Bone Marrow Densitometry by Clinical High Resolution Computed Tomography of Human Vertebrae

JAIME PEÑA; GRAEME CAMPBELL; TIMO DAMM; REINHARD BARKMANN; SONJA WLADHAUSEN; FELIX THOMSEN; MATTHIAS KRAUSE; KLAUS PÜSCHEL; CLAUS-C. GLÜER

Prague

Congreso; European Calcified Tissue Society Congress 2014; 2014

European Calcified Tissue Society

Bone mineral density (BMD) as measured by quantitative computed tomography (QCT) is biased by the variable marrow composition since marrow fat reduces the apparent BMD. We developed a marrow densitometry (MD) method that identifies subvolumes in the vertebrae that consist of bone marrow only (bone voids) and determines their average mineral equivalent density values. The method was developed for use with in vivo high resolution QCT (HR-QCT). HR-QCT still has limited spatial resolution and thus we aimed to determine the accuracy of our MD method by comparison with HR-pQCT as gold standard using the Scanco Medical XtremeCT device. 11 excised frozen human T10 vertebrae analysed within the BioAsset study were scanned using the standard HR-pQCT protocol and by HRQCT (Siemens Somatom, 120 kVp, 360 mAs, 0.6 mm slice thickness and 0.3 mm slice increment). HR-pQCT images were filtered with an anisotropic diffusion filter (Matlab, Natick, MA). Both HR-pQCT and HR-QCT images were binarized to exclude air-voxels. Morphological filters using a binary, sphere as structuring element were applied to find spherical subvolumes of bone voids with increasing radii (0.1?1.2 mm, 0.1 mm increment). Mineral equivalent marrow density (MEMD) within the bone-void regions was estimated for each sphere-radius. A convergent MEMD value was observed for larger radii. The corresponding detected regions in both techniques were co-registered (ITK implementation under Structural Insight, in-house software). BMD and tissue mineral density (TMD) measurements within the same search-region were obtained as well. Significant correlations were obtained between the HR-QCT and HR-pQCT based MEMD: (R2=0.42, RMSE=9.9 mg/cc, P<0.03). Strong correlations between both techniques were observed for BMD (R2=0.97, RMSE=3.0 mg/cc, P<0.0001) and TMD of HR-QCT vs HR-pQCT (R2=0.86, RMSE=12.0 mg/cc, P<0.0001). Our results demonstrate that marrow composition can be estimated by HR-QCT with a residual error of about 10 mg/cc mineral equivalent density despite the limitations due to partial volume effects. This MD method is applicable to in vivo assessments of treatment effects and thus can be used to distinguish between true mineral and apparent changes induced by changes in marrow composition.