Towards Real-Time Simulation of a Finite Element Generic Lumbar Spine Model

  • Author / Creator
    Maeda, Nathanial
  • Real-time simulation of biomechanical models has provided exciting potential for improving the outcomes of some medical interventions by allowing clinicians to visualize the displacements and strains of biomechanical tissues in real-time during the intervention. Through real-time visualization, the clinician could plan the intervention quicker and adjust the intervention as they operate in response to the displacements and strains, in addition to improving the training of clinicians by using realistic biomechanics in simulation scenarios. Although investigated in other areas of medicine, no study has attempted to create a real-time simulation of the spine for implementation into spinal interventions, despite the biomechanical nature of many spinal conditions and their treatments. One significant barrier for typical FE spine models is that they require huge amounts of memory and computation time. Considering the lack of developments towards real-time spine simulation and the numerous possible applications, initial work should focus on development of a generic lumbar modelling methodology from which application-specific models may be derived. Also, given the difficulty of FE contact formulations in other real-time biomechanical simulations, facet contact is a complex and difficult problem, and thus, it requires a separate investigation in itself. Therefore, the proposed thesis aims to initiate and develop improvements to clinical applicability of FE spine models by increasing computation speed close to real-time rates. To achieve this primary objective, the research was broken into three studies: create and validate a FE lumbar spine model for gross physiologic movements, to ensure generalization of the current work, using simpler element types and materials than conventional models (Study 1); develop real-time FE techniques for spine models using graphic processing unit (GPU) in conjunction with the CUDA language (Study 2); and apply the real-time FE techniques to the FE lumbar spine model without facet contact (Study 3). In Study 1, a proposed FE lumbar spine model was created using conventional methodologies but comprised of purely tetrahedral elements and relatively stable material properties. In comparison to a conventional FE lumbar spine model using ANSYS (a conventional FE program), the proposed model exhibited similar accuracy and improved parallel computation capability with approximately 1.6X faster speeds for physiological movements. Then, in Study 2, a custom CUDA program and a simple cube model, of a similar size and material to the proposed spine model, were generated to develop and evaluate numerous parallel real-time FE techniques. In comparison with ANSYS, the CUDA program demonstrated a computation speed-up of approximately 3-4X with similar accuracy. Lastly, in Study 3, the proposed model from Study 1 (but without contact conditions) was implemented into the CUDA program from Study 2 (now the CUDA model), in addition to adding a novel composite element for the annulus fibrosus. In comparison with the conventional ANSYS model, the CUDA model demonstrated similar accuracy with approximately 20.9X speed-up versus the conventional model, in which the CUDA model’s computation time was approximately 12 seconds for flexion and lateral bending. Although not real-time, this result represents possibilities toward creating application-specific spine models that may be implemented into clinical scenarios. Still, further work is necessary to reach that long-term real-time goal, including the development of real-time FE facet contact. Future research should also focus on improving the linear solver through efficient GPU implementation and testing the CUDA program on better GPUs with more cores. Altogether, the proposed thesis demonstrated significant advancements towards improving the computation speed, and thus real-time clinical use, of FE spine models.

  • Subjects / Keywords
  • Graduation date
    Fall 2018
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.