Modélisation biomécanique de la trajectoire des vis sacro-iliaques : application pour optimiser l'ancrage tenant compte de la qualité osseuse spécifique au patient

Adult spinal deformities encompass various spinal pathologies that typically occur with age and affect 30% to 60% of the population. Surgical intervention is often necessary in elderly patients with severe deformities, and there has been a significant increase of 141% in long-segment spinal fusion procedures in the United States between 2004 and 2014. However, complications such as L5-S1 disc degeneration and sagittal imbalance can occur with an incidence of 61%, especially when the fusion extends to L5. To address these issues, S2-alar-iliac screws have been used as an alternative to iliac screws due to their lower revision rates with an odds ratio of 7.84, and their less invasive nature. However, there is a significant risk of screw loosening and insufficient healing, particularly in patients aged 70 years or older, with reported rates ranging from 50% to 66%. This risk is heightened by the commonly observed poor bone quality in the elderly, which increases the likelihood of screw loosening and pseudarthrosis in the lumbosacral region. Finite element analysis has been used in recent studies to analyze and biomechanically optimize spinopelvic implants. However, previous research has often focused on combining finite element models with medical imaging to improve the representation of bone quality or on using optimization methods in conjunction with finite element models to numerically optimize fixation strength. There is a missing link between these two aspects for realistic numerical optimization in the literature. Thus, the objective of this research project is to develop and validate a method that integrates the mechanical properties of bone tissue extracted from multi-energy biplanar radiographs, as well as using this personalized model to optimize the trajectory of S2-alar-iliac screws. Calibration between multi-energy biplanar radiographs and CT was performed to estimate volumetric bone mineral density from surface bone mineral density. Multi-energy biplanar radiographs were generated from CT data and calibrated using regression analysis. A region of interest was defined in the bone surrounding the screw to evaluate pelvic fixation, and trabecular elements in this region were grouped based on their density values. Mechanical properties were determined for each group, and a Johnson-Cook elastoplastic law was used to simulate the behavior of the bone. Cortical bone density was reduced by 35% accounting for the patient's age, and other material properties were adjusted accordingly. A validation study was conducted to assess a pelvic screw insertion approach. A specific test case was selected, using the CT-scans of an 81-year-old woman to generate corresponding radiographs. A pelvic screw was inserted into the right ilium, and its extraction force and initial stiffness were determined. Convergence studies were performed to determine the optimal number of clusters for assigning bone materials. The size of the region of interest around the screw was evaluated through extraction simulations with varying diameters and decreasing velocities. Validation was carried out by comparing the results with experimental studies and conducting statistical tests, considering uncertainties related to calibration and material properties. Sensitivity analysis was performed to compare patient-specific non-homogeneous models with homogeneous models, and different configurations were tested. Finally, a genetic algorithm was used to optimize the dimensions and trajectories of S2-alar-iliac screws for an 81-year-old woman with osteoporosis and a healthy 35-year-old woman, based on their personalized finite element models obtained following the aforementioned steps. The variables to optimize included the screw insertion point, trajectory angles, and screw size. Constraints were applied to eliminate infeasible solutions. A score was assigned to each solution based on the estimated bone mass around the screw (optimized approach) or the maximum diameter of the screw corridor (standard approach). Selection, crossover, and mutation operations were used to progressively improve the results. Simulations were performed to assess screw extraction and tipping load conditions, using force-displacement curves to calculate initial stiffness and failure force. The results were compared using paired Student t-tests. The results revealed a mean square error of 74.7 mg/cm3 between the actual Hounsfield unit values obtained from CT and the predicted values. As the size of the region of interest increased, the values of initial stiffness and extraction force converged. The validity range of the simulations compared to the experimental groups is summarized, with p-values ranging from 0.38 to 0.42 for comparison with experimental group 1 and from 0.11 to 0.83 for comparison with experimental group 2. In the sensitivity analysis, different models and trajectories yielded varying values of extraction force and initial stiffness, such as 794 N and 10964 N/mm for a homogeneous model over the entire ilium, 561 N and 7556 N/mm for a homogeneous model composed of two regions, and 939 N and 15271 N/mm for the different trajectory. The optimal trajectory of S2-alar-iliac screws was more lateral and caudal with the optimized approach, with insertion angles ranging from 49° to 66° in the sagittal plane and from 29° to 35° in the transverse plane, compared to the standard approach where the ranges were 44° to 54° and 24° to 30°, respectively. The extraction forces ranged from 5 kN to 18.4 kN with the optimized approach, which was 2.4 times higher than with the standard approach (ranging from 2.1 kN to 7.7 kN). Simulated tipping loads generated failure forces ranging from 0.8 kN to 10.1 kN with the optimized approach, which were higher than those of the standard approach (ranging from 0.9 kN to 2.9 kN). The bone mass surrounding the screw, representing the quality score and the extraction forces in the osteoporotic case, were correlated (R2 > 0.8). In conclusion, this research project developed and validated a method for integrating the mechanical properties of bone tissue extracted from multi-energy biplanar radiographs into finite element models. The personalized models were used to optimize the trajectory of S2-alar-iliac screws for patients with spinal deformities. The results demonstrated the feasibility and effectiveness of the optimized approach, which provided better screw-bone interaction and improved screw performance compared to the standard approach. This method has the potential to enhance surgical outcomes and reduce complications in elderly patients with spinal deformities. Further research and clinical validation are necessary to confirm the findings and establish the practical applicability of this approach in real-world scenarios.