Automated quantification of the anatomic accuracy of muscle paths and its application in an image-based subject-specific modeling workflow for adult spinal deformity

Background

Musculoskeletal models (MSKM) can non-invasively evaluate the effect of altered muscle geometry and physiology on locomotor function. Nevertheless, this requires anatomically accurate muscle paths throughout functional ranges of motion. However, reference data to evaluate such muscle paths is rarely available, including in adult spinal deformity (ASD). Although this degenerative disorder alters muscle geometry and physiology, these parameters are not available to inform clinical decision-making as generic MSKM cannot account for these alterations, and reliable ASD-specific modeling workflows do not currently exist.

Research question

Can an efficient workflow be developed for evaluating and optimizing the anatomic accuracy of dynamic muscle paths in personalized MSKM and applied for reliable spinal motion simulations in ASD?

Methods

A workflow was developed to automatically analyze anatomic muscle accuracy throughout predefined ranges of motion in terms of muscle-bone penetration, muscle action, moment arm magnitude, and discontinuities. Erector spinae, multifidus, and psoas muscles were semi-automatically segmented in magnetic resonance images of one healthy and two ASD subjects. Next, their muscle representation, insertion sites, and complexity were iteratively refined with the above workflow to generate subject-specific MSKM and compare them against state-of-the-art generic MSKM.

Results

All muscles in the subject-specific MSKM were anatomically accurate, except for discontinuities in 3.81 % (psoas) and 0.37 % (multifidus) of moment arm curves across motions and subjects. In contrast, scaled generic MSKM were consistently associated with muscle-bone penetration, decreased moment arm magnitude, and opposite muscle actions.

Significance

This novel workflow is the first to allow for an efficient evaluation of the anatomic accuracy of dynamic muscle paths. Its application in MSKM of ASD patients resulted in subject-specific muscle paths, with an anatomically correct muscle geometry, while preventing bone penetration during the representative range of motions. The workflow is promising to enable biomechanical analyses of ASD with an accuracy beyond that of scaled generic models.