Skip to main content
NCBI home page
JBMR Plus logoLink to JBMR Plus
. 2025 Dec 18;10(2):ziaf194. doi: 10.1093/jbmrpl/ziaf194

Repeatability of objective bone and joint measures in the knee using weight-bearing computed tomography: three-dimensional quantification of bone mineral density, joint space width, and subchondral bone plate thickness

Tadiwa H Waungana 1,2, Ria Mangat 3, Chloe Chen 4, Ainsley C J Smith 5,6, Yousif Al-Khoury 7,8, Michael T Kuczynski 9, Tom Turmezei 10,11, Donald D Anderson 12, William Brent Edwards 13,14, Sarah L Manske 15,16,17,✉
PMCID: PMC12815259  PMID: 41562112

Abstract

Bone alterations and degenerative joint structural changes are frequently observed in knee osteoarthritis (OA). Objective measures of subchondral bone plate thickness (SBP.Th), apparent BMD, and joint space width (JSW) have begun to be used to better assess and understand disease progression. Weight-bearing CT (WBCT) allows 3D assessment of multiple bone and joint parameters; however, there is a paucity of literature investigating factors that might affect our ability to track changes over time. The purpose of this study was to investigate the repeatability of BMD, JSW, and SBP.Th measures obtained from WBCT images. Same-day, scan–rescan, knee WBCT images were acquired from 37 healthy adults (20 female, mean age: 24.6 yr). We quantified trabecular bone BMD at the proximal tibia and implemented joint space and cortical bone mapping to measure tibiofemoral JSW and SBP.Th, respectively. Test–retest repeatability was evaluated using coefficients of variation (CVRMS%) and least significant change (LSC). Mean trabecular BMD exhibited CVRMS% ranging from 1.87% to 2.85% and LSCs between 13.53 and 17.39 mgHA/cm3. For JSW, CVRMS% and LSC were less than 1% and 0.2 mm, respectively. SBP.Th measures had higher CVRMS% (9.75%-11.29%) but similar LSC values (0.09-0.26 mm) in comparison to JSW values. The high repeatability of subchondral bone and joint parameters underscore the potential of WBCT for quantifying and monitoring bone and joint changes in knee OA progression and management. However, standardization of acquisition and analysis methods will likely be required to ensure reliable evaluation over multiple time points.

Keywords: knee, osteoarthritis, BMD, subchondral bone, joint space width, weight-bearing CT, repeatability

Graphical Abstract

Graphical Abstract.

Graphical Abstract

Open in a new tab

Introduction

The knee is the most common site for the development of osteoarthritis (OA) accounting for approximately 85% of all OA cases globally.1,2 The presence of osteophytes, subchondral bone sclerosis, joint space narrowing, and subchondral bone cysts have been identified as hallmark features of knee OA.3,4

Bone has emerged as a potential therapeutic target for OA management.1,5 Subchondral bone changes, such as thickening of the subchondral bone plate and subchondral trabecular BMD alterations, are well recognized features of knee OA that may play a role in OA initiation and progression.6,7 While it has recently been shown that objective measures of subchondral bone sclerosis were associated with knee OA disease severity,8 there remains a need to develop our understanding of bone behavior and distribution in OA.

Although weight-bearing radiographs remain the recommended imaging modality of choice to evaluate structural features in knee OA, anatomical joint overlap and potential errors in X-ray beam angulation may lead to poor reproducibility and sensitivity to change over time.9–12 These limitations can be overcome by cone beam, weight-bearing CT (WBCT), which is unhindered by the projectional limitations of conventional radiography, leading to a more detailed and accurate representation of functional joint behavior.13,14

These benefits of WBCT have been most notable in the evaluation of joint space narrowing, a surrogate measure for cartilage thickness and meniscal integrity. Indeed, previous studies have demonstrated the feasibility and improved diagnostic performance of 3D joint space width (JSW) measurements in knee OA using WBCT.15–17 3D JSW measurements from WBCT are more sensitive at detecting disease-related changes over time than conventional radiography, further underscoring their strength as a tool to understand knee OA.18

However, the lack of standardized image acquisition and analysis methods for WBCT present a barrier for establishing its clinical utility.19 Key in this context is obtaining repeatable positioning and stance during acquisitions and evaluating the robustness and precision of different analysis methods. For example, differences in knee flexion may cause inconsistencies in repeated JSW measurements that can be as large as 1-yr’s worth of joint space narrowing in OA knees.12,20

In terms of OA-related changes in BMD, we have recently shown that WBCT can be used to accurately measure BMD at the knee.21 However, the image grayscale values (attenuation units, AU) provided by WBCT may be affected by beam-hardening and X-ray scatter artifacts, resulting in inaccurate, and potentially sporadic, BMD measurements that may compromise the ability to track changes over time.22–24 Similarly, although sclerotic bone changes have been quantified through subchondral bone plate thickness (SBP.Th) estimates that are unconstrained by the CT image resolution,8,25 the effects of positioning and grayscale variability on measurement precision are not well understood.

Therefore, the aim of this study was to assess the repeatability of BMD, tibiofemoral JSW, and SBP.Th measures using cone beam WBCT. We quantified the effects of incidental minor variations in image acquisition and patient stance, despite a standardized protocol, as well as the influence of systematic variation in analysis inputs and methodology. Our findings are useful for evaluating the robustness of analysis techniques and lay the foundation for establishing guidelines to monitor bone and joint changes using WBCT.

Materials and methods

Study design and participant recruitment

We recruited 37 healthy participants (20 female, mean age: 24.5 ± 6.5 yr, mean height: 171.9 ± 10.9 cm, mean weight: 73.2 ± 15.7 kg) for this study. Participants were recruited from Calgary, Alberta and the surrounding area. All participants provided written consent, and ethics approval for this study was granted under the Conjoint Health Research Ethics Board at the University of Calgary (REB16-0014). Participant inclusion was limited to those with healthy knees as this was deemed a necessary first step to ensure that the methods produce precise results in a homogeneous sample of knees. Consequently, participants were excluded, if they had metal implants in either knee or were unable to bear their full body weight in bipedal stance.

Scanning by WBCT

Bilateral knee images were acquired using WBCT (HiRise, CurveBeam) in the McCaig Institute for Bone and Joint Health at the University of Calgary. A standard weight bearing, bilateral knee scan protocol was used (130 kVp, 6.5 mA, 0.3 mm isotropic voxels, 720 projections over a 360° projection angle) with volumetric back projection producing a single image stack with a 19.6 cm axial scan length. Each bilateral scan took ~40 s delivering 30 μSv of radiation dose.

Participants were scanned twice on the same day, and each participant was instructed to leave the scanner between scans for repositioning (approximately 5 min between scans). For both scans, participants stood with the toes and medial surfaces of the feet placed against vertical platforms, externally rotating the feet by 10° (Figure S1).21,26 The thighs of participants were positioned against a vertical position plate resulting in approximately 20° of knee flexion depending on the relative foot/leg length ratio of each participant.27

A bone density calibration phantom (QRM), containing 3 cylindrical hydroxyapatite (HA) inserts (18 mm diameter × 200 mm, 100.4 mgHA/cm3, 399.9 mgHA/cm3, and 795.7 mgHA/cm3) housed in a homogenous water-equivalent resin, was scanned at the beginning and end of each session to facilitate asynchronous BMD calibration. Calibration was performed using phantom image data acquired at the beginning of the scanning session, with the end-of-session phantom image data being used to verify the stability of image grayscale values (AU) over the scan session.

Image analysis

An enhance-and-segment image segmentation pipeline with manual post-correction was used to obtain periosteal masks of the distal femur and proximal tibia.28 Briefly, an edge-enhancement filter was used to emphasize the bone edges and joint space, while a label image derived from the original grayscale image was used to sparsely identify the femur, tibia, and patella. These images were then combined to generate bone masks using a graph cut segmentation algorithm. All subsequent imaging analysis was performed using custom Python (v3.8.5) scripts developed using SimpleITK (v2.0.2), Visualization Toolkit (v9.3.1) and PyVista (v0.44.1) libraries.

Bone mineral density

We considered the subchondral trabecular bone as the tibial volume between the epiphyseal line and the subchondral bone surface. Principal component analysis was used to estimate the principal axes of the tibia from the original bone mask. Landmark images of the epiphyseal line were manually generated for each grayscale image using ITK-Snap (v3.8.0). 3D morphological erosion was performed to “peel” the tibial bone mask by 8 voxels to exclude the cortical bone and subchondral bone plate.29 Eight voxels (~2.4 mm) were chosen as this was approximately the minimum number of voxels needed to exclude the thickest regions of the cortical bone/subchondral bone plate in our dataset. A quadratic plane of best-fit was obtained from the epiphyseal landmark data and used to exclude bone distal to the epiphyseal line—creating a subchondral trabecular bone mask. The principal axes were subsequently used to split the subchondral trabecular bone mask into anteroposterior and mediolateral quadrants for BMD analysis (Figure 1). To derive BMD, image grayscale values (AU) were converted to equivalent BMD (mgHA/cm3) utilizing a linear calibration equation determined using asynchronous calibration, as previously validated.21

Figure 1.

Figure 1

Open in a new tab

Definition of subchondral trabecular bone regions of interest for quadrant-based analyses at the tibia. Principal axes were determined by applying principal component analysis (PCA) on the tibial bone mask. Morphological erosion was performed to exclude the cortical bone regions before landmarks were placed at the epiphyseal line to exclude bone distal to the epiphyseal line. Principal axes from PCA were used to split the resulting bone mask into quadrants for BMD analysis.

3D JSW and subchondral bone plate thickness

Triangulated 3D bone surface meshes, with uniformly sized triangles, were generated using the corrected bone masks from image segmentation. Methods described by Turmezei et al. were implemented to quantify the 3D JSW.17,30 From the vertices of each femoral mesh element, rays normal to the bone surface mesh were cast into the 3D image space to identify rays that intersected with the adjacent tibia bone surface mesh. For rays that intersected the tibial surface, their point of origin on the femoral bone surface was included in the joint space region for analysis. From each vertex within the joint space region, the grayscale image data were sampled along rays normal to the femoral bone surface to obtain intensity line profiles possessing 2 peaks, between which JSW was estimated as the full width at half-maximum distance.31

Accurate visualization of thin bone sections is dependent on the spatial resolution of the CT scanner. Specifically, image blurring will make thin, dense sections of bone appear thicker and less dense. By estimating the unblurred grayscale intensity of the bone, an accurate estimate of the bone thickness can be obtained by a deconvolution of the grayscale image data.32 As a result, we implemented cortical bone mapping, a model-based surface mapping technique, to obtain sub-voxel, SBP.Th estimates at the tibiofemoral joint surfaces.33 With this method, the grayscale image data were sampled along vectors normal to the bone surface model to obtain line intensity profiles across the cortical bone. An equation that modeled both the Gaussian blurring properties of the imaging system and the grayscale intensity (AU) variation across the cortical bone was fit to the intensity profile from each sample location. This provided preliminary cortical thickness estimates based on the maximum cortical density in each profile. In a second step, the maximum cortical density and thickness estimates from each sample location were pooled and used to estimate a global cortical density (Ct.BMD). For consistency, the same Ct.BMD estimation parameters and initial conditions were used for Ct.BMD estimation for all bones across all participants. In the final step, Ct.BMD was used to refine the local thickness estimates by accounting for the cortical density variations that arose due to partial volume effects that produce lower grayscale intensities (AU) in thin bone regions. The subchondral bone plate region of interest for each bone was extracted based on proximity to the opposing bone surface using JSW patches obtained using the described JSW approach. An overview of the estimation approaches is shown in Figure 2.

Figure 2.

Figure 2

Open in a new tab

Estimation of joint space width (JSW) using full width at half maximum distance between the two highest grayscale intensities across the joint space (top). Estimation of subchondral bone plate thickness (SBP.Th) by fitting a Gaussian-blur model to the grayscale intensity variation across the bone (bottom).

Secondary analyses

A best estimate of the tibiofemoral knee flexion angle was automatically calculated from anatomical bone axes derived using bone masks from image segmentation. Flexion angle estimates from scan–rescan images were used to investigate the sensitivity of 3D JSW measures to potential changes in knee flexion.

A sensitivity analysis was also conducted to investigate the effect of the Ct.BMD estimate on the SBP.Th measurements and repeatability. This was done by sequentially cropping the bone surface models in the axial direction. The effect of this simulation was to not only vary the amount of bone image data used to estimate Ct.BMD, but to also vary the locations, where the grayscale image data were sampled from (Figure S2). Additionally, we tested 4 different analysis methods to assess how each might affect the repeatability of SBP.Th measurements. For the primary analysis, no registration was performed and SBP.Th was estimated using unique Ct.BMD values and bone surface meshes derived from each scan–rescan image. For the secondary, sensitivity analyses, the following approaches were tested: (1) a common bone mesh was used to estimate a unique Ct.BMD for each scan–rescan image by transforming the bone mesh from one scan into the image space of the paired scan using mesh registration, (2) either the scan or rescan image was registered to the corresponding paired image after which a common bone mesh was used to estimate a unique Ct.BMD for each image, and (3) the same mesh registration approach was performed as in (1) except the Ct.BMD was fixed to the same value for scan–rescan images (Figure S3).

Quantitative outcomes

Bone and joint outcomes

The following bone and joint parameters were estimated for each scan: (1) mean BMD in each subchondral compartment of the proximal tibia, (2) mean SBP.Th at the articulating surface of each bone, and (3) mean JSW for each tibiofemoral compartment.

Statistical analyses

We followed procedures recommended for quantifying the unbiased short-term precision error of a measurement technique based on duplicate measures on each participant.34 The repeatability of BMD, JSW, and SBP.Th estimates were assessed using the 95% root-mean-square coefficient of variation (CVRMS%)34 and least significant change (LSC),35 which indicate the precision error and the least amount of change in each outcome metric that can be considered statistically significant, respectively. The sensitivity of JSW to potential changes in knee flexion between scan–rescan acquisitions were reported as absolute JSW change relative to changes in flexion angle. For SBP.Th, sensitivity was evaluated by comparing CVRMS% and LSC values after varying the Ct.BMD estimates using the described methods.

Results

Participant demographics are summarized in Table 1. Of the 37 participants whose knees were imaged, 5 had to be excluded because at least 1 of their scans showed excessive cone beam and/or motion artifacts in both knees. This left 32 participants for analysis, which still met the minimum sample size of 27 required for accurate assessment of precision errors based on duplicate measures on each participant.34 We arbitrarily chose to analyze 1 knee per participant, with the same knee being analyzed for both scan and re-scan images. For consistency, we considered the right knee for each participant, except for 1 participant where the left knee was analyzed, because artifacts were present in the right knee.

Table 1.

Participant demographics.

No. of individuals 32
No. of female 17
Age (yr) a 24.6 ± 6.5
Height (cm) a 171.9 ± 10.9
Weight (kg) a 73.2 ± 15.7
Open in a new tab
a

Data are mean ± SD.

Repeatability of BMD measurements

Mean subchondral trabecular BMD for the lateral anterior (LA), lateral posterior (LP), medial anterior (MA), and medial posterior (MP) tibia quadrants were 228.9, 324.6, 266.1, and 292.9 mgHA/cm3, respectively (Table 2). CVRMS% ranged from 1.87% to 2.85% in the LP and LA compartments, respectively. Least significant change was lowest in the LP compartment and highest in the MP compartment reaching values of 13.53 and 17.39 mgHA/cm3, respectively.

Table 2.

Scan–rescan repeatability of knee joint measurements.

Parameter No. of knees (right) Mean CVRMS%a LSCb
BMD (mgHA/cm 3 ) 32 (31)
 Lateral anterior 228.92 2.85 16.82
 Lateral posterior 324.58 1.87 13.53
 Medial anterior 266.11 2.39 16.83
 Medial posterior 292.85 2.46 17.39
JSW (mm) 32 (31)
 Medial 5.07 0.79 0.11
 Lateral 6.08 0.94 0.15
Femur SBP.Th (mm) 30 (29)
 Medial 0.34 10.55 0.10
 Lateral 0.30 11.29 0.09
Tibia SBP.Th (mm) 30 (29)
 Medial 0.76 9.75 0.20
 Lateral 0.80 11.05 0.26
Open in a new tab
a

Root mean square coefficient of variation.

b

Least significant change.

Repeatability of tibiofemoral JSW measurements

Group average values for mean JSW were 5.07 and 6.08 mm in the medial and lateral compartments, respectively. For mean medial JSW, the CVRMS% and LSC were 0.79% and 0.11 mm, while for mean lateral JSW, they were 0.94% and 0.15 mm, respectively (Table 2). A representative example of scan–rescan JSW measurements mapped onto the distal femoral surface is presented in Figure 3A. There was a trend suggesting that a positive change in knee flexion angle produced a decrease in average JSW for both the medial and lateral compartment; however, this effect was not significant (medial: R = −0.3, p = .09; lateral: R = −0.2, p = .26) (Figure 4).

Figure 3.

Figure 3

Open in a new tab

Representative example of paired 3-D JSW distribution maps from scan–rescan images plotted on the inferior surface of the distal femur (A). Representative examples of paired cortical thickness maps (B—distal femur, C—proximal tibia). Subchondral bone plate regions of interest were defined as the overlapping area from the JSW and cortical thickness maps and are shown traced in black.

Figure 4.

Figure 4

Open in a new tab

Correlation between scan–rescan changes in knee flexion angle and scan–rescan changes in JSW.

Repeatability of SBP.Th measurements

Subchondral bone plate thickness (SBP.Th) was estimated in the medial and lateral compartments of both the proximal tibia and distal femur (Figure 3B and C). Two bones were further excluded from both the tibia and femur datasets as the cortical bone mapping algorithm failed to converge to a reliable Ct.BMD estimate. For the femur, group average values for mean SBP.Th were 0.34 and 0.30 mm for medial and lateral compartment, respectively, with the highest femoral CVRMS% and LSC being reached in the lateral (11.29%) and medial (0.10 mm) compartments. Tibia group average values for mean SBP.Th were 0.76 and 0.80 mm for the medial and lateral compartment, respectively, with the highest tibial CVRMS% and LSC being reached in the lateral compartment (11.05% and 0.26 mm, respectively). When we modified the model parameters to obtain reliable Ct.BMD estimates for the four non-convergent bones that were originally excluded in the SBP.Th analysis, their inclusion did not change the overall repeatability (less than 1% increase in CVRMS%, no change in LSC) and their within-subject coefficients of variation (CVFem [0.031-0.277]; CVTib [0.027-0.239]) were similar to the convergent cases (CVFem [0.001-0.326]; CVTib [0.000-0.236]). Our results showed a maximum 40% change in SBP.Th based on differences in the cortical density estimate (Ct.BMD).

Our SBP.Th sensitivity analysis showed an inverse relationship between Ct.BMD and SBP.Th (Figure 5). Our simulation results demonstrated that including more of the thicker, diaphyseal cortical bone has a marked impact on Ct.BMD estimation and subsequent SBP.Th measures. Additionally, we found that different image analysis methods derived from distinct registration strategies (Figure S3), caused changes up to 8% and 0.08 mm for CVRMS% and LSC, respectively (Figure S4). Specifically, we found that (1) using a common bone surface mesh together with image registration or (2) bone surface mesh alignment together with fixing the Ct.BMD to a constant value for both scan–rescan images decreased both the CVRMS% and LSC for SBP.Th measures.

Figure 5.

Figure 5

Open in a new tab

Effect of global cortical density (Ct.BMD) estimate on mean subchondral bone thickness (SBP.Th) estimates for individual bones (left). Effect of Ct.BMD on mean SBP.Th expressed as percentages for pooled data (right). Data are shown for the medial compartment of four femurs and tibiae.

Discussion

In this study, we quantified the repeatability of subchondral trabecular BMD, JSW, and SBP.Th estimates at the knee using WBCT. We found good repeatability of BMD and JSW, along with moderate repeatability of SBP.Th estimates in a sample of younger healthy adults. For all parameters other than SBP.Th, CVRMS% was less than 3%. For BMD, LSC was less than 18 mgHA/cm3, while LSC for both JSW and SBP.Th was less than the isotropic voxel spacing (0.3 mm) of our reconstructed WBCT images.

Weight-bearing CT employs cone-beam CT technology, which is reported to have unreliable image grayscale values due to increased X-ray scatter and beam hardening artifacts36–39 that can adversely affect BMD estimates. Indeed, we have previously found measurement errors of up to 15% relative to reference phantom densities using WBCT.21 However, our tibia BMD analysis revealed repeatability values (CVRMS%) that were comparable to other tibia BMD studies conducted using different CT modalities.40,41 In comparison to a study by Burnett et al.,42 all our regional LSC values were lower than the minimum BMD change (44 mg/cm3) associated with a perceivable change in OA-related pain. Despite measurement errors which may compromise BMD measurement accuracy, our present results suggest that WBCT may be used to detect clinically meaningful changes in trabecular BMD.

Similar to previous studies evaluating 3D JSW repeatability at the knee, we found high repeatability of measures in both the medial and lateral tibiofemoral compartments. Segal et al. demonstrated high repeatability of 3D JSW using a nearest-neighbor, mesh-to-mesh distance mapping approach.26 Using the same dataset as Segal et al., Turmezei et al. found high repeatability using a joint space mapping technique that estimated JSW using variations in image grayscale intensities (AU) sampled across the joint space, with bone meshes serving only to inform the sampling direction.17,30 While we used the same patient positioning protocol as Segal et al. and implemented a similar joint space mapping approach to Turmezei et al., we applied them to newly acquired data and achieved excellent repeatability. Our results therefore demonstrate impressive robustness and reproducibility for 3D tibiofemoral JSW estimation from cone beam WBCT data. Although not statistically significant, the finding of an inverse relationship between knee flexion and JSW estimates was consistent with previous radiographic studies20,43 and underscores the potential influence of flexion differences for future studies. Consequently, we not only recommend future studies to utilize positioning devices or digital goniometers to ensure consistent knee flexion, but also the exploration of more poses to strengthen our understanding of flexion influences in WBCT knee assessment.

We quantified SBP.Th at the femur and tibia in both tibiofemoral compartments using an implementation of cortical bone mapping described by Treece et al.33 Our LSC results demonstrated sub-voxel precision across all compartments, suggesting that it is possible to detect thickness changes that are smaller than the isotropic voxel spacing of our WBCT images. While tibia LSC values were also sub-voxel, these were at least twice as large compared to femoral values and may reflect the thicker bone that is present at the tibial plateau. For comparison, Kroker et al. investigated SBP.Th differences in persons with unilateral anterior cruciate ligament reconstruction—a known risk factor for knee OA development—and found SBP.Th in the lateral femur of injured knees to be 0.1-0.14 mm thicker than uninjured contralateral/control knees.44 These femoral SBP.Th differences match very closely with femoral LSC values from our study, suggesting that the presented approach possesses adequate precision to detect differences in femoral SBP.Th using WBCT. Additionally, work by Mastbergen et al.25 showed SBP.Th thickness changes of up to 0.25 mm 1 yr after knee joint distraction—which are larger than the LSCs reported in this study. While these investigations were not in OA patients, who may experience less acute bone changes, these studies shed some light on the magnitude of SBP.Th changes that can be expected.

On average, our absolute JSW estimates were at least 6 times larger than our SBP.Th, which would partially explain the larger CVRMS% found for SBP.Th compared to JSW estimates. While our sensitivity simulation indicated that SBP.Th estimates could vary by up to 180%, our repeatability data only showed a maximum 40% change in SBP.Th based on differences in the cortical density (Ct.BMD) estimate (data not shown). Importantly, we found that improved SBP.Th repeatability was achieved when density estimates were derived using a common bone surface mesh, and further improved when the same Ct.BMD estimate was used for both scan and re-scan images. While using the same Ct.BMD for both scans had the best repeatability, this approach is more appropriate when Ct.BMD is assumed to remain unchanged between time points, limiting its clinical applications. Consequently, based on our results, we recommend a common mesh approach (registering one image to a paired reference image and using a common bone mesh or aligning one surface mesh to a reference mesh) together with a Ct.BMD estimate that is independently determined for each grayscale image. This ensures that the same bone regions are used for bone analyses from repeat scans. Additionally, we strongly recommend the implementation of methods, such as a scout view with reference line, to ensure the same bone and joint regions are captured within the imaging field of view for any given participant in future repeatability and longitudinal studies.

Our study benefited from a well-defined patient positioning protocol, comparable to that frequently used for radiographic knee imaging, allowing straightforward adoption in clinical settings. Additionally, we were able to identify factors that contribute toward measurement variability in JSW and bone thickness estimates in WBCT images. There are some limitations to our study. We limited our BMD analysis to the subchondral trabecular bone of the tibia where OA-related differences have previously been observed.42,45 Whereas quantifying long-term repeatability would determine whether measurement variability remains acceptable over a period that is clinically relevant for knee OA, we limited our analysis to short-term precision from same day, repeat scans. This was chosen to eliminate the influence of any joint changes that may occur over time and focus our assessment on the repeatability determined from the image acquisition and analysis methods.34 Second, while we only included healthy participants without a knee OA diagnosis and recognize that repeatability may vary with OA status, determining repeatability in healthy participants establishes baseline performance of the described methods and will be useful for monitoring early structural joint changes in individuals at risk of developing knee OA such as those with anterior cruciate ligament and meniscal injuries. This will aid in the clinical interpretation of findings from future studies which start with healthy joints at baseline. SBP.Th and JSW estimation accuracy at the knee was not quantified in this study. While both cortical bone mapping and joint space mapping at the hip have previously been validated by comparing estimates from conventional CT to those from HR-pQCT,30,32,33,46 we did not collect high resolution image data to conduct similar comparisons at the knee. Instead, we verified our cortical bone mapping implementation using synthetic CT image data comprised of shapes (spheres, plates, and cylinders) with known wall thicknesses (Figures S5 and S6). Thickness estimates in this data showed strong correlations (R2 > 0.99, p < .001) to the true wall thickness, with average biases that ranged from −0.149 to 0.065 mm depending on the isotropic voxel spacing of the image data. For JSW, previous work by Turmezei et al. indicated a slight overestimation (0.13 mm) by conventional CT (0.31 mm in-plane pixel spacing, 1.5 mm slice thickness) compared to estimates from HR-pQCT (0.082 mm isotropic voxel spacing) using a similar joint space mapping approach.30 Although both the verification of our cortical bone mapping implementation and the similarities in voxel spacing between our WBCT scans (0.3 mm) and the conventional CT scans used for JSW validation would suggest that our mapping approaches retained sufficient accuracy, further studies comparing WBCT-derived estimates with more established imaging modalities or measurement techniques are necessary. Finally, for JSW and SBP.Th, repeatability was assessed on estimates that were averaged over the 3D joint surface. As a result, these measures provide an overall sense of repeatability that may not reflect how repeatability might vary spatially across the joint surface.

Previous studies have demonstrated improved diagnostic performance of WBCT for detecting knee OA features15,16,47–49 and disease-related changes over time.18 However, as the role of WBCT in knee and OA assessment continues to evolve, repeatability and reproducibility studies are essential to evaluate our ability to monitor the effects of surgical interventions and disease progression. Additionally, because cone beam CT is subject to known artifacts—such as inconsistencies of grayscale values (AU) that may affect JSW and SBP.Th estimates obtained using the presented methods—further work is needed to verify their accuracy and agreement with estimates obtained from established 3D imaging modalities. Notably, despite implementing a well-defined positioning protocol, some scans had to be excluded highlighting 2 important considerations for WBCT image acquisition. First, the upright patient position increases the risk for motion compared to supine, fan beam CT. Second, unlike fan beam CT, motion during WBCT acquisition affects larger regions of the image due to cone beam geometry that captures a large volumetric stack at once.50 While a vertical thigh positioning plate helps to stabilize patients and minimize motion, image-based motion correction approaches51 could be explored in future studies. This study contributes to the limited research that currently exists on measurement repeatability in knee WBCT and will deepen our understanding of methods to reproduce scanning conditions and image analysis techniques over time.

Conclusion

Our results demonstrate repeatable objective quantification of multiple bone and joint parameters in 3D, with minimal variability between scans. Although further studies in OA cohorts are necessary, our results demonstrate repeatable BMD, JSW, and SBP.Th estimates made using WBCT, which when considered alongside previous work, suggests its potential to uncover the natural history and disease trajectory of knee OA through more sensitive disease stratification and progression monitoring. Additionally, our study explored alternative approaches to the cortical bone mapping technique and presented recommendations for repeatable bone and joint analyses in future knee WBCT studies. Importantly, SBP.Th repeatability was influenced by the amount of bone imaged within the field of view and subsequent model parameters utilized in the cortical bone mapping technique. This highlights the importance of controlling these sources of variation whenever possible, as this may affect the interpretation of bone thickness changes in longitudinal studies.

Supplementary Material

WBCT_Repeatability_Supplementary_Material-Revised_ziaf194
wbct_repeatability_supplementary_material-revised_ziaf194.docx (17.4MB, docx)

Acknowledgments

Additional support was provided by a Killam Visiting Scholar Award at the University of Calgary (D.D.A.).

Contributor Information

Tadiwa H Waungana, Department of Biomedical Engineering, University of Calgary, Calgary, AB T2N 4V8, Canada; McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Ria Mangat, McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Chloe Chen, McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Ainsley C J Smith, Department of Biomedical Engineering, University of Calgary, Calgary, AB T2N 4V8, Canada; McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Yousif Al-Khoury, Department of Biomedical Engineering, University of Calgary, Calgary, AB T2N 4V8, Canada; McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Michael T Kuczynski, McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Tom Turmezei, Department of Radiology, Norfolk and Norwich University Hospital, Norwich, NR4 7UY, United Kingdom; Norwich Medical School, University of East Anglia, Norwich, NR4 7TJ, United Kingdom.

Donald D Anderson, Department of Orthopedics and Rehabilitation, University of Iowa, Iowa City, IA 52242, United States.

William Brent Edwards, Department of Biomedical Engineering, University of Calgary, Calgary, AB T2N 4V8, Canada; McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada.

Sarah L Manske, Department of Biomedical Engineering, University of Calgary, Calgary, AB T2N 4V8, Canada; McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, AB T2N 4Z6, Canada; Department of Radiology, Cumming School of Medicine, University of Calgary, Calgary, AB T2N 4N1, Canada.

Author contributions

Tadiwa H. Waungana (Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing—original draft, Writing—review & editing), Ria Mangat (Investigation, Writing—review & editing), Chloe Chen (Investigation, Writing—review & editing), Ainsley C.J. Smith (Investigation, Writing—review & editing), Yousif Al-Khoury (Software, Writing—review & editing), Michael T. Kuczynski (Software, Writing—review & editing), Tom Turmezei (Conceptualization, Methodology, Validation, Writing—review & editing), Donald D. Anderson (Methodology, Writing—review & editing), W. Brent Edwards (Supervision, Writing—review & editing), and Sarah L. Manske (Conceptualization, Formal analysis, Project administration, Resources, Supervision, Validation, Writing—review & editing)

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Conflicts of interest

C.C. reports financial support was provided by Natural Sciences and Engineering Research Council of Canada. A.C.J. Smith reports financial support was provided by Canadian Institutes of Health Research. Y.A.-K. reports financial support was provided by Natural Sciences and Engineering Research Council of Canada. S.L.M. reports financial support was provided by Natural Sciences and Engineering Research Council of Canada. S.L.M. reports financial support was provided by Canada Foundation for Innovation. T.T. reports a relationship with CurveBeam AI that includes: consulting or advisory. T.T. reports a relationship with Computed Tomography in Osteoarthritis Research (OCTA) that includes: board membership. T.T. reports a relationship with International Society of Osteoarthritis Imaging that includes: board membership. T.T. reports a relationship with Chondrometrics GmbH that includes: board membership and travel reimbursement. T.T. reports a relationship with Osteoarthritis Research Society International that includes: travel reimbursement. T.T. reports a relationship with International Weight Bearing CT Society that includes: board membership. T.T. reports a relationship with KNEE3D Ltd. that includes: board membership and equity or stocks. Co-author previously received royalties or licenses from Elsevier Ltd.—T.T. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

S.L.M. holds the position of Guest Editor for JBMR Plus and has been recused from reviewing or making decisions for the manuscript.

Data availability

Data will be made available upon reasonable request. Image analysis code is provided at https://github.com/ManskeLab/WBCT_Repeatability.

References

  • 1. Hunter  DJ, Bierma-Zeinstra  S. Osteoarthritis. Lancet. 2019; 393(10182):1745–1759. 10.1016/S0140-6736(19)30417-9 [DOI] [PubMed] [Google Scholar]
  • 2. Vos  T, Allen  C, Arora  M, et al.  Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet. 2016;388(10053):1545–1602. 10.1016/S0140-6736(16)31678-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Altman  RD, Gold  GE. Atlas of individual radiographic features in osteoarthritis, revised. Osteoarthr Cartil. 2007;15(Suppl A):A1–A56. 10.1016/j.joca.2006.11.009 [DOI] [Google Scholar]
  • 4. Kellgren  JH, Lawrence  JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis. 1957;16(4):494–502. 10.1136/ard.16.4.494 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Zhu  X, Chan  YT, Yung  PSH, Tuan  RS, Jiang  Y. Subchondral bone remodeling: a therapeutic target for osteoarthritis. Front Cell Dev Biol. 2021;8:607764. 10.3389/fcell.2020.607764 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Goldring  SR, Goldring  MB. Changes in the osteochondral unit during osteoarthritis: structure, function and cartilage-bone crosstalk. Nat Rev Rheumatol. 2016;12(11):632–644. 10.1038/nrrheum.2016.148 [DOI] [PubMed] [Google Scholar]
  • 7. Burr  DB, Gallant  MA. Bone remodelling in osteoarthritis. Nat Rev Rheumatol. 2012;8(11):665–673. 10.1038/nrrheum.2012.130 [DOI] [PubMed] [Google Scholar]
  • 8. Turmezei  TD, Low  SB, Rupret  S, et al.  Multiparametric 3-D analysis of bone and joint space width at the knee from weight bearing computed tomography. Osteoarthr Imaging. 2022;2(2):100069. 10.1016/j.ostima.2022.100069 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Roemer  FW, Demehri  S, Omoumi  P, et al.  State of the art: imaging of osteoarthritis—revisited 2020. Radiology. 2020;296(1):5–21. 10.1148/radiol.2020192498 [DOI] [PubMed] [Google Scholar]
  • 10. Guermazi  A, Hayashi  D, Roemer  FW, Felson  DT. Osteoarthritis: a review of strengths and weaknesses of different imaging options. Rheum Dis Clin North Am. 2013;39(3):567–591. 10.1016/j.rdc.2013.02.001 [DOI] [PubMed] [Google Scholar]
  • 11. Sakellariou  G, Conaghan  PG, Zhang  W, et al.  EULAR recommendations for the use of imaging in the clinical management of peripheral joint osteoarthritis. Ann Rheum Dis. 2017;76(9):1484–1494. 10.1136/annrheumdis-2016-210815 [DOI] [PubMed] [Google Scholar]
  • 12. Vignon  E. Radiographic issues in imaging the progression of hip and knee osteoarthritis. J Rheumatol. 2004;31(Suppl 70):36-44. Available: www.jrheum.org [Google Scholar]
  • 13. Thawait  GK, Demehri  S, AlMuhit  A, et al.  Extremity cone-beam CT for evaluation of medial tibiofemoral osteoarthritis: initial experience in imaging of the weight-bearing and non-weight-bearing knee. Eur J Radiol. 2015;84(12):2564–2570. 10.1016/j.ejrad.2015.09.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Tuominen  EKJ, Kankare  J, Koskinen  SK, Mattila  KT. Weight-bearing CT imaging of the lower extremity. AJR Am J Roentgenol. 2013;200(1):146–148. 10.2214/AJR.12.8481 [DOI] [PubMed] [Google Scholar]
  • 15. Segal  NA, Frick  E, Duryea  J, et al.  Comparison of tibiofemoral joint space width measurements from standing CT and fixed flexion radiography. J Orthop Res. 2017;35(7):1388–1395. 10.1002/jor.23387 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Segal  NA, Frick  E, Duryea  J, et al.  Correlations of medial joint space width on fixed-flexed standing computed tomography and radiographs with cartilage and meniscal morphology on magnetic resonance imaging. Arthritis Care Res (Hoboken). 2016;68(10):1410–1416. 10.1002/acr.22888 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Turmezei  TD, B. Low  S, Rupret  S, et al.  Quantitative three-dimensional assessment of knee joint space width from weight-bearing CT. Radiology. 2021;299(3):649–659. 10.1148/radiol.2021203928 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Segal  NA, Nevitt  MC, Morales Aquino  M, et al.  Improved responsiveness to change in joint space width over 24-month follow-up: comparison of 3D JSW on weight-bearing CT vs 2D JSW on radiographs in the MOST study. Osteoarthr Cartil. 2023;31(3):406–413. 10.1016/j.joca.2022.12.002 [DOI] [Google Scholar]
  • 19. Brinch  S, Wellenberg  RHH, Boesen  MP, et al.  Weight-bearing cone-beam CT: the need for standardised acquisition protocols and measurements to fulfill high expectations—a review of the literature. Skeletal Radiol. 2023;52(6):1073–1088. 10.1007/s00256-022-04223-1 [DOI] [PubMed] [Google Scholar]
  • 20. Piperno  M, Le Graverand  MPH, Conrozier  T, Bochu  M, Mathieu  P, Vignon  E. Quantitative evaluation of joint space width in femorotibial osteoarthritis: comparison of three radiographic views. Osteoarthr Cartil. 1998;6(4):252–259. 10.1053/joca.1998.0118 [DOI] [Google Scholar]
  • 21. Waungana  TH, Qiu  K, Tse  JJ, et al.  Accuracy of volumetric bone mineral density measurement in weight bearing, cone beam computed tomography. J Clin Densitom. 2024;27(3):101504. 10.1016/j.jocd.2024.101504 [DOI] [PubMed] [Google Scholar]
  • 22. Turunen  MJ, Töyras  J, Kokkonen  HT, Jurvelin  JS. Quantitative evaluation of knee subchondral bone mineral density using cone beam computed tomography. IEEE Trans Med Imaging. 2015;34(10):2186–2190. 10.1109/TMI.2015.2426684 [DOI] [PubMed] [Google Scholar]
  • 23. Zbijewski  W, de Jean  P, Prakash  P, et al.  A dedicated cone-beam CT system for musculoskeletal extremities imaging: design, optimization, and initial performance characterization. Med Phys. 2011;38(8):4700–4713. 10.1118/1.3611039 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Siewerdsen  JH, Jaffray  DA. Cone-beam computed tomography with a flat-panel imager: magnitude and effects of X-ray scatter. Med Phys. 2001;28(2):220–231. 10.1118/1.1339879 [DOI] [PubMed] [Google Scholar]
  • 25. Mastbergen  SC, Ooms  A, Turmezei  TD, et al.  Subchondral bone changes after joint distraction treatment for end stage knee osteoarthritis. Osteoarthr Cartil. 2022;30(7):965–972. 10.1016/j.joca.2021.12.014 [DOI] [Google Scholar]
  • 26. Segal  NA, Bergin  J, Kern  A, Findlay  C, Anderson  DD. Test–retest reliability of tibiofemoral joint space width measurements made using a low-dose standing CT scanner. Skeletal Radiol. 2017;46(2):217–222. 10.1007/s00256-016-2539-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Hellio Le Graverand  MP, Mazzuca  S, Duryea  J, Brett  A. Radiographic-based grading methods and radiographic measurement of joint space width in osteoarthritis. Radiol Clin North Am. 2009;47(4):567–579. 10.1016/j.rcl.2009.04.004 [DOI] [PubMed] [Google Scholar]
  • 28. Besler  BA, Michalski  AS, Kuczynski  MT, Abid  A, Forkert  ND, Boyd  SK. Bone and joint enhancement filtering: application to proximal femur segmentation from uncalibrated computed tomography datasets. Med Image Anal. 2021;67:101887. 10.1016/j.media.2020.101887 [DOI] [Google Scholar]
  • 29. Sekhon  K, Kazakia  GJ, Burghardt  AJ, Hermannsson  B, Majumdar  S. Accuracy of volumetric bone mineral density measurement in high-resolution peripheral quantitative computed tomography. Bone. 2009;45(3):473–479. 10.1016/j.bone.2009.05.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Turmezei  TD, Treece  GM, Gee  AH, Houlden  R, Poole  KES. A new quantitative 3D approach to imaging of structural joint disease. Sci Rep. 2018;8(1):9280. 10.1038/s41598-018-27486-y [DOI] [Google Scholar]
  • 31. McFadden  E, Dibbern  K, Segal  N, Lynch  J, Felson  T, Anderson  D. Fully automated 3D joint space width analysis of the tibiofemoral joint from weight bearing CT. Osteoarthr Cartil. 2022;30(Suppl 1):S284–S285. 10.1016/j.joca.2022.02.382 [DOI] [Google Scholar]
  • 32. Treece  GM, Gee  AH, Mayhew  PM, Poole  KES. High resolution cortical bone thickness measurement from clinical CT data. Med Image Anal. 2010;14(3):276–290. 10.1016/j.media.2010.01.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Treece  GM, Poole  KES, Gee  AH. Imaging the femoral cortex: thickness, density and mass from clinical CT. Med Image Anal. 2012;16(5):952–965. 10.1016/j.media.2012.02.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Glüer  C-C, Blake  G, Lu  Y, Blunt  BA, Jergas  M, Genant  HK. Accurate assessment of precision errors: how to measure the reproducibility of bone densitometry techniques. Osteoporos Int. 1995;5(4):262-270. [Google Scholar]
  • 35. Baim  S, Wilson  CR, Lewiecki  EM, Luckey  MM, Downs  RW, Lentle  BC. Precision assessment and radiation safety for dual-energy X-ray absorptiometry position paper of the International Society for Clinical Densitometry. J Clin Densitom. 2005;8(4):371–378. 10.1220/1094-6950 [DOI] [PubMed] [Google Scholar]
  • 36. Parsa  A, Ibrahim  N, Hassan  B, van der Stelt  P, Wismeijer  D. Bone quality evaluation at dental implant site using multislice CT, micro-CT, and cone beam CT. Clin Oral Implants Res. 2015;26(1):e1–e7. 10.1111/clr.12315 [DOI] [Google Scholar]
  • 37. Selvaraj  A, Jain  RK, Nagi  R, Balasubramaniam  A. Correlation between gray values of cone-beam computed tomograms and Hounsfield units of computed tomograms: a systematic review and meta-analysis. Imaging Sci Dent. 2022;52(2):133–140. 10.5624/isd.20210274 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Becker  AE, Hernandez  AM, Schwoebel  PR, Boone  JM. Cone beam CT multisource configurations: evaluating image quality, scatter, and dose using phantom imaging and Monte Carlo simulations. Phys Med Biol. 2020;65(23):235032. 10.1088/1361-6560/abc306 [DOI] [Google Scholar]
  • 39. Katsumata  A, Hirukawa  A, Okumura  S, et al.  Relationship between density variability and imaging volume size in cone-beam computerized tomographic scanning of the maxillofacial region: an in vitro study. Oral Surg Oral Med Oral Pathol Oral Radiol Endod. 2009;107(3):420–425. 10.1016/j.tripleo.2008.05.049 [DOI] [PubMed] [Google Scholar]
  • 40. Johnston  JD, McLennan  CE, Hunter  DJ, Wilson  DR. In vivo precision of a depth-specific topographic mapping technique in the CT analysis of osteoarthritic and normal proximal tibial subchondral bone density. Skeletal Radiol. 2011;40(8):1057–1064. 10.1007/s00256-010-1001-6 [DOI] [PubMed] [Google Scholar]
  • 41. Sievänen  H, Koskue  V, Rauhio  A, Kannus  P, Heinonen  A, Vuori  I. Peripheral quantitative computed tomography in human long bones: evaluation of in vitro and in vivo precision. J Bone Miner Res. 1998;13(5):871–882. 10.1359/jbmr.1998.13.5.871 [DOI] [PubMed] [Google Scholar]
  • 42. Burnett  WD, Kontulainen  SA, McLennan  CE, et al.  Proximal tibial trabecular bone mineral density is related to pain in patients with osteoarthritis. Arthritis Res Ther. 2017;19(1):200. 10.1186/s13075-017-1415-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43. Vignon  E, Piperno  M, Le Graverand  MP, et al.  Measurement of radiographic joint space width in the tibiofemoral compartment of the osteoarthritic knee: comparison of standing anteroposterior and Lyon schuss views. Arthritis Rheum. 2003;48(2):378–384. 10.1002/art.10773 [DOI] [PubMed] [Google Scholar]
  • 44. Kroker  A, Bhatla  JL, Emery  CA, Manske  SL, Boyd  SK. Subchondral bone microarchitecture in ACL reconstructed knees of young women: a comparison with contralateral and uninjured control knees. Bone. 2018;111:1–8. 10.1016/j.bone.2018.03.006 [DOI] [PubMed] [Google Scholar]
  • 45. Burnett  WD, Kontulainen  SA, McLennan  CE, et al.  Knee osteoarthritis patients with severe nocturnal pain have altered proximal tibial subchondral bone mineral density. Osteoarthr Cartil. 2015;23(9):1483–1490. 10.1016/j.joca.2015.04.012 [DOI] [Google Scholar]
  • 46. Treece  GM, Gee  AH. Independent measurement of femoral cortical thickness and cortical bone density using clinical CT. Med Image Anal. 2015;20(1):249–264. 10.1016/j.media.2014.11.012 [DOI] [PubMed] [Google Scholar]
  • 47. Segal  NA, Nevitt  MC, Lynch  JA, Niu  J, Torner  JC, Guermazi  A. Diagnostic performance of 3D standing CT imaging for detection of knee osteoarthritis features. Phys Sportsmed. 2015;43(3):213–220. 10.1080/00913847.2015.1074854 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Segal  NA, Murphy  MT, Everist  BM, et al.  Clinical value of weight-bearing CT and radiographs for detecting patellofemoral cartilage visualized by MRI in the MOST study. Osteoarthr Cartil. 2021;29(11):1540–1548. 10.1016/j.joca.2021.07.011 [DOI] [Google Scholar]
  • 49. Kaneda  K, Harato  K, Oki  S, et al.  Increase in tibial internal rotation due to weight-bearing is a key feature to diagnose early-stage knee osteoarthritis: a study with upright computed tomography. BMC Musculoskelet Disord. 2022;23(1):253. 10.1186/s12891-022-05190-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50. Jacobson  MW. In: Sarment  D, ed. Cone Beam Computed Tomography: Oral and Maxillofacial Diagnosis and Applications. Ames, Iowa: John Wiley & Sons, Incorporated; 2013:ch. 1. [Google Scholar]
  • 51. Berger  M, Müller  K, Aichert  A, et al.  Marker-free motion correction in weight-bearing cone-beam CT of the knee joint. Med Phys. 2016;43(3):1235–1248. 10.1118/1.4941012 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

WBCT_Repeatability_Supplementary_Material-Revised_ziaf194
wbct_repeatability_supplementary_material-revised_ziaf194.docx (17.4MB, docx)

Data Availability Statement

Data will be made available upon reasonable request. Image analysis code is provided at https://github.com/ManskeLab/WBCT_Repeatability.

Articles from JBMR Plus are provided here courtesy of Oxford University Press

RESOURCES