Modeling and Estimation of Personalized Spine and Torso Biomechanics

Abstract

Adolescent idiopathic scoliosis is a complex spinal deformity with a considerable prevalence that can lead to deterioration of life. Scoliosis problems of moderate degree on adolescents are typically treated using orthotic brace structures that push the spine; it is usually the only way to avoid or delay surgery. The development of personalized biomechanical models of the torso opens the door to computational solutions for the design of such braces. However, the development of patient-specific biomechanical models faces two challenges: Fitting the geometry of the torso skeleton to the patient’s anatomy and characterizing their personalized stiffness response. Before handling these challenges, the design of such models faces several unknowns, such as the correct identification of relevant mechanical elements, or the required accuracy of model parameters. In this thesis, we first design a methodology for the identification of the relevant mechanical elements, with the purpose of creating personalized models suited for patient-specific brace design and the definition of parameter estimation criteria. Next, we present a method to fit personalized geometric models of the torso skeleton that takes as input biplanar low-dose radiographs. The method relies on a biomecahnically inspired regularizer for robust fitting, and minimizes the radiation exposure compared to existing works. Finally, we describe a methodology to personalize the stiffness response of a biomechanical model of the torso and the spine. In high contrast to previous work, the proposed methodology uses controlled forcedeformation data that mimic the conditions of spinal bracing for scoliosis, which leads to personalized biomechanical models that are suitable for computational brace design. The novel method relies on a prototype system that includes controlled force-deformation measurements, a model of differentiable biomechanics of the torso, which becomes the key building block for robust parameter estimation, and an optimization procedure for parameter estimation which relies on differentiability of the biomechanics and the image generation process. Altogether, we present a method that deals with the full personalization of torso and spine models under end-to-end representative adolescent idiopathic scoliosis conditions. We demonstrate its applicability on a cohort of scoliosis patients and we show quantitative validation and improvement over previous work.