Muscle strength assessment is a standard part of any clinical evaluation. Due to the kinematic and muscular redundancy in the human musculoskeletal system, muscle strength can not be measured directly in vivo. Clinicians utilize specific postures and forces to bias the muscles of interest, then infer the muscle strength from an indirect measurement. This is particularly challenging in the cervical spine, where more than 25 muscle pairs with multiple attachments and multi-planar actions act across 7 intervertebral joints. As a result, there is currently no method for measuring individual neck muscle strengths. Model-based parameter estimation techniques infer the values of the unobservable quantities by fitting measured data to a model. There is an opportunity to apply and expand these methods to aid clinicians and researchers in evaluation of individual muscle strengths in complex musculoskeletal systems. In a clinical setting, limiting the number of measurements is a priority due to time constraints and patient fatigue and pain. This thesis presents an active learning approach to parameter estimation that utilizes information theoretic measures to query the clinician for the next measurement that will maximize information gain relative to the unknown parameters, thus avoiding unnecessary measurements. This thesis is motivated by clinical questions that arose from imaging studies in individuals with persistent whiplash-associated disorders (WAD). Is there a link between compositional changes in the deep cervical extensor muscles and motor dysfunction in WAD? We begin with two studies confirming and expanding upon the compositional muscle changes and exploring the potential biomechanical consequences of those changes using a musculoskeletal model of the neck. The results illustrate the role of the deep extensors in multi-directional neck strength and provide simulated evidence of altered motor control patterns in WAD. The lack of an existing method for non-invasive measurement of individual muscle strengths motivates the main contribution of this thesis, a novel framework for active model-based Bayesian inference for muscle strength diagnostics. We demonstrate the utility of the approach for estimating individual neck muscle strengths from multi-directional isometric neck strength measurements from 5 healthy participants. A framework for a clinician-in-the-loop muscle strength estimator is presented, where in addition to accurately estimating the strength of the deep cervical extensor muscles, the algorithm predicts the benefit of taking additional measurements and selects the most informative next measurement to take. The framework is a step towards producing a clinically translatable test for individual neck muscle strengths.

Creator

• Abbott, Rebecca Elizabeth

DOI

• https://doi.org/10.21985/n2-ctyc-yg39

Subject

• Mechanical engineering
• Biomechanics

Language

• en

Alternate Identifier

• etdadmin_upload_797807
• http://dissertations.umi.com/northwestern:15487

Keyword

• Active Learning
• Parameter Estimation
• Musculoskeletal Model
• Cervical Spine

Date created

• 2021-01-01

Resource type

• Dissertation

Rights statement

• In Copyright