Three-dimensional Image Registration Improves the Long-term Precision of In Vivo Micro-Computed Tomographic Measurements in Anabolic and Catabolic Mouse Models

Authors:
Publisher website
Google Scholar
Abstract
Micro-computed tomography (micro-CT) is a widely used technique to track bone structural and mineral changes in small animals in vivo. Precise definition of volumes of interest (VOIs) in follow-up scans is required to accurately quantify these changes. To improve precision, VOIs can be transferred from baseline images onto follow-ups using image registration. We studied the performance of a registration procedure applied to in vivo data sets of anabolic and osteoporotic bone changes in mice. Micro-CT image data from two separate CD1 mouse data sets were studied. The first included a group treated with parathyroid hormone (PTH) and control and the second, an ovariectomy (OVX) group and control. Micro-CT was performed once per week for 4 weeks at the proximal tibia starting at treatment onset (PTH data set) or after surgery (OVX data set). A series consisting entirely of user-defined VOIs and a registered series where VOIs defined at baseline were transferred to follow-ups were created. Standard bone structural and mineral measurements were calculated. Image registration resulted in a 13–56 % reduction in precision error. Significant effects of registration to detect PTH-induced changes in BV/TV and trabecular BMD were observed. When changes were very pronounced or small, the qualitative improvement observed for the registered data set did not reach statistical significance. This study documents an increase in long-term precision of micro-CT measurements with image registration. Sensitivity to detect changes was improved but not uniform for all parameters. Future study of this technique on images with a smaller voxel size (<19 μm) may capture the effect in greater detail, in particular for trabecular thickness, where changes may be too small to be observed with the voxel size used here. Our results document the value of registration and indicate that the magnitude of improvement depends on the model and treatment chosen.

This publication has 32 references indexed in Scilit: