Magnetic resonance osteodensitometry in human heel bones: correlation with quantitative computed tomography using different measuring parameters.

Abstract

Density of trabecular bone structures in human heel bones was assessed by 3D magnetic resonance (MR) gradient echo imaging (GEI) with multiple echoes. Different spatial resolutions were applied to investigate the influence of the pixel size on signal characteristics in GEI and to find suitable measuring parameters for a maximum correlation between GEI and bone mineral density obtained by quantitative computed tomography (QCT). Thirty-five patients aged 31 to 65 years with suspected osteoporosis underwent MR and QCT examinations of the heel bones. The MR protocol included 3D GEI with three echo times (TE1 = 9.3, TE2 = 27.9, and TE3 = 46.5 ms) and isotropic pixel sizes of (0.6 mm)3, (1.2 mm)3, and (2.4 mm)3. Several subregions in the heel bones were analyzed. For determination of signal reduction with increasing TE, signal intensity ratios were calculated pixelwise from images with TE2/TE1 and TE3/TE1. All examinations showed that the T2*-related signal decrease was more pronounced for lower spatial resolution. In the dorsal part of the heel bones, the correlation between signal ratios in GEI and QCT-based bone mineral density values was between r = -0.86 for a spatial resolution of (0.6 mm)3 and r = -0.73 for (2.4 mm)3. Areas with low trabecular density in the ventral part of the heel bones showed clearly lower correlation coefficients (-0.65 < r < -0.67). Spatial resolution in 3D GEI clearly influences the T2*-related signal characteristics. Despite measuring different physical properties of spongy bone by GEI and QCT, a relatively high correlation between GEI with small pixel sizes and QCT was obtained in the dorsal part of the heel bones, but not in the ventral part with partly thickened trabeculae and irregular distribution. However, standardized measuring protocols with preferably small pixel sizes (as low as [0.6 mm]3) should be applied, and correlation curves must be determined, dependent on the actual bone marrow site, before clinical routine MR osteodensitometry becomes possible.