Micro-computed tomography (microCT) has become an important tool for morphological characterization of cortical and trabecular bone. Quantitative assessment of bone tissue mineral density (TMD) from microCT images may be possible; however, the methods for calibration and accuracy have not been thoroughly evaluated. This study investigated hydroxyapatite (HA) phantom sampling limitations, short-term reproducibility of phantom measurements, and accuracy of TMD measurements by correlation to ash density. Additionally, the performance of a global and a local threshold for determining TMD was tested. The full length of a commercial density phantom was imaged by microCT, and mean calibration parameters were determined for a volume of interest (VOI) at 10 random positions along the longitudinal axis. Ten different VOI lengths were used (0.9-13 mm). The root mean square error (RMSE) was calculated for each scan length. Short-term reproducibility was assessed by five repeat phantom measurements for three source voltage settings. Accuracy was evaluated by imaging rat cortical bone (n = 16) and bovine trabecular bone (n = 15), followed by ash gravimetry. Phantom heterogeneity was associated with <0.5% RMSE. The coefficient of variation for five repeat measurements was generally <0.25% across all energies and phantom densities. Bone mineral content was strongly correlated to ash weight (R (2) = 1.00 for both specimen groups and both threshold methods). Ash density was well correlated for the trabecular bone specimens (R (2) > 0.80). In cortical bone specimens, the correlation was somewhat weaker when a global threshold was applied (R (2) = 0.67) compared to the local threshold method (R (2) = 0.78).