Combining Register and Radiological Visits Data Allows to Reliably Identify Incident Wrist Fractures.

Purpose: To evaluate how comprehensively wrist fractures can be tracked from the national medical registers, and to propose a method for complementing the register data using time stamps of wrist radiography visits recorded in the radiological image archive. Patients and Methods: For the Kuopio Osteoporosis Risk Factor and Prevention Study (OSTPRE) cohort of 14220 post-menopausal women, we analysed the data from the Care Register for Health Care, Register for Primary Health Care Visits, self-reports, radiological image archive PACS, and patient records to identify the wrist fractures occurred between 2011 and 2021. Using this gold standard of fractures, we validated the coverage of the registers and image archive and created algorithms to automatically identify fracture events from the registers and/or metadata of wrist radiography visits. Results: We show that wrist fractures cannot be comprehensively identified based on national registers. To remedy this, our proposed method of combining register and image archive data can lift the coverage from 81% to 94% and reduce false discoveries from 6% to 2%. Conclusion: The proposed method offers a more reliable way of gathering fracture information. Comprehensive fracture identification is essential in many research settings, such as incidence statistics, prevention studies, and risk assessment models.

3D Finite Element Models Reconstructed From 2D Dual-Energy X-Ray Absorptiometry (DXA) Images Improve Hip Fracture Prediction Compared to Areal BMD in Osteoporotic Fractures in Men (MrOS) Sweden Cohort.

Bone strength is an important contributor to fracture risk. Areal bone mineral density (aBMD) derived from dual-energy X-ray absorptiometry (DXA) is used as a surrogate for bone strength in fracture risk prediction tools. 3D finite element (FE) models predict bone strength better than aBMD, but their clinical use is limited by the need for 3D computed tomography and lack of automation. We have earlier developed a method to reconstruct the 3D hip anatomy from a 2D DXA image, followed by subject-specific FE-based prediction of proximal femoral strength. In the current study, we aim to evaluate the method's ability to predict incident hip fractures in a population-based cohort (Osteoporotic Fractures in Men [MrOS] Sweden). We defined two subcohorts: (i) hip fracture cases and controls cohort: 120 men with a hip fracture (<10 years from baseline) and two controls to each hip fracture case, matched by age, height, and body mass index; and (ii) fallers cohort: 86 men who had fallen the year before their hip DXA scan was acquired, 15 of which sustained a hip fracture during the following 10 years. For each participant, we reconstructed the 3D hip anatomy and predicted proximal femoral strength in 10 sideways fall configurations using FE analysis. The FE-predicted proximal femoral strength was a better predictor of incident hip fractures than aBMD for both hip fracture cases and controls (difference in area under the receiver operating characteristics curve, ΔAUROC = 0.06) and fallers (ΔAUROC = 0.22) cohorts. This is the first time that FE models outperformed aBMD in predicting incident hip fractures in a population-based prospectively followed cohort based on 3D FE models obtained from a 2D DXA scan. Our approach has potential to notably improve the accuracy of fracture risk predictions in a clinically feasible manner (only one single DXA image is needed) and without additional costs compared to the current clinical approach. © 2023 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).

Statistical Shape and Appearance Models: Development Towards Improved Osteoporosis Care.

PURPOSE OF REVIEW: Statistical models of shape and appearance have increased their popularity since the 1990s and are today highly prevalent in the field of medical image analysis. In this article, we review the recent literature about how statistical models have been applied in the context of osteoporosis and fracture risk estimation. RECENT FINDINGS: Recent developments have increased their ability to accurately segment bones, as well as to perform 3D reconstruction and classify bone anatomies, all features of high interest in the field of osteoporosis and fragility fractures diagnosis, prevention, and treatment. An increasing number of studies used statistical models to estimate fracture risk in retrospective case-control cohorts, which is a promising step towards future clinical application. All the reviewed application areas made considerable steps forward in the past 5-6 years. Heterogeneities in validation hinder a thorough comparison between the different methods and represent one of the future challenges to be addressed to reach clinical implementation.

Detecting pathological features and predicting fracture risk from dual-energy X-ray absorptiometry images using deep learning.

Dual-energy X-ray absorptiometry (DXA) is the gold standard imaging method for diagnosing osteoporosis in clinical practice. The DXA images are commonly used to estimate a numerical value for bone mineral density (BMD), which decreases in osteoporosis. Low BMD is a known risk factor for osteoporotic fractures. In this study, we used deep learning to identify lumbar scoliosis and structural abnormalities that potentially affect BMD but are often neglected in lumbar spine DXA analysis. In addition, we tested the approach's ability to predict fractures using only DXA images. A dataset of 2949 images gathered by Kuopio Osteoporosis Risk Factor and Prevention Study was used to train a convolutional neural network (CNN) for classification. The model was able to classify scoliosis with an AUC of 0.96 and structural abnormalities causing unreliable BMD measurement with an AUC of 0.91. It predicted fractures occurring within 5 years from the lumbar spine DXA scan with an AUC of 0.63, meeting the predictive performance of combined BMD measurements from the lumbar spine and hip. In an independent test set of 574 clinical patients, AUC for lumbar scoliosis was 0.93 and AUC for unreliable BMD measurements was 0.94. In each classification task, neural network visualizations indicated the model's predictive strategy. We conclude that deep learning could complement the well established DXA method for osteoporosis diagnostics by analyzing incidental findings and image reliability, and improve its predictive ability in the future.

Validation of 3D finite element models from simulated DXA images for biofidelic simulations of sideways fall impact to the hip.

Computed tomography (CT)-derived finite element (FE) models have been proposed as a tool to improve the current clinical assessment of osteoporosis and personalized hip fracture risk by providing an accurate estimate of femoral strength. However, this solution has two main drawbacks, namely: (i) 3D CT images are needed, whereas 2D dual-energy x-ray absorptiometry (DXA) images are more generally available, and (ii) quasi-static femoral strength is predicted as a surrogate for fracture risk, instead of predicting whether a fall would result in a fracture or not. The aim of this study was to combine a biofidelic fall simulation technique, based on 3D computed tomography (CT) data with an algorithm that reconstructs 3D femoral shape and BMD distribution from a 2D DXA image. This approach was evaluated on 11 pelvis-femur constructs for which CT scans, ex vivo sideways fall impact experiments and CT-derived biofidelic FE models were available. Simulated DXA images were used to reconstruct the 3D shape and bone mineral density (BMD) distribution of the left femurs by registering a projection of a statistical shape and appearance model with a genetic optimization algorithm. The 2D-to-3D reconstructed femurs were meshed, and the resulting FE models inserted into a biofidelic FE modeling pipeline for simulating a sideways fall. The median 2D-to-3D reconstruction error was 1.02 mm for the shape and 0.06 g/cm3 for BMD for the 11 specimens. FE models derived from simulated DXAs predicted the outcome of the falls in terms of fracture versus non-fracture with the same accuracy as the CT-derived FE models. This study represents a milestone towards improved assessment of hip fracture risk based on widely available clinical DXA images.

Elucidating failure mechanisms in human femurs during a fall to the side using bilateral digital image correlation.

An improved understanding of the mechanical properties of human femurs is a milestone towards a more accurate assessment of fracture risk. Digital image correlation (DIC) has recently been adopted to provide full-field strain measurements during mechanical testing of femurs. However, it has typically been used to measure strains on the anterior side of the femur, whereas in both single-leg-stance and sideways fall loading conditions, the highest deformations result on the medial and lateral sides of the femoral neck. The goal of this study was to measure full-field deformations simultaneously on the medial and lateral side of the femoral neck in a configuration resembling a fall to the side. Twelve female cadaver femurs were prepared for DIC measurements and tested in sideways fall at 5 mm/s displacement rate. Two pairs of cameras recorded the medial and lateral side of the femoral neck, and deformations were calculated using DIC. The samples exhibited a two-stage failure: first, a compressive collapse on the superolateral side of the femoral neck in conjunction with peak force, followed by complete femoral neck fracture at the force drop following the post-elastic phase. DIC measurements corroborated this observation by reporting no tensile strains above yield limit for the medial side of the neck up to peak force. DIC measurements registered onto the bone micro-architecture showed strain localizations in proximity of cortical pores due to, for instance, blood vessels. This could explain previously reported discrepancies between simulations and experiments in regions rich with large pores, like the superolateral femoral neck.

Clinical Contrast-Enhanced Computed Tomography With Semi-Automatic Segmentation Provides Feasible Input for Computational Models of the Knee Joint.

Computational models can provide information on joint function and risk of tissue failure related to progression of osteoarthritis (OA). Currently, the joint geometries utilized in modeling are primarily obtained via manual segmentation, which is time-consuming and hence impractical for direct clinical application. The aim of this study was to evaluate the applicability of a previously developed semi-automatic method for segmenting tibial and femoral cartilage to serve as input geometry for finite element (FE) models. Knee joints from seven volunteers were first imaged using a clinical computed tomography (CT) with contrast enhancement and then segmented with semi-automatic and manual methods. In both segmentations, knee joint models with fibril-reinforced poroviscoelastic (FRPVE) properties were generated and the mechanical responses of articular cartilage were computed during physiologically relevant loading. The mean differences in the absolute values of maximum principal stress, maximum principal strain, and fibril strain between the models generated from semi-automatic and manual segmentations were <1 MPa, <0.72% and <0.40%, respectively. Furthermore, contact areas, contact forces, average pore pressures, and average maximum principal strains were not statistically different between the models (p >0.05). This semi-automatic method speeded up the segmentation process by over 90% and there were only negligible differences in the results provided by the models utilizing either manual or semi-automatic segmentations. Thus, the presented CT imaging-based segmentation method represents a novel tool for application in FE modeling in the clinic when a physician needs to evaluate knee joint function.

Automated segmentation of cortical and trabecular bone to generate finite element models for femoral bone mechanics.

Finite element (FE) models based on quantitative computed tomography (CT) images are better predictors of bone strength than conventional areal bone mineral density measurements. However, FE models require manual segmentation of the femur, which is not clinically applicable. This study developed a method for automated FE analyses from clinical CT images. Clinical in-vivo CT images of 13 elderly female subjects were collected to evaluate the method. Secondly, proximal cadaver femurs were harvested and imaged with clinical CT (N = 17). Of these femurs, 14 were imaged with µCT and three had earlier been tested experimentally in stance-loading, while collecting surface deformations with digital image correlation. Femurs were segmented from clinical CT images using an automated method, based on the segmentation tool Stradwin. The method automatically distinguishes trabecular and cortical bone, corrects partial volume effect and generates input for FE analysis. The manual and automatic segmentations agreed within about one voxel for in-vivo subjects (0.99 ±â€¯0.23 mm) and cadaver femurs (0.21 ±â€¯0.07 mm). The strains from the FE predictions closely matched with the experimentally measured strains (R2 = 0.89). The method can automatically generate meshes suitable for FE analysis. The method may bring us one step closer to enable clinical usage of patient-specific FE analyses.

Correction to: Prediction of femoral strength using 3D finite element models reconstructed from DXA images: validation against experiments.

In the original publication of the article, Fig. 3 and Tables 2, 4 and 5 were published with errors. The issue was caused by an error in the code used to predict femoral strength in the finite element (FE) models.

Method for Segmentation of Knee Articular Cartilages Based on Contrast-Enhanced CT Images.

Segmentation of contrast-enhanced computed tomography (CECT) images enables quantitative evaluation of morphology of articular cartilage as well as the significance of the lesions. Unfortunately, automatic segmentation methods for CECT images are currently lacking. Here, we introduce a semiautomated technique to segment articular cartilage from in vivo CECT images of human knee. The segmented cartilage geometries of nine knee joints, imaged using a clinical CT-scanner with an intra-articular contrast agent, were compared with manual segmentations from CT and magnetic resonance (MR) images. The Dice similarity coefficients (DSCs) between semiautomatic and manual CT segmentations were 0.79-0.83 and sensitivity and specificity values were also high (0.76-0.86). When comparing semiautomatic and manual CT segmentations, mean cartilage thicknesses agreed well (intraclass correlation coefficient = 0.85-0.93); the difference in thickness (mean ± SD) was 0.27 ± 0.03 mm. Differences in DSC, when MR segmentations were compared with manual and semiautomated CT segmentations, were statistically insignificant. Similarly, differences in volume were not statistically significant between manual and semiautomatic CT segmentations. Semiautomation decreased the segmentation time from 450 ± 190 to 42 ± 10 min per joint. The results reveal that the proposed technique is fast and reliable for segmentation of cartilage. Importantly, this is the first study presenting semiautomated segmentation of cartilage from CECT images of human knee joint with minimal user interaction.

Dual Contrast CT Method Enables Diagnostics of Cartilage Injuries and Degeneration Using a Single CT Image.

Cartilage injuries may be detected using contrast-enhanced computed tomography (CECT) by observing variations in distribution of anionic contrast agent within cartilage. Currently, clinical CECT enables detection of injuries and related post-traumatic degeneration based on two subsequent CT scans. The first scan allows segmentation of articular surfaces and lesions while the latter scan allows evaluation of tissue properties. Segmentation of articular surfaces from the latter scan is difficult since the contrast agent diffusion diminishes the image contrast at surfaces. We hypothesize that this can be overcome by mixing anionic contrast agent (ioxaglate) with bismuth oxide nanoparticles (BINPs) too large to diffuse into cartilage, inducing a high contrast at the surfaces. Here, a dual contrast method employing this mixture is evaluated by determining the depth-wise X-ray attenuation profiles in intact, enzymatically degraded, and mechanically injured osteochondral samples (n = 3 × 10) using a microCT immediately and at 45 min after immersion in contrast agent. BiNPs were unable to diffuse into cartilage, producing high contrast at articular surfaces. Ioxaglate enabled the detection of enzymatic and mechanical degeneration. In conclusion, the dual contrast method allowed detection of injuries and degeneration simultaneously with accurate cartilage segmentation using a single scan conducted at 45 min after contrast agent administration.

In Vivo Contrast-Enhanced Cone Beam CT Provides Quantitative Information on Articular Cartilage and Subchondral Bone.

In post-traumatic osteoarthritis, both articular cartilage and subchondral bone undergo characteristic pathological changes. This study investigates potential of delayed cone beam computed tomography arthrography (dCBCTa) to simultaneously detect variations in cartilage and subchondral bone. The knees of patients (n = 17) with suspected joint injuries were imaged using a clinical CBCT scanner at 5 and 45 min after the intra-articular injection of anionic contrast agent (Hexabrix™) with hydroxyapatite phantoms around the knee. Normalized attenuation (i.e., contrast agent partition, an indicator of tissue composition) in cartilage, bone mineral density (BMD) in subchondral bone plate (SBP), subchondral bone and trabecular bone, and thicknesses of SBP and cartilage were determined. Lesions of cartilage were scored using International Cartilage Repair Society (ICRS) grading. Normalized attenuation in the delayed image (t = 45 min) increased along the increase of ICRS grade (p = 0.046). Moreover, BMD was significantly higher in SBPs under damaged cartilage (ICRS = 1-2 or ICRS ≥ 3; p = 0.047 and p = 0.038, respectively) than in SBP under non-injured tissue (ICRS = 0). For the first time, dCBCTa enabled the detection of articular cartilage injuries and subchondral bone alterations simultaneously in vivo. Significant relations between ICRS grading and both cartilage and bone parameters suggest that dCBCTa has potential for quantitative imaging of the knee joint.

Prediction of femoral strength using 3D finite element models reconstructed from DXA images: validation against experiments.

Computed tomography (CT)-based finite element (FE) models may improve the current osteoporosis diagnostics and prediction of fracture risk by providing an estimate for femoral strength. However, the need for a CT scan, as opposed to the conventional use of dual-energy X-ray absorptiometry (DXA) for osteoporosis diagnostics, is considered a major obstacle. The 3D shape and bone mineral density (BMD) distribution of a femur can be reconstructed using a statistical shape and appearance model (SSAM) and the DXA image of the femur. Then, the reconstructed shape and BMD could be used to build FE models to predict bone strength. Since high accuracy is needed in all steps of the analysis, this study aimed at evaluating the ability of a 3D FE model built from one 2D DXA image to predict the strains and fracture load of human femora. Three cadaver femora were retrieved, for which experimental measurements from ex vivo mechanical tests were available. FE models were built using the SSAM-based reconstructions: using only the SSAM-reconstructed shape, only the SSAM-reconstructed BMD distribution, and the full SSAM-based reconstruction (including both shape and BMD distribution). When compared with experimental data, the SSAM-based models predicted accurately principal strains (coefficient of determination >0.83, normalized root-mean-square error <16%) and femoral strength (standard error of the estimate 1215 N). These results were only slightly inferior to those obtained with CT-based FE models, but with the considerable advantage of the models being built from DXA images. In summary, the results support the feasibility of SSAM-based models as a practical tool to introduce FE-based bone strength estimation in the current fracture risk diagnostics.

How accurately can subject-specific finite element models predict strains and strength of human femora? Investigation using full-field measurements.

Subject-specific finite element models have been proposed as a tool to improve fracture risk assessment in individuals. A thorough laboratory validation against experimental data is required before introducing such models in clinical practice. Results from digital image correlation can provide full-field strain distribution over the specimen surface during in vitro test, instead of at a few pre-defined locations as with strain gauges. The aim of this study was to validate finite element models of human femora against experimental data from three cadaver femora, both in terms of femoral strength and of the full-field strain distribution collected with digital image correlation. The results showed a high accuracy between predicted and measured principal strains (R(2)=0.93, RMSE=10%, 1600 validated data points per specimen). Femoral strength was predicted using a rate dependent material model with specific strain limit values for yield and failure. This provided an accurate prediction (<2% error) for two out of three specimens. In the third specimen, an accidental change in the boundary conditions occurred during the experiment, which compromised the femoral strength validation. The achieved strain accuracy was comparable to that obtained in state-of-the-art studies which validated their prediction accuracy against 10-16 strain gauge measurements. Fracture force was accurately predicted, with the predicted failure location being very close to the experimental fracture rim. Despite the low sample size and the single loading condition tested, the present combined numerical-experimental method showed that finite element models can predict femoral strength by providing a thorough description of the local bone mechanical response.

Generation of 3D shape, density, cortical thickness and finite element mesh of proximal femur from a DXA image.

Areal bone mineral density (aBMD), as measured by dual-energy X-ray absorptiometry (DXA), predicts hip fracture risk only moderately. Simulation of bone mechanics based on DXA imaging of the proximal femur, may help to improve the prediction accuracy. Therefore, we collected three (1-3) image sets, including CT images and DXA images of 34 proximal cadaver femurs (set 1, including 30 males, 4 females), 35 clinical patient CT images of the hip (set 2, including 27 males, 8 females) and both CT and DXA images of clinical patients (set 3, including 12 female patients). All CT images were segmented manually and landmarks were placed on both femurs and pelvises. Two separate statistical appearance models (SAMs) were built using the CT images of the femurs and pelvises in sets 1 and 2, respectively. The 3D shape of the femur was reconstructed from the DXA image by matching the SAMs with the DXA images. The orientation and modes of variation of the SAMs were adjusted to minimize the sum of the absolute differences between the projection of the SAMs and a DXA image. The mesh quality and the location of the SAMs with respect to the manually placed control points on the DXA image were used as additional constraints. Then, finite element (FE) models were built from the reconstructed shapes. Mean point-to-surface distance between the reconstructed shape and CT image was 1.0 mm for cadaver femurs in set 1 (leave-one-out test) and 1.4 mm for clinical subjects in set 3. The reconstructed volumetric BMD showed a mean absolute difference of 140 and 185 mg/cm(3) for set 1 and set 3 respectively. The generation of the SAM and the limitation of using only one 2D image were found to be the most significant sources of errors in the shape reconstruction. The noise in the DXA images had only small effect on the accuracy of the shape reconstruction. DXA-based FE simulation was able to explain 85% of the CT-predicted strength of the femur in stance loading. The present method can be used to accurately reconstruct the 3D shape and internal density of the femur from 2D DXA images. This may help to derive new information from clinical DXA images by producing patient-specific FE models for mechanical simulation of femoral bone mechanics.

Full-field strain measurement during mechanical testing of the human femur at physiologically relevant strain rates.

Understanding the mechanical properties of human femora is of great importance for the development of a reliable fracture criterion aimed at assessing fracture risk. Earlier ex vivo studies have been conducted by measuring strains on a limited set of locations using strain gauges (SGs). Digital image correlation (DIC) could instead be used to reconstruct the full-field strain pattern over the surface of the femur. The objective of this study was to measure the full-field strain response of cadaver femora tested at a physiological strain rate up to fracture in a configuration resembling single stance. The three cadaver femora were cleaned from soft tissues, and a white background paint was applied with a random black speckle pattern over the anterior surface. The mechanical tests were conducted up to fracture at a constant displacement rate of 15 mm/s, and two cameras recorded the event at 3000 frames per second. DIC was performed to retrieve the full-field displacement map, from which strains were derived. A low-pass filter was applied over the measured displacements before the crack opened in order to reduce the noise level. The noise levels were assessed using a dedicated control plate. Conversely, no filtering was applied at the frames close to fracture to get the maximum resolution. The specimens showed a linear behavior of the principal strains with respect to the applied force up to fracture. The strain rate was comparable to the values available in literature from in vivo measurements during daily activities. The cracks opened and fully propagated in less than 1 ms, and small regions with high values of the major principal strains could be spotted just a few frames before the crack opened. This corroborates the hypothesis of a strain-driven fracture mechanism in human bone. The data represent a comprehensive collection of full-field strains, both at physiological load levels and up to fracture. About 10,000 points were tracked on each bone, providing superior spatial resolution compared to ∼15 measurements typically collected using SGs. These experimental data collection can be further used for validation of numerical models, and for experimental verification of bone constitutive laws and fracture criteria.

Repeatability of digital image correlation for measurement of surface strains in composite long bones.

Digital image correlation (DIC) can measure full-field surface strains during mechanical testing of hard and soft tissues. When compared to traditional methods, such as strain gauges, DIC offers larger validation data (∼50,000 points) for, e.g., finite element models. Our main aim was to evaluate the repeatability of surface strain measurements with DIC during compressive testing of composite femurs mimicking human bones. We also studied the similarity of the composite femur samples using CT. Composite femurs were chosen as test material to minimize the uncertainties associated with the use of cadaveric tissues and to understand the variability of the DIC measurement itself. Six medium-sized fourth generation composite human proximal femora (Sawbones) were CT imaged and mechanically tested in stance configuration. The force-displacement curves were recorded and the 3D surface strains were measured with DIC on the anterior surface of the femurs. Five femurs fractured at the neck-trochanter junction and one at the site below the minor trochanter. CT image of this bone showed an air cavity at the initial fracture site. All femurs fractured through a sudden brittle crack. The fracture force for the composite bones was 5751±650N (mean±SD). The maximum von Mises strain during the fractures was 2.4±0.8%. Noise in one experiment was 5-30µÎµ. When applied loads were equalized the variation in strains between the bones was 20-25%, and when the maximum strains were equalized, variation in the other regions was 5-10%. DIC showed that the ability of nominally identical composite bones to bear high strains and loads before fracturing may vary between the samples.

Experimental validation of finite element model for proximal composite femur using optical measurements.

Patient-specific finite element models have been used to predict femur strength and fracture risk in individuals. Validation of the adopted finite element modelling procedure against mechanical testing data is a crucial step when aiming for clinical applications. The majority of the works available in literature used data from strain gages to validate the model, thus having up to 15 experimental measurements. Optical techniques, such as digital image correlation, can help to improve the models by providing a continuous field of deformation data over a femoral surface. The main objective of this study was to validate finite element models of six composite femora against strain data from digital image correlation, obtained during fracture tests performed in quasi-axial loading configuration. The finite element models were obtained from CT scans, by means of a semi-automatic segmentation. The principal strains both during the elastic phase and close to the fracture were compared, and showed a correlation coefficient close to 0.9. In the linear region, the slope and intercept were close to zero and unity, while for the case when fracture load was simulated, the slope decreased somewhat. The accuracy of the obtained results is comparable with the state-of-the-art literature, with the significant improvement of having around 50,000 data points for each femur. This large number of measurements allows a more comprehensive validation of the predictions by the finite element models, since thousand of points are tracked along the femoral neck and trochanter region, i.e., the sites that are most critical for femur fracture. Moreover, strain measurement biases due to the strain gage reinforcement effect, were avoided. The combined experimental-numerical approach proved to be ready for application to in-vitro tests of human cadaver femurs, thus helping to develop a suitable mechanistic fracture risk criterion.

Estimation of 3D rotation of femur in 2D hip radiographs.

Femoral radiographs are affected by the degree of rotation of the femur with respect to the plane of projection. We aimed to determine the 3D rotation of the proximal femur in 2D radiographs. A 3D Statistical Appearance Model (SAM), which was built from CT images of cadaver proximal femurs (n=33) was randomly sampled to form a training set of 500 bones. Nineteen clinical CT images were collected for testing. All CT images were rotated to ±20° in 2° division around the shaft axis, ±10° around medial-lateral axis, and by simultaneous rotation of both axes (±16° and ±8° around shaft and medial-lateral axes). In each orientation, a 2D projection was recorded for generating a 2D SAM. The outcome parameters of the 2D SAM were used as input for a linear regression model and an artificial neural network to predict the rotation. The artificial neural network estimated the rotation more accurately than the linear regression. For artificial neural networks the mean errors were 4.0° and 2.0° around the shaft and medial-lateral axes, respectively. For an individual radiograph, the confidence interval of estimation was still relatively large. However, this method has high potential to differentiate the amount of rotations in two image sets.