The center of rotation is a physical location in the microCT scanner, defined by the axis of rotation of the sample stage. This physical location is always well defined during calibration of the instrument and fitted by an appropriate algorithm. However, in real images of limited contrast and with X-ray photon noise, this algorithm exhibits poorer precision and the optimum center of rotation cannot be always acquired. Thus, adjustment by operator is necessary to determine whether the center of rotation was correct, in order that the structural information of the sample can be correctly interpreted. In this paper, the effect of center of rotation on the assessment of densitometric and structural properties of trabecular bone was firstly evaluated. Twenty female Sprague-Dawley rats of 7-month-old were randomly assigned to ovariectomized (OVX) and SHAM-operated (SHAM) groups. The left tibiae were harvested at 3 weeks postoperatively. High resolution microCT was used to identify the densitometric and microstructural properties of trabeculae in the proximal ends of tibia. After CT scanning, the best artificial center of rotation for each scan was obtained. Bone parameters analyses were performed on the centers at different places away from the best artificial center of +/-0.2, +/-0.5, +/-1.0, +/-1.5, and +/-2.0 pixels, respectively. The general linear model (GLM) repeated measures procedure was used to investigate the difference in the parameters between the two groups (OVX vs. SHAM) and the possible effects of center displacements. A significant difference between OVX and SHAM groups was found in all parameters (p < 0.05) except Tb.Th, DA, and BS/BV. TBMD, DA, BS/BV, and Conn.D were decreased while BV/TV and Tb.Th were increased with the center deflection. Variations of these parameters were acceptable when the displacements were limited within +/-1.5 pixels for tBMD, BV/TV, DA, and Conn.D, and +/-1.0 pixels for Tb.Th and BS/BV. These changes were similar in both OVX and SHAM groups. The changing curves of bone parameters vs. centers could be well fitted by quadric regression models, by which the real center could be acquired, and thus the precision of microCT analysis would be improved. There were some inevitable differences between the best artificial and real centers.