No abstract

Patients with chronic kidney disease (CKD) who undergo kidney transplantation experience bone loss and increased risk of fracture. However, the mechanisms of this bone loss are unclear. Our objective was to use image registration to define the cortex to assess changes in cortical porosity (Ct.Po) in patients undergoing first-time kidney transplantation. We obtained serial measurements of parathyroid hormone (PTH) and bone turnover markers and used high-resolution peripheral quantitative computed tomography (HRpQCT) to scan the distal radius and tibia in 31 patients (21 men, 10 women; aged 51.9 AE 13.4 years) at transplant and after 1 year. Baseline and 1-year images were aligned using a fully automated, intensity-based image registration framework. We compared three methods to define the cortical region of interest (ROI) and quantify the changes: 1) cortical bone was independently defined in baseline and follow-up scans; 2) cortical bone was defined as the common cortical ROI; and 3) the cortical ROI at baseline was carried forward to 1-year follow-up (baseline-indexed). By the independently defined ROI, Ct.Po increased 11.7% at the radius and 9.1% at the tibia, whereas by the common ROI, Ct.Po increased 14.6% at the radius and 9.1% at the tibia. By the baseline-indexed ROI, which provides insight into changes at the endocortical region, Ct.Po increased 63.4% at the radius and 17.6% at the tibia. We found significant relationships between changes in Ct.Po and bone formation and resorption markers at the radius. The strongest associations were found between markers and Ct.Po using the baseline-index method. We conclude that Ct.Po increases throughout the cortex after kidney transplant, and this increase is particularly marked at the endocortical surface. These methods may prove useful for all HR-pQCT longitudinal studies, particularly when changes are expected at the endocortical region.

In this study, we propose interactive graph cut image segmentation for fast creation of femur finite element (FE) models from clinical computed tomography scans for hip fracture prediction. Using a sample of N=48 bone scans representing normal, osteopenic and osteoporotic subjects, the proximal femur was segmented using manual (gold standard) and graph cut segmentation. Segmentations were subsequently used to generate FE models to calculate overall stiffness and peak force in a sideways fall simulations. Results show that, comparable FE results can be obtained with the graph cut method, with a reduction from 20 minutes to 2–5 minutes interaction time. Average differences between segmentation methods of 0.22 mm were not significantly correlated with differences in FE derived stiffness (R2 = 0.08, p = 0.05) and weakly correlated to differences in FE derived peak force (R2 = 0.16, p = 0.01). We further found that changes in automatically assigned boundary conditions as a consequence of small segmentation differences were significantly correlated with FE derived results. The proposed interactive graph cut segmentation software MITK-GEM is freely available online at https://simtk.org/home/mitk-gem.

No abstract