Seyedmahdi Hosseinitabatabaei

The Open and Reproducible Musculoskeletal Imaging Research (ORMIR) community is a scientific community dedicated to promoting openness and reproducibility in musculoskeletal imaging, image processing, and computational modelling. In this perspective paper, we outline the motivations for conducting transparent research and provide practical guidelines to implement it. We start with defining open and reproducible research and describing the benefits and challenges of working transparently. Next, we redefine the outputs of a computational research study as—ideally—a combination of data, code, and a publication, recommend a folder and file structure that reflects these three study outcomes, and describe how to maintain and update such a structure during the study and at study publication. Finally, we emphasize that working in an open and reproducible manner is a learning process and the best way to acquire the necessary competencies is simply to start. Lay summary: The ORMIR community promotes openness and reproducibility in musculoskeletal imaging research. In this perspective paper, we explain why transparency matters and recommend how to conduct a computational study in an open and reproducible manner focusing on its three outputs: data, code, and publication. Finally, we highlight that the best way to learn these practices is simply to start.

Volumetric imaging using X-ray microcomputed tomography (µCT) is a mainstream method for 3D structural studies of bones and teeth. Quantitative analysis of mineralized features in general relies on accurate segmentation. The larger the image and the higher the resolution, the greater the incentive to use (semi)automated high-throughput segmentation approaches, primarily using deep learning-aided methods. In this study we pretrained and evaluated two U-Net segmentation models, called Universal Bone and Universal Jaw, using a diverse 3D image library from contributing colleagues. Our assumption was that pretraining a U-Net on skeletal and dental features would result in higher accuracy and throughput, in comparison with a naïve U-Net trained by an end-user on a limited local dataset. The Universal Bone binary segmentor pretrained on a relatively heterogeneous subset of images showed satisfactory generalization capacity, and was applicable to unseen images with minimal tuning. The Universal Jaw four-class segmentor (background, bone, dentin, enamel) was pretrained on a narrower range of images and required rather extensive retraining on manually retouched image segmentations. The performance of pretrained segmentation U-Nets is defined not only by the number of segmentation classes (the fewer the better), similarity of the unseen image spacing to the training set (the closer the better), but even more so by class balance. For particularly large 3D images, extensive manual retouching and re-training was nevertheless valuable attributable to the high throughput of the pretrained segmentor. Despite achieving partial success in pretraining specialized segmentation models, we advocate for future concerted efforts to merge and expand image collections, share pretrained models and utilize expert segmentations for benchmarking.

Introduction: Osteogenesis imperfecta (OI) is characterized by bone fragility with frequent fractures, especially in children. Some studies have used peripheral-QCT (pQCT) to examine bone density in children with OI and one cross-sectional study used HR-pQCT in 9 children with OI. Materials and Methods: We compared bone density, microstructure, and strength changes over 1 yr at the metaphysis and diaphysis of the radius and tibia, as well as lumbar spine (LS) bone mass and area, in 20 children with OI under bisphosphonate treatment and 20 age- and sex-matched controls. For the double-stack metaphyseal scans, we developed an algorithm to correct stack misalignment. For the LS, we used DXA. Results: At the tibial metaphysis, children with OI had lower baseline total and trabecular volumetric BMD (vBMD), trabecular and cortical microstructure, and strength. At the radial metaphysis, baseline trabecular microstructure and strength were lower in children with OI. At the tibial and radial diaphysis, children with OI had lower bone area and strength. At the tibial and radial metaphysis, 1-yr increases in most measurements were similar for both groups. Cortical vBMD and trabecular separation (radius) only increased in the OI group. Trabecular vBMD, volume fraction (tibia), number (tibia), and total vBMD (radius) only increased in the control group. At the tibia and radius diaphysis, cortical vBMD increased in the OI group. Tibial Ct.Po decreased in children with OI. At the LS, the OI group had lower bone mass, density, and area, but similar longitudinal changes. Discussion: In summary, baseline trabecular bone density, trabecular and cortical microstructure, and strength were lower in metaphyseal regions of children with OI. At diaphysis, children with OI had lower bone area and strength. While longitudinal changes over 1 yr were generally comparable, certain outcomes demonstrated differences. These data are essential for powering future clinical trials.

Cerebral palsy (CP) is a non-progressive neurological syndrome resulting in abnormal muscle tone, movement, and posture. It is unclear whether ambulatory children with CP have deficits in bone quantity or quality. Furthermore, the relationship between abnormal muscle tone, altered function, and bone health remains largely unexplored. This observational study investigated bone mineral density (BMD) and microarchitecture in ambulatory children with spastic CP and associations of BMD with function, muscle spasticity, and gait. Children with spasticity in both lower limbs (n = 12) aged 3-8 years were recruited. Areal BMD was measured with dual energy x-ray absorptiometry (DXA) at the proximal femur and lateral distal femur and compared to normative data. High resolution peripheral quantitative computed tomography (HR-pQCT) was performed at the metaphyseal tibia and radius in a subset of participants (n = 5) and compared to healthy children (n = 7). Gait pathology and cardiopulmonary function were investigated with the Gait Deviation Index, Edinburgh Visual Gait Score, and energy expenditure index. DXA areal BMD (aBMD) Z-scores at the lateral distal femur were within a normal range. However, the CP group’s median aBMD Z-score at the proximal femur was −1.8 (IQR: −2.2, −1.2, p = .03) indicating potential skeletal fragility. Strong correlations were found between gait pathology and DXA-based bone outcomes (correlation coefficient 0.62 [p = .04] to 0.73 [p = .01]) as well as energy expenditure index and DXA-based bone outcomes (correlation coefficient −0.63 [p = .03] to −0.98 [p ≤ .001]). At the metaphyseal tibia, children with spastic CP had significant deficits in HR-pQCT-measured bone geometry and trabecular microarchitecture: 35% lower total area, 42% lower trabecular area, and 48% lower trabecular number than controls. HR-pQCT parameters were similar between groups at the metaphyseal radius. These differences in tibial metaphysis size and trabecular microarchitecture are similar to those observed in disuse and thus could be a result of abnormal biomechanics or low levels of physical activity.

Timelapse imaging using high-resolution peripheral quantitative computed tomography has emerged as a non-invasive method to quantify bone (re)modeling. However, there is no consensus on how to perform the procedure. As part of the ASTEROID phase-2b multicenter trial, we used 29 same-day repeated scans from adults with OI to identify a method that minimized measurement error. We evaluated input image type, registration method, segmentation mask, and for grayscale images various values for the voxel density difference considered formed or resorbed, minimum formation/resorption cluster size, and Gaussian smoothing sigma. We verified the accuracy of our method and then used it on longitudinal scans (baseline, 6, 12, 18, and 24 mo) from 78 participants to assess bone formation and resorption induced by an anabolic (setrusumab) and anti-catabolic (zoledronic acid) treatments as part of the ASTEROID trial. Regardless of image registration method, binary input images resulted in large errors ~13% and ~8% for first- and second-generation scanners, respectively. For the grayscale input images, errors were smaller for 3D compared to matched angle registration. For both scanner generations, a density threshold of 200 mgHA/cm3 combined with Gaussian noise reduction resulted in errors <1%. We verified the method was accurate by showing that similar regions of bone formation and resorption were identified when comparing each scan from the same-day repeated scans with a scan from another timepoint. Timelapse analysis revealed a dose-dependent increase in bone formation and resorption with setrusumab treatment. Zoledronic acid altered bone changes in favor of formation, although no changes reached statistical significance. This study identifies a timelapse method that minimizes measurement error, which can be used in future studies to improve the uniformity of results. This non-invasive imaging biomarker revealed dose dependent bone (re)modeling outcomes from 1 year of setrusumab treatment in adults with OI.

Osteogenesis imperfecta (OI) is a rare genetic disorder commonly caused by variants of the type I collagen genes COL1A1 and COL1A2. OI is associated with increased bone fragility, bone deformities, bone pain, and reduced growth. Setrusumab, a neutralizing antibody to sclerostin, increased areal bone mineral density (aBMD) in a 21-week phase 2a dose escalation study. The phase 2b Asteroid (NCT03118570) study evaluated the efficacy and safety of setrusumab in adults. Adults with a clinical diagnosis of OI type I, III, or IV, a pathogenic variant in COL1A1/A2, and a recent fragility fracture were randomized 1:1:1:1 to receive 2, 8, or 20 mg/kg setrusumab doses or placebo by monthly intravenous infusion during a 12-month treatment period. Participants initially randomized to the placebo group were subsequently reassigned to receive setrusumab 20 mg/kg open label. Therefore, only results from the 2, 8, and 20 mg/kg double-blind groups are presented herein. The primary endpoint of Asteroid was change in distal radial trabecular vBMD from baseline at Month 12, supported by changes in high-resolution peripheral quantitative computed tomography micro finite element-derived bone strength. A total of 110 adults were enrolled with similar baseline characteristics across treatment groups. At 12 months, there was a significant increase in mean (SE) failure load in the 20 mg/kg group (3.17% [1.26%]) and stiffness in the 8 (3.06% [1.70%]) and 20 mg/kg (3.19% [1.29%]) groups from baseline. There were no changes in radial trabecula vBMD (p > 0.05). Gains in failure load and stiffness were similar across OI types. There were no significant differences in annualized fracture rates between doses. Two adults in the 20 mg/kg group experienced related serious adverse reactions. Asteroid demonstrated a beneficial effect of setrusumab on estimates of bone strength across the different types of OI and provides the basis for additional phase 3 evaluation. Osteogenesis imperfecta, or OI, is a rare disorder affecting patients’ bones causing pain and an increased chance of the bone breaking. Setrusumab is a possible treatment for OI being studied in a clinical trial called Asteroid. The goal of Asteroid was to determine which dose of setrusumab helped adults with OI the most: 2, 8, or 20 mg/kg. Researchers looked at the density of patients’ bones and estimated how strong their bones were before setrusumab and again after 12 months of treatment to see how they improved with treatment. Researchers could compare these improvements to see which dose of setrusumab helped patients the most. Patients on the highest dose of setrusumab (20 mg/kg) experienced improvements in the density of their arm bones (radius) and leg bones (tibia) after 12 months. The strength of these bones also improved. The density of other bones including the spine, hip, and the overall skeleton (total body) also improved with treatment. Of patients who had side effects after receiving setrusumab, most were mild or moderate intensity. Overall, setrusumab improved the bones of patients with OI with no serious safety concerns. More studies will include even more patients to see how setrusumab can improve their bones.

Purpose of Review: The purpose of this review is to summarize current approaches and provide recommendations for imaging bone in pediatric populations using high-resolution peripheral quantitative computed tomography (HR-pQCT). Recent Findings: Imaging the growing skeleton is challenging and HR-pQCT protocols are not standardized across centers. Adopting a single-imaging protocol for all studies is unrealistic; thus, we present three established protocols for HR-pQCT imaging in children and adolescents and share advantages and disadvantages of each. Limiting protocol variation will enhance the uniformity of results and increase our ability to compare study results between different research groups. We outline special cases along with tips and tricks for acquiring and processing scans to minimize motion artifacts and account for growing bone. Summary: The recommendations in this review are intended to help researchers perform HR-pQCT imaging in pediatric populations and extend our collective knowledge of bone structure, architecture, and strength during the growing years.

Repositioning error in longitudinal high-resolution peripheral-quantitative computed tomography (HR-pQCT) imaging can lead to different bone volumes being assessed over time. To identify the same bone volumes at each time point, image registration is used. While cross-sectional area image registration corrects axial misalignment, 3D registration additionally corrects rotations. Other registration methods involving matched angle analysis (MA) or boundary transformations (3D-TB) can be used to limit interpolation error in 3D-registering micro-finite-element data. We investigated the effect of different image registration methods on short-term in vivo precision in adults with osteogenesis imperfecta, a collagen-related genetic disorder resulting in low bone mass, impaired quality, and increased fragility. The radii and tibiae of 29 participants were imaged twice on the same day with full repositioning. We compared the precision error of different image registration methods for density, microstructural, and micro-finite-element outcomes with data stratified based on anatomical site, motion status, and scanner generation. Regardless of the stratification, we found that image registration improved precision for total and trabecular bone mineral densities, trabecular and cortical bone mineral contents, area measurements, trabecular bone volume fraction, separation, and heterogeneity, as well as cortical thickness and perimeter. 3D registration marginally outperformed cross-sectional area registration for some outcomes, such as trabecular bone volume fraction and separation. Similarly, precision of micro-finite-element outcomes was improved after image registration, with 3D-TB and MA methods providing greatest improvements. Our regression model confirmed the beneficial effect of image registration on HR-pQCT precision errors, whereas motion had a detrimental effect on precision even after image registration. Collectively, our results indicate that 3D registration is recommended for longitudinal HR-pQCT imaging in adults with osteogenesis imperfecta. Since our precision errors are similar to those of healthy adults, these results can likely be extended to other populations, although future studies are needed to confirm this. © 2022 American Society for Bone and Mineral Research (ASBMR).

High‐resolution peripheral quantitative computed tomography (HR‐pQCT) imaging, together with computational finite element analysis (FEA), offers an attractive, non‐invasive tool to quantify bone strength development in pediatric studies. Evidence of annual changes and errors in repeated HR‐pQCT measures is limited, and time intervals required to reliably capture changes in children's bone strength or micro‐architecture have not yet been defined. Our objectives were to: 1) quantify annual changes in bone strength and micro‐architectural properties, 2) define precision errors for pediatric bone strength outcomes; 3) characterize annual changes in contrast to pediatric precision errors; and 4) estimate monitoring time intervals (MTIs) required to reliably characterize bone development at the distal radius and tibia. We obtained distal radius (7% of ulnar length) and tibia (8%) bone properties using HR‐pQCT and FEA from 38 follow‐up study participants (21 girls) at baseline (mean age 10.6, SD 1.7 years) and after 1‐year; and from 32 precision study participants (16 girls) at baseline (11.3, 1.6 years) and after 1‐week. We characterized mean annual changes (paired t‐tests) contrasted to pediatric precision errors (CV%_RMS) and estimated MTIs. Annual increases in bone strength, total area, cortical thickness and density ranged between 3.0‐25.3% and 2.4‐15.6% at the distal radius and tibia, respectively. Precision errors for all bone strength outcomes were ≤6.8% and ≤5.1% at the distal radius and tibia, respectively, and appeared lower than annual gains in bone strength at both sites. Cortical porosity decreased 19.6% at the distal radius and 6.6% at the distal tibia and these changes exceeded respective precision errors, indicating cortical bone consolidation. MTIs ranged between 0.5‐∞ years at the distal radius and 0.5‐5.9 years at the distal tibia. Estimated MTIs suggest that pediatric bone strength, cortical bone density and porosity development can be reliably monitored with annual measurements. This article is protected by copyright. All rights reserved