Document Type

Dissertation

Date of Award

2018

Keywords

Biological sciences, Applied sciences, Artificial neural network, Finite element analysis, Lumbar spine, Multi-objective optimization, Total disc arthroplasty

Department

Mechanical Engineering

First Advisor

Ryan Willing

Subject Heading(s)

Biological sciences; Applied sciences; Artificial neural network; Finite element analysis; Lumbar spine; Multi-objective optimization; Total disc arthroplasty

Abstract

Total disc arthroplasty (TDA) is a motion-preserving surgical technique used to treat spinal disorders, when more conservative medical therapies fail. Unfortunately, a high incidence of revision surgery exists due to postoperative complications including abnormal kinematics, facet joint arthritis, and implant failures. However, TDA is still an attractive option, since an optimally designed artificial disc is expected to reproduce native segmental biomechanics. Correspondingly, it would mitigate the development of adjacent segment diseases (a major concern of spinal fusion) caused by altered segmental biomechanics.

Design optimization is a process of finding the best design parameters for a component/system to satisfy one/multiple design requirements using optimization algorithms. The shape of a candidate design is parametrized using computer-aided design, such that design parameters are manipulated to minimize one/multiple objective functions subject to performance constraints and design space bounds. Optimization algorithms typically require the gradients of the objective/constraint functions with respect to each design variable. In the traditional design optimization, due to the high computational cost to calculate the gradients by performing finite element analysis in each optimization iteration, it often results in a slow process to seek the optimal solution. To address the problem, an artificial neural network (ANN) was implemented to derive the analytical expressions of the objective/constraint function and their gradients. By incorporating analytical gradients, we successfully developed a multiobjective optimization (MOO) framework considering three performance metrics simultaneously.

Furthermore, a new mobile-bearing TDA design concept featuring a biconcave polyethylene (PE) core was proposed, to strengthen the PE rim, where a high risk of fracture exists. It was hypothesized that there is a trade-off relationship among postoperative performance metrics in terms of spinal kinematics, facet joint loading, and metal-on-polyethylene contact mechanics. We tested this hypothesis by refining the new TDA to match normal segmental biomechanics and alleviate PE core stress. After performing MOO, the best-trade-off TDA design was determined by the solved three-dimensional Pareto frontier. The novel MOO framework can be also used to improve existing TDA designs, as well as to push the cutting edge of surgical techniques for the treatment of spinal disorders.

Share

COinS