed tomography for architectural characterization of human lumbar cancellous bone: relationships with histomorphometry and biomechanics.

The aim of the present study on human vertebral cancellous bone was to validate structural parameters measured with high-resolution (150 microm) computed tomography (HRCT) by referring to histomorphometry and to try to predict mechanical properties of bone using HRCT. Two adjacent vertical cores were removed from the central part of human L2 vertebral body taken after necropsy in 22 subjects aged 47-95 years (10 women, 12 men; mean age 79 +/- 14 years). The right core was used for structural analysis performed by both HRCT and histomorphometry. Two cancellous bone specimens were extracted from the left core: a cube for HRCT and a compression test, and a cylinder for a shear test. Significant correlations were found between HRCT and histomorphometric measurements (BV/TV, trabecular thickness, separation and number, and node-strut analysis), but with higher values for most of the tomographic parameters (BV/TV and trabecular thickness determined by HRCT were overestimated by a factor 3.5 and 2.5 respectively, as compared with histomorphometry). The maximum compressive strength and Young's modulus were highly correlated (rho = 0.99, p<0.0005). Significant correlation was obtained between bone mineral density (determined using dual-energy X-ray absorptiometry) and the maximum compressive strength (rho = 0. 64, p = 0.002). In addition the maximum compressive strength and architectural parameters determined by HRCT or histomorphometry showed significant correlations (e.g., for HRCT, BV/TV: rho = 0.88, p<0.0005, N.Nd/TV: rho = 0.73, p<0.001). The shear strength was significantly correlated with BV/TV (rho = 0.62, p = 0.002), Tb.Sp (rho = -0.58, p = 0.004) and TSL (rho = 0.55, p = 0.006) measured by HRCT. In conclusion, an HRCT system with 150 microm resolution is not sufficient to predict the true values of the structural parameters measured by histomorphometry, although high correlations were found between the two methods. However, we showed that a resolution of 150 microm allowed us to predict the mechanical properties of human cancellous bone. In vivo peripheral systems with such a resolution should be of interest and would deliver an acceptable radiation dose to the patient.

Mechanical properties of ewe vertebral cancellous bone compared with histomorphometry and high-resolution computed tomography parameters.

The goal of the present study was to determine if a high-resolution computed tomography (HRCT) system with 150 microns resolution was sufficient to predict mechanical properties in ewe lumbar vertebrae. To answer this question, we used a triangular comparison between: HRCT; biomechanics (compression and shear tests); and histomorphometry, which was the reference method for the measurements of morphometric parameters. Two dissected lumbar vertebrae (L-4 and L-5) from 32 ewes were used. Both compressive and shear properties correlated significantly with amount of bone and structural parameters evaluated by histomorphometry (bone volume/tissue volume, trabecular thickness, trabecular separation), but no significant correlation was found with the trabecular number. With our shear test involving the trabecular architecture itself more significant correlations were found with the node-strut analysis parameters than from the compressive test. Significant correlations were also found between HRCT and histological parameters (bone volume/tissue volume, bone surface/bone volume, trabecular separation, trabecular number, total strut length, number of nodes, and number of termini). Correlations between HRCT structural parameters and mechanical properties on L-4 were of the same magnitude as the correlations between the histomorphometric structural parameters and mechanical results on L-5 but with the remarkable advantage the HRCT is a noninvasive method. In spite of the resolution (150 microns) of our HRCT system, which entailed mainly an enlargement of the thinnest trabeculae or their loss during the segmentation process, we obtained coherent relationships between mechanical and tomographic parameters. The thinnest trabeculae probably had little effect on the mechanical strength. Also, this type of resolution allows us to consider the possibility of perfecting an in vivo HRCT system. However, physical density and bone mineral density correlated much better with strength than either classical histomorphometric or tomographic parameters. The current conclusion is fairly negative with respect to the ability of HRCT to assess mechanical properties nondestructively as compared with dual-energy X-ray absorptiometry. But, the noninvasive nature of the imaging modality and the capacity for three-dimensional imaging at arbitrary orientation make HRCT a promising tool in the quantitative assessment of cancellous architecture.