BACKGROUND:Vertebral metastases compromise bone integrity and may lead to fractures with substantial clinical and economic consequences. Finite element (FE) modeling based on computed tomography (CT) scans enables individualized biomechanical assessment but remains underused clinically due to workflow limitations. OBJECTIVE:To evaluate the sensitivity and accuracy of a patient-specific finite element (FE) simulation pipeline that incorporates automated vertebral segmentation for predicting vertebral strength in metastatic cases. METHODS:Thirty vertebrae (18 metastatic, 12 healthy) from 12 patients were analyzed using two finite element (FE) models (perfectly elasto-plastic vs. linear elastic) within an automated segmentation pipeline. Intra- and inter-operator reproducibility, sensitivity to boundary condition definition, and model-type comparisons were performed. Failure loads were normalized to derive equivalent stresses. RESULTS:Automated segmentation demonstrated excellent agreement with manual reference (R2 = 0.996 p < 0.001; no systematic bias). Sensitivity to operator-defined boundary conditions was minimal (<1%). The elasto-plastic model provided significantly lower failure loads than the linear elastic model (-25% on average). CONCLUSION:This study supports the feasibility of a fully-automated CT-based finite-element (FE) pipeline for vertebral fracture risk assessment in both healthy and metastatic vertebrae.

Computed tomography (CT) based finite element (FE) models for bone strength prediction of metastasised femurs require calibration of Hounsfield Units (HU) to bone mineral density, to then derive mechanical parameters. Calibration is generally performed using a calibration phantom, which is a limitation for clinical application. The purpose of this study was to evaluate a phantomless calibration method for linear and non-linear FE models of femoral strength prediction. The constitutive laws of the linear and non-linear models, respectively, are defined by two and six density-dependent parameters, respectively. Femurs of 54 patients with bone metastases were CT-scanned with a calibration phantom. Linear and non-linear FE models were created for each femur, following calibration using (i) the phantom, and (ii) the air-fat-muscle (AFM) calibration method, where HU peaks of air, fat, and muscle are extracted and linearly fitted to reference density values. Apparent density and failure loads were compared between each calibration method, and the intra- and inter- operator variability was computed, for each model. A very good reproducibility of the AFM method was found for both models, and a significant correlation with the gold standard phantom-based calibration in terms of apparent density (R2 = 0.78, p<0.01) and femoral strength (R2 = 0.85, p<0.01). Limits of agreement between each method were narrower for the linear model than for the non-linear model, suggesting that the linear model is less sensitive to density calibration. The phantomless calibration method is a promising approach to broaden routine use of bone strength numerical simulation in bone metastatic patients.

Highlights• Automated segmentation of vertebral body showed excellent agreement with expert• Sensitivity to operator-defined boundary conditions was minimal• Discrepancies between elastic and elastoplastic models

Bone segmentation is an important step to perform biomechanical failure load simulations on in-vivo CT data of patients with bone metastasis, as it is a mandatory operation to obtain meshes needed for numerical simulations. Segmentation can be a tedious and time consuming task when done manually, and expert segmentations are subject to intra- and inter-operator variability. Deep learning methods are increasingly employed to automatically carry out image segmentation tasks. These networks usually need to be trained on a large image dataset along with the manual segmentations to maximize generalization to new images, but it is not always possible to have access to a multitude of CT-scans with the associated ground truth. It then becomes necessary to use training techniques to make the best use of the limited available data. In this paper, we propose a dedicated pipeline of preprocessing, deep learning based segmentation method and post-processing for in-vivo human femurs and vertebrae segmentation from CT-scans volumes. We experimented with three U-Net architectures and showed that out-of-the-box models enable automatic and high-quality volume segmentation if carefully trained. We compared the failure load simulation results obtained on femurs and vertebrae using either automatic or manual segmentations and studied the sensitivity of the simulations on small variations of the automatic segmentation. The failure loads obtained using automatic segmentations were comparable to those obtained using manual expert segmentations for all the femurs and vertebrae tested, demonstrating the effectiveness of the automated segmentation approach for failure load simulations.

Purpose: Bone metastasis have a major impact on the quality of life of patients and they are diverse in terms of size and location, making their segmentation complex. Manual segmentation is time-consuming, and expert segmentations are subject to operator variability, which makes obtaining accurate and reproducible segmentations of bone metastasis on CT-scans a challenging yet important task to achieve. Materials and Methods: Deep learning methods tackle segmentation tasks efficiently but require large datasets along with expert manual segmentations to generalize on new images. We propose an automated data synthesis pipeline using 3D Denoising Diffusion Probabilistic Models (DDPM) to enchance the segmentation of femoral metastasis from CT-scan volumes of patients. We used 29 existing lesions along with 26 healthy femurs to create new realistic synthetic metastatic images, and trained a DDPM to improve the diversity and realism of the simulated volumes. We also investigated the operator variability on manual segmentation. Results: We created 5675 new volumes, then trained 3D U-Net segmentation models on real and synthetic data to compare segmentation performance, and we evaluated the performance of the models depending on the amount of synthetic data used in training. Conclusion: Our results showed that segmentation models trained with synthetic data outperformed those trained on real volumes only, and that those models perform especially well when considering operator variability.

Metastases increase the risk of fracture when affecting the femur. Consequently, clinicians need to know if the patients femur can withstand the stress of daily activities. The current tools used in clinics are not sufficiently precise. A new method, the CT-scan-based finite element analysis, gives good predictive results. However, none of the existing models were tested for reproducibility. This is a critical issue to address in order to apply the technique on a large cohort around the world to help evaluate bone metastatic fracture risk in patients. Please see pdf file

The mechanical properties of the extracellular matrix are essential for regulating cancer cell behavior, but how they change depending on tumour type remains unclear. The aim of the current study was to determine how the mechanical properties of tumours that frequently metastasize to bones were affected depending on histological type. Human breast, kidney, and thyroid specimens containing tumour and normal tissue were collected during surgery. The elastic modulus and elastic fraction of each sample were characterized using atomic force microscopy and compared with histopathological markers. We observed that tumour mechanical properties were differentially affected depending on organ and histological type. Indeed, clear cell renal carcinoma and poorly differentiated thyroid carcinoma displayed a decrease in the elastic modulus compared to their normal counterpart, while breast tumours, papillary renal carcinoma and fibrotic thyroid tumours displayed an increase in the elastic modulus. Elastic fraction decreased only for thyroid tumour tissue, indicating an increase in viscosity. These findings suggest a unique mechanical profile associated with each subtype of cancer. Therefore, viscosity could be a discriminator between tumour and normal thyroid tissue, while elasticity could be a discriminator between the subtypes of breast, kidney and thyroid cancers.