Optimization of pedicle screw position using finite element method and neural networks

Abstract

The application of pedicle screws in instrumentation of the spine is widely used in spine correction surgeries. High stresses can be developed in the region surrounding the screw, and it is important to understand how to minimize this stress in order to reduce risk of patient harm. Traditional studies have used FEM (finite element method) to evaluate the best position and orientation of the pedicle screws; however, such approach demands iterative simulations to find the minimum stress position. This work has used an optimization-based method, based on neural network and genetic algorithm, to test different pedicular screw positions. By performing simulations of different screw positions through finite element analysis and using the results to train an artificial neural network, a less computationally intensive method of determining stress was found. The mathematical model created was optimized through a genetic algorithm and used to find a trajectory that results in lower mechanical stress between the pedicle screw and the vertebra. The results of the ANN approach show a reduction of 5.25% in the von Mises stress when compared to the FEA. Regarding the optimal trajectory, the angles differences were 2° in sagittal plane and 1.4° in transverse plane, hence showing good results.

Comprehensive analysis of pedicle screw implantation in the C7 vertebra using computed tomography-based three-dimensional models

“…Various factors affected the pullout strength of pedicle screw have been reported [ 20 – 25 ]. A biomechanical study conducted by Kueny et al [ 23 ] showed that increasing the screw diameter by 1 mm can increase 24% pullout force.…”

“…A biomechanical study conducted by Kueny et al [ 23 ] showed that increasing the screw diameter by 1 mm can increase 24% pullout force. In another study, the researchers proved that the pullout strength of pedicle screw can be increased when screw contact with more cortical bone [ 25 ]. Besides, the bone mineral density of entry point and vertebral body can also affect the axial pullout strength [ 24 ].…”

Performance Prediction of the Elastic Support Structure of a Wind Turbine Based on Multi-Task Learning

“…Moreover, Wang et al [25] used a BP neural network to model the relationship between weight, dynamic characteristics, and surface shape errors of a large space-based mirror, and its structural parameters, with training data obtained through finite element simulation analysis. Furthermore, Silva et al [26] employed a finite element simulation model to assess the stress impact of the pedicle screw installation position and direction on the installation area, training a neural network with simulation experimental data to minimize the mechanical stress between the pedicle screw and vertebrae.…”