Voxel size dependency, reproducibility and sensitivity of an in vivo bone loading estimation algorithm

Bone remodeling and responsiveness to mechanical stimuli in individuals with type 1 diabetes mellitus

Type 1 diabetes mellitus (T1DM) has been linked to increased osteocyte apoptosis, local accumulation of mineralized lacunar spaces, and microdamage suggesting an impairment of the mechanoregulation network in affected individuals. Diabetic neuropathy might exacerbate this dysfunction through direct effects on bone turnover, and indirect effects on balance, muscle strength, and gait. However, the in vivo effects of impaired bone mechanoregulation on bone remodeling in humans remain underexplored. This longitudinal cohort study assessed consenting participants with T1DM and varying degree of distal symmetric sensorimotor polyneuropathy (T1DM, n = 20, median age 46.5 yr, eight female) and controls (CTRL; n = 9, median age 59.0 yr, four female) at baseline and 4-yr follow-up. Nerve conduction in participants with T1DM was tested using DPNCheck and bone remodeling was quantified with longitudinal high-resolution peripheral quantitative-computed tomography (HR-pQCT, 82 μm) at the standard distal sites. Local trabecular bone formation (Tb.F) and resorption (Tb.R) sites were captured by implementing 3D rigid image registration of HR-pQCT images, and the mechanical environment across the bone microarchitecture at these sites was simulated using micro-finite element analysis. We calculated odds ratios to determine the likelihood of bone formation (ORF) and resorption (ORR) with increasing/decreasing strain in percent as markers for mechanoregulation. At the distal radius, Tb.F was 47% lower and Tb.R was 59% lower in T1DM participants compared with CTRL (P < .05). Tb.F correlated positively with nerve conduction amplitude (R = 0.69, P < .05) in participants with T1DM and negatively with glycated hemoglobin (HbA1c) (R = -0.45, P < .05). Additionally, ORF was 34% lower and ORR was 18% lower in T1DM compared with CTRL (P < .05). Our findings represent in vivo evidence suggesting that bone remodeling in individuals with T1DM is in a state of low responsiveness to mechanical stimuli, resulting in impaired bone formation and resorption rates; these correlate to the degree of neuropathy and level of diabetes control.

The sacroiliac joint: An original and highly sensitive tool to highlight altered bone phenotype in murine models of skeletal disorders

Bone disorders may affect the skeleton in different ways, some bones being very impaired and others less severely. In translational studies using murine models of human skeletal diseases, the bone phenotype is mainly evaluated at the distal femur or proximal tibia. The sacroiliac joint (SIJ), which connects the spine to the pelvis, is involved in the balanced transfer of mechanical energy from the lumbar spine to the lower extremities. Because of its role in biomechanical stress, the SIJ is a region of particular interest in various bone diseases. Here we aimed to characterize the SIJ in several murine models to develop a highly reliable tool for studying skeletal disorders. We performed a 12-month in vivo micro-computed tomography (micro-CT) follow-up to characterize the SIJ in wild-type (WT) C57BL/J6 mice and compared the bone microarchitecture of the SIJ and the distal femur at 3 months by micro-CT and histology. To test the sensitivity of our methodology, the SIJ and distal femur were evaluated at 3 and 6 months, in 2 murine models of skeletal disorder, X-linked hypophosphatemia (Hyp mice) and HLA-B27 transgenic mice and compared to WT mice. A multimodal analysis was performed, using a combination of microCT and histological analysis. With the Hyp model, the SIJ displayed more bone microarchitecture alterations than the distal femur. Hyp mice showed a significant reduction in trabecular bone at both the distal femur and sacral slope as compared with WT mice, with a significant positive correlation between trabecular bone parameters of the distal femur and sacral side of the SIJ. Furthermore, trabecular bone parameters (Bone Volume/Total Volume (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), trabecular number (Tb.N), trabecular pattern factor (Tb.Pf)) were significantly increased compared to femoral parameters at the SIJ. The sacral articular cortical bone, which is indicative of osteoarticular lesions, was altered in Hyp mice. Interestingly, in accordance to previous studies, HLA-B27 transgenic mice did not show any osteoarticular lesions as compared with WT mice. Cortical bone parameters (thickness, porosity), as well as scoring performed with double blinding, did not show difference between the 2 genotypes. The characterization and evaluation of the SIJ surface appears very sensitive to emphasize alterations of bone and joint. The SIJ may represent a valuable tool to investigate both bone and local osteoarticular alterations in murine models of skeletal disorders and might be a relevant site for assessing the response to treatment of chronic bone diseases.

Hip joint load prediction using inverse bone remodeling with homogenized FE models: Comparison to micro-FE and influence of material modeling strategy

BACKGROUND AND OBJECTIVE

Measuring physiological loading conditions in vivo can be challenging, as methods are invasive or pose a high modeling effort. However, the physiological loading of bones is also imprinted in the bone microstructure due to bone (re)modeling. This information can be retrieved by inverse bone remodeling (IBR). Recently, an IBR method based on micro-finite-element (µFE) modeling was translated to homogenized-FE (hFE) to decrease computational effort and tested on the distal radius. However, this bone has a relatively simple geometry and homogeneous microstructure. Therefore, the objective of this study was to assess the agreement of hFE-based IBR with µFE-based IBR to predict hip joint loading from the head of the femur; a bone with more complex loading as well as more heterogeneous microstructure.

METHODS

hFE-based IBR was applied to a set of 19 femoral heads using four different material mapping laws. One model with a single homogeneous material for both trabecular and cortical volume and three models with a separated cortex and either homogeneous, density-dependent inhomogeneous, or density and fabric-dependent orthotropic material. Three different evaluation regions (full bone, trabecular bone only, head region only) were defined, in which IBR was applied. µFE models were created for the same bones, and the agreement of the predicted hip joint loading history obtained from hFE and µFE models was evaluated. The loading history was discretized using four unit load cases.

RESULTS

The computational time for FE solving was decreased on average from 500 h to under 1 min (CPU time) when using hFE models instead of µFE models. Using more information in the material model in the hFE models led to a better prediction of hip joint loading history. Inhomogeneous and inhomogeneous orthotropic models gave the best agreement to µFE-based IBR (RMSE% <14%). The evaluation region only played a minor role.

CONCLUSIONS

hFE-based IBR was able to reconstruct the dominant joint loading of the femoral head in agreement with µFE-based IBR and required considerably lower computational effort. Results indicate that cortical and trabecular bone should be modeled separately and at least density-dependent inhomogeneous material properties should be used with hFE models of the femoral head to predict joint loading.

Personalized loading conditions for homogenized finite element analysis of the distal sections of the radius

The microstructure of trabecular bone is known to adapt its morphology in response to mechanical loads for achieving a biomechanical homeostasis. Based on this form-function relationship, previous investigators either simulated the remodeling of bone to predict the resulting density and architecture for a specific loading or retraced physiological loading conditions from local density and architecture. The latter inverse approach includes quantifying bone morphology using computed tomography and calculating the relative importance of selected load cases by minimizing the fluctuation of a tissue loading level metric. Along this concept, the present study aims at identifying an optimal, personalized, multiaxial load case at the distal section of the human radius using in vivo HR-pQCT-based isotropic, homogenized finite element (hFE) analysis. The dataset consisted of HR-pQCT reconstructions of the 20 mm most distal section of 21 human fresh-frozen radii. We simulated six different unit canonical load cases (FX palmar-dorsal force, FY ulnar-radial force, FZ distal-proximal force, MX moment about palmar-dorsal, MY moment about ulnar-radial, MZ moment about distal-proximal) using a simplified and efficient hFE method based on a single isotropic bone phase. Once we used a homogeneous mean density (shape model) and once the original heterogeneous density distribution (shape + density model). Using an analytical formulation, we minimized the deviation of the resulting strain tensors ε(x) to a hydrostatic compressive reference strain ε0, once for the 6 degrees of freedom (DOF) optimal (OPT) load case and for all individual 1 DOF load cases (FX, FY, FZ, MX, MY, MZ). All seven load cases were then extended in the nonlinear regime using the scaled displacements of the linear load cases as loading boundary conditions (MAX). We then compared the load cases and models for their objective function (OF) values, the stored energies and their ultimate strength using a specific torsor norm. Both shape and shape + density linear-optimized OPT models were dominated by a positive force in the z-direction (FZ). Transversal force DOFs were close to zero and mean moment DOFs were different depending on the model type. The inclusion of density distribution increased the influence and changed direction of MX and MY, while MZ was small in both models. The OPT load case had 12-15% lower objective function (OF) values than the FZ load case, depending on the model. Stored energies at the optimum were consistently 142-178% higher for the OPT load case than for the FZ load case. Differences in the nonlinear response maximum torsor norm ‖t‖ were heterogeneous, but consistently higher for OPT_MAX than FZ_MAX. We presented the proof of concept of an optimization procedure to estimate patient-specific loading conditions for hFE methods. In contrast to similar models, we included canonical load cases in all six DOFs and used a strain metric that favors hydrostatic compression. Based on a biomechanical analysis of the distal joint surfaces at the radius, the estimated load directions are plausible. For our dataset, the resulting OPT load case is close to the standard axial compression boundary conditions, usually used in HR-pQCT-based FE analysis today. But even using the present simplified hFE model, the optimized linear six DOF load case achieves a more homogeneous tissue loading and can absorb more than twice the energy than the standard uniaxial load case. The ultimate strength calculated with a torsor norm was consistently higher for the 6-DOF nonlinear model (OPT_MAX) than for the 1-DOF nonlinear uniaxial model (FZ_MAX). Defining patient-specific boundary conditions may decrease angulation errors during CT measurements and improve repeatability as well as reproducibility of bone stiffness and strength estimated by HR-pQCT-based hFE analysis. These results encourage the extension of the present method to anisotropic hFE models and their application to repeatability data sets to test the hypothesis of reduced angulation errors during measurement.

A Density-Dependent Target Stimulus for Inverse Bone (Re)modeling with Homogenized Finite Element Models

Inverse bone (re)modeling (IBR) can infer physiological loading conditions from the bone microstructure. IBR scales unit loads, imposed on finite element (FE) models of a bone, such that the trabecular microstructure is homogeneously loaded and the difference to a target stimulus is minimized. Micro-FE (µFE) analyses are typically used to model the microstructure, but computationally more efficient, homogenized FE (hFE) models, where the microstructure is replaced by an equivalent continuum, could be used instead. However, also the target stimulus has to be translated from the tissue to the continuum level. In this study, a new continuum-level target stimulus relating relative bone density and strain energy density is proposed. It was applied using different types of hFE models to predict the physiological loading of 21 distal radii sections, which was subsequently compared to µFE-based IBR. The hFE models were able to correctly identify the dominant load direction and showed a high correlation of the predicted forces, but mean magnitude errors ranged from − 14.7 to 26.6% even for the best models. While µFE-based IBR can still be regarded as a gold standard, hFE-based IBR enables faster predictions, the usage of more sophisticated boundary conditions, and the usage of clinical images.

Bone Mechanoregulation Allows Subject-Specific Load Estimation Based on Time-Lapsed Micro-CT and HR-pQCT in Vivo

Patients at high risk of fracture due to metabolic diseases frequently undergo long-term antiresorptive therapy. However, in some patients, treatment is unsuccessful in preventing fractures or causes severe adverse health outcomes. Understanding load-driven bone remodelling, i.e., mechanoregulation, is critical to understand which patients are at risk for progressive bone degeneration and may enable better patient selection or adaptive therapeutic intervention strategies. Bone microarchitecture assessment using high-resolution peripheral quantitative computed tomography (HR-pQCT) combined with computed mechanical loads has successfully been used to investigate bone mechanoregulation at the trabecular level. To obtain the required mechanical loads that induce local variances in mechanical strain and cause bone remodelling, estimation of physiological loading is essential. Current models homogenise strain patterns throughout the bone to estimate load distribution in vivo, assuming that the bone structure is in biomechanical homoeostasis. Yet, this assumption may be flawed for investigating alterations in bone mechanoregulation. By further utilising available spatiotemporal information of time-lapsed bone imaging studies, we developed a mechanoregulation-based load estimation (MR) algorithm. MR calculates organ-scale loads by scaling and superimposing a set of predefined independent unit loads to optimise measured bone formation in high-, quiescence in medium-, and resorption in low-strain regions. We benchmarked our algorithm against a previously published load history (LH) algorithm using synthetic data, micro-CT images of murine vertebrae under defined experimental in vivo loadings, and HR-pQCT images from seven patients. Our algorithm consistently outperformed LH in all three datasets. In silico-generated time evolutions of distal radius geometries (n = 5) indicated significantly higher sensitivity, specificity, and accuracy for MR than LH (p < 0.01). This increased performance led to substantially better discrimination between physiological and extra-physiological loading in mice (n = 8). Moreover, a significantly (p < 0.01) higher association between remodelling events and computed local mechanical signals was found using MR [correct classification rate (CCR) = 0.42] than LH (CCR = 0.38) to estimate human distal radius loading. Future applications of MR may enable clinicians to link subtle changes in bone strength to changes in day-to-day loading, identifying weak spots in the bone microstructure for local intervention and personalised treatment approaches.

Cortical and trabecular bone structure of the hominoid capitate

Morphological variation in the hominoid capitate has been linked to differences in habitual locomotor activity due to its importance in movement and load transfer at the midcarpal joint proximally and carpometacarpal joints distally. Although the shape of bones and their articulations are linked to joint mobility, the internal structure of bones has been shown experimentally to reflect, at least in part, the loading direction and magnitude experienced by the bone. To date, it is uncertain whether locomotor differences among hominoids are reflected in the bone microarchitecture of the capitate. Here, we apply a whole‐bone methodology to quantify the cortical and trabecular architecture (separately and combined) of the capitate across bipedal (modern Homo sapiens), knuckle‐walking (Pan paniscus, Pan troglodytes, Gorilla sp.), and suspensory (Pongo sp.) hominoids (n = 69). It is hypothesized that variation in bone microarchitecture will differentiate these locomotor groups, reflecting differences in habitual postures and presumed loading force and direction. Additionally, it is hypothesized that trabecular and cortical architecture in the proximal and distal regions, as a result of being part of mechanically divergent joints proximally and distally, will differ across these portions of the capitate. Results indicate that the capitate of knuckle‐walking and suspensory hominoids is differentiated from bipedal Homo primarily by significantly thicker distal cortical bone. Knuckle‐walking taxa are further differentiated from suspensory and bipedal taxa by more isotropic trabeculae in the proximal capitate. An allometric analysis indicates that size is not a significant determinate of bone variation across hominoids, although sexual dimorphism may influence some parameters within Gorilla. Results suggest that internal trabecular and cortical bone is subjected to different forces and functional adaptation responses across the capitate (and possibly other short bones). Additionally, while separating trabecular and cortical bone is normal protocol of current whole‐bone methodologies, this study shows that when applied to carpals, removing or studying the cortical bone separately potentially obfuscates functionally relevant signals in bone structure.

Validation of HR-pQCT against micro-CT for morphometric and biomechanical analyses: A review

High-resolution peripheral quantitative computed-tomography (HR-pQCT) has the potential to become a powerful clinical assessment and diagnostic tool. Given the recent improvements in image resolution, from 82 to 61 μm, this technology may be used to accurately quantify in vivo bone microarchitecture, a key biomarker of degenerative bone diseases. However, computational methods to assess bone microarchitecture were developed for micro computed tomography (micro-CT), a higher-resolution technology only available for ex vivo studies, and validation of these computational analysis techniques against the gold-standard micro-CT has been inconsistent and incomplete. Herein, we review methods for segmentation of bone compartments and microstructure, quantification of bone morphology, and estimation of mechanical strength using finite-element analysis, highlighting the need throughout for improved standardization across the field. Studies have relied on homogenous datasets for validation, which does not allow for robust comparisons between methods. Consequently, the adaptation and validation of novel segmentation approaches has been slow to non-existent, with most studies still using the manufacturer's segmentation for morphometric analysis despite the existence of better performing alternative approaches. The promising accuracy of HR-pQCT for capturing morphometric indices is overshadowed by considerable variability in outcomes between studies. For finite element analysis (FEA) methods, the use of disparate material models and FEA tools has led to a fragmented ability to assess mechanical bone strength with HR-pQCT. Further, the scarcity of studies comparing 62 μm HR-pQCT to the gold standard micro-CT leaves the validation of this imaging modality incomplete. This review revealed that without standardization, the capabilities of HR-pQCT cannot be adequately assessed. The need for a public, extendable, heterogeneous dataset of HR-pQCT and corresponding gold-standard micro-CT images, which would allow HR-pQCT users to benchmark existing and novel methods and select optimal methods depending on the scientific question and data at hand, is now evident. With more recent advancements in HR-pQCT, the community must learn from its past and provide properly validated technologies to ensure that HR-pQCT can truly provide value in patient diagnosis and care.

Inverse remodelling algorithm identifies habitual manual activities of primates based on metacarpal bone architecture

Previously, a micro-finite element (micro-FE)-based inverse remodelling method was presented in the literature that reconstructs the loading history of a bone based on its architecture alone. Despite promising preliminary results, it remains unclear whether this method is sensitive enough to detect differences of bone loading related to pathologies or habitual activities. The goal of this study was to test the sensitivity of the inverse remodelling method by predicting joint loading histories of metacarpal bones of species with similar anatomy but clearly distinct habitual hand use. Three groups of habitual hand use were defined using the most representative primate species: manipulation (human), suspensory locomotion (orangutan), and knuckle-walking locomotion (bonobo, chimpanzee, gorilla). Nine to ten micro-computed tomography scans of each species ([Formula: see text] in total) were used to create micro-FE models of the metacarpal head region. The most probable joint loading history was predicted by optimally scaling six load cases representing joint postures ranging from [Formula: see text] (extension) to [Formula: see text] (flexion). Predicted mean joint load directions were significantly different between knuckle-walking and non-knuckle-walking groups ([Formula: see text]) and in line with expected primary hand postures. Mean joint load magnitudes tended to be larger in species using their hands for locomotion compared to species using them for manipulation. In conclusion, this study shows that the micro-FE-based inverse remodelling method is sensitive enough to detect differences of joint loading related to habitual manual activities of primates and might, therefore, be useful for palaeoanthropologists to reconstruct the behaviour of extinct species and for biomedical applications such as detecting pathological joint loading.

Virtual supersampling as post-processing step preserves the trabecular bone morphometry in human peripheral quantitative computed tomography scans

In the clinical field of diagnosis and monitoring of bone diseases, high-resolution peripheral quantitative computed tomography (HR-pQCT) is an important imaging modality. It provides a resolution where quantitative bone morphometry can be extracted in vivo on patients. It is known that HR-pQCT provides slight differences in morphometric indices compared to the current standard approach micro-computed tomography (micro-CT). The most obvious reason for this is the restriction of the radiation dose and with this a lower image resolution. With advances in micro-CT evaluation techniques such as patient-specific remodeling simulations or dynamic bone morphometry, a higher image resolution would potentially also allow the application of such novel evaluation techniques to clinical HR-pQCT measurements. Virtual supersampling as post-processing step was considered to increase the image resolution of HR-pQCT scans. The hypothesis was that this technique preserves the structural bone morphometry. Supersampling from 82 μm to virtual 41 μm by trilinear interpolation of the grayscale values of 42 human cadaveric forearms resulted in strong correlations of structural parameters (R²: 0.96–1.00). BV/TV was slightly overestimated (4.3%, R²: 1.00) compared to the HR-pQCT resolution. Tb.N was overestimated (7.47%; R²: 0.99) and Tb.Th was slightly underestimated (-4.20%; R²: 0.98). The technique was reproducible with PE_%CV between 1.96% (SMI) and 7.88% (Conn.D). In a clinical setting with 205 human forearms with or without fracture measured at 82 μm resolution HR-pQCT, the technique was sensitive to changes between groups in all parameters (p < 0.05) except trabecular thickness. In conclusion, we demonstrated that supersampling preserves the bone morphometry from HR-pQCT scans and is reproducible and sensitive to changes between groups. Supersampling can be used to investigate on the resolution dependency of HR-pQCT images and gain more insight into this imaging modality.

Plausibility and parameter sensitivity of micro-finite element-based joint load prediction at the proximal femur

A micro-finite element-based method to estimate the bone loading history based on bone architecture was recently presented in the literature. However, a thorough investigation of the parameter sensitivity and plausibility of this method to predict joint loads is still missing. The goals of this study were (1) to analyse the parameter sensitivity of the joint load predictions at one proximal femur and (2) to assess the plausibility of the results by comparing load predictions of ten proximal femora to in vivo hip joint forces measured with instrumented prostheses (available from www.orthoload.com ). Joint loads were predicted by optimally scaling the magnitude of four unit loads (inclined [Formula: see text] to [Formula: see text] with respect to the vertical axis) applied to micro-finite element models created from high-resolution computed tomography scans ([Formula: see text]m voxel size). Parameter sensitivity analysis was performed by varying a total of nine parameters and showed that predictions of the peak load directions (range 10[Formula: see text]-[Formula: see text]) are more robust than the predicted peak load magnitudes (range 2344.8-4689.5 N). Comparing the results of all ten femora with the in vivo loading data of ten subjects showed that peak loads are plausible both in terms of the load direction (in vivo: [Formula: see text], predicted: [Formula: see text]) and magnitude (in vivo: [Formula: see text], predicted: [Formula: see text]). Overall, this study suggests that micro-finite element-based joint load predictions are both plausible and robust in terms of the predicted peak load direction, but predicted load magnitudes should be interpreted with caution.

Computational modeling of long-term effects of prophylactic vertebroplasty on bone adaptation

Cement augmentation in vertebrae (vertebroplasty) is usually used to restore mechanical strength after spinal fracture but could also be used as a prophylactic treatment. So far, the mechanical competence has been determined immediately post-treatment, without considering long-term effects of bone adaptation. In this work, we investigated such long-term effects of vertebroplasty on the stiffness of the augmented bone by means of computational simulation of bone adaptation. Using micro-finite element analysis, we determined sites of increased mechanical stress (stress raisers) and stress shielding and, based on the simulations, regions with increased or decreased bone loss due to augmentation. Cement volumes connecting the end plates led to increased stress shielding and bone loss. The increased stiffness due to the augmentation, however, remained constant over the simulation time of 30 years. If the intervention was performed at an earlier time point, it did lead to more bone loss, but again, it did not affect long-term stability as this loss was compensated by bone gains in other areas. In particular, around the augmentation cement, bone structures were preserved, suggesting a long-term integration of the cement in the augmented bone. We conclude that, from a biomechanical perspective, the impact of vertebroplasty on the bone at the microstructural level is less detrimental than previously thought.

In vivo Visualisation and Quantification of Bone Resorption and Bone Formation from Time-Lapse Imaging

PURPOSE OF REVIEW

Mechanoregulation of bone cells was proposed over a century ago, but only now can we visualise and quantify bone resorption and bone formation and its mechanoregulation. In this review, we show how the newest advances in imaging and computational methods paved the way for this breakthrough.

RECENT FINDINGS

Non-invasive in vivo assessment of bone resorption and bone formation was demonstrated by time-lapse micro-computed tomography in animals, and by high-resolution peripheral quantitative computed tomography in humans. Coupled with micro-finite element analysis, the relationships between sites of bone resorption and bone formation and low and high tissue loading, respectively, were shown. Time-lapse in vivo imaging and computational methods enabled visualising and quantifying bone resorption and bone formation as well as its mechanoregulation. Future research includes visualising and quantifying mechanoregulation of bone resorption and bone formation from molecular to organ scales, and translating the findings into medicine using personalised bone health prognosis.

Micro-computed tomography characterization of tissue engineering scaffolds: effects of pixel size and rotation step

Quantitative assessment of micro-structure of materials is of key importance in many fields including tissue engineering, biology, and dentistry. Micro-computed tomography (µ-CT) is an intensively used non-destructive technique. However, the acquisition parameters such as pixel size and rotation step may have significant effects on the obtained results. In this study, a set of tissue engineering scaffolds including examples of natural and synthetic polymers, and ceramics were analyzed. We comprehensively compared the quantitative results of µ-CT characterization using 15 acquisition scenarios that differ in the combination of the pixel size and rotation step. The results showed that the acquisition parameters could statistically significantly affect the quantified mean porosity, mean pore size, and mean wall thickness of the scaffolds. The effects are also practically important since the differences can be as high as 24% regarding the mean porosity in average, and 19.5 h and 166 GB regarding the characterization time and data storage per sample with a relatively small volume. This study showed in a quantitative manner the effects of such a wide range of acquisition scenarios on the final data, as well as the characterization time and data storage per sample. Herein, a clear picture of the effects of the pixel size and rotation step on the results is provided which can notably be useful to refine the practice of µ-CT characterization of scaffolds and economize the related resources.

Towards in silico prognosis using big data

Clinical diagnosis and prognosis usually rely on few or even single measurements despite clinical big data being available. This limits the exploration of complex diseases such as adolescent idiopathic scoliosis (AIS) where the associated low bone mass remains unexplained. Observed low physical activity and increased RANKL/OPG, however, both indicate a mechanobiological cause. To deepen disease understanding, we propose an in silico prognosis approach using clinical big data, i.e. medical images, serum markers, questionnaires and live style data from mobile monitoring devices and explore the role of inadequate physical activity in a first AIS prototype. It employs a cellular automaton (CA) to represent the medical image, micro-finite element analysis to calculate loading, and a Boolean network to integrate the other biomarkers. Medical images of the distal tibia, physical activity scores, and vitamin D and PTH levels were integrated as measured clinically while the time development of bone density and RANKL/OPG was observed. Simulation of an AIS patient with normal physical activity and patient-specific vitamin D and PTH levels showed minor changes in bone density whereas the simulation of the same AIS patient but with reduced physical activity led to low density. Both showed unchanged RANKL/OPG and considerable cortical resorption. We conclude that our integrative in silico approach allows to account for a variety of clinical big data to study complex diseases.