Challenges in longitudinal measurements with HR-pQCT: evaluation of a 3D registration method to improve bone microarchitecture and strength measurement reproducibility

Definition of identical regions between repeated computed tomography (CT) scans is a key factor to monitor changes in bone microarchitecture. In longitudinal studies, accurate determination of the volume of interest (VOI), using three dimensional (3D) registration may improve precision. Therefore, the aim of our study was to investigate the short-term reproducibility of bone geometry, density, microstructure and biomechanical parameters assessed by HR-pQCT and micro-finite element (μFE) derived analyses, using the cross-sectional area (CSA) registration method in comparison with the use of 3D registration, to find overlapping regions between scans. Fifteen healthy individuals (aged 21-47 years) underwent 3 separate scans at the distal radius and tibia, within a one-month interval. Reproducibility was assessed after double contouring the cortical compartment and after applying three different methods to determine the common region between repeated scans: (i) the VOI was determined with no registration, i.e., on 110 slices, (ii) the VOI was determined after CSA-based registration, and (iii) the VOI was determined after 3D registration. Both pre- and post-registration short-term reproducibility for each subject was determined. With no registration, CVrms of geometry parameters ranged from 0.5 to 3.7%, showing a slight variation in the CSA between scans. When the CSA registration method was employed, the variability of geometry (CVrms<1.8%) and density parameters (CVrms<1.8%), was better than that obtained without registration. By removing the effect of repositioning, the 3D registration further improved the reproducibility of cortical bone measurements compared to other methods. Indeed, significant improvements were found for cortical geometry and microstructure measurements (CVrms ranged from 0.4% to 10.7% at both sites; p<0.05), whereas the impact on trabecular bone measurements was restricted to its geometry parameter. The repositioning error was significantly reduced, most markedly at the radius compared to the tibia. For μFE measures, the impact of 3D registration on whole bone stiffness was negligible, indicating adequate assessment of longitudinal changes in estimated biomechanical properties, even without registration. In conclusion, we have shown that the 3D registration improved the identification of the common region retained for longitudinal analysis, contributing to improve the reproducibility of cortical bone parameter measurements. We also quantified the minimally detectable bone changes to help designing future studies with HR-pQCT.