Processing and analysis of in vivo high-resolution MR images of trabecular bone for longitudinal studies: Reproducibility of structural measures and micro-finite element analysis derived mechanical properties

The authors have developed a system for the characterization of trabecular bone structure from high-resolution MR images. It features largely automated coil inhomogeneity correction, trabecular bone region segmentation, serial image registration, bone/marrow binarization, and structural calculation steps. The system addresses problems of efficiency and inter- and intraoperator variability inherent in previous analyses. The system is evaluated on repetitive scans of 8 volunteers for both two-dimensional (2D) apparent structure calculations and three-dimensional (3D) mechanical calculations using micro-finite element analysis. Coil correction methods based on a priori knowledge of the coil sensitivity and on low-pass filtering of the high-resolution mages are compared and found to perform similarly. Image alignment is found to cause small but significant changes in some structural parameters. Overall the automated system provides on the order of a 3-fold decrease in trained operator time over previous manual methods. Reproducibility is found to be dependent on image quality for most parameters. For 7 subjects with good image quality, reproducibility of 24% is found for 2D structural parameters, while 3D mechanical parameters vary by 4-9%, with percent standardized coefficients of variation in the ranges of 15-34% and 20-38% respectively.