Frontiers in Bioengineering and Biotechnology (Jul 2021)
The Influence of Kinematic Constraints on Model Performance During Inverse Kinematics Analysis of the Thoracolumbar Spine
Abstract
Motion analysis is increasingly applied to spine musculoskeletal models using kinematic constraints to estimate individual intervertebral joint movements, which cannot be directly measured from the skin surface markers. Traditionally, kinematic constraints have allowed a single spinal degree of freedom (DOF) in each direction, and there has been little examination of how different kinematic constraints affect evaluations of spine motion. Thus, the objective of this study was to evaluate the performance of different kinematic constraints for inverse kinematics analysis. We collected motion analysis marker data in seven healthy participants (4F, 3M, aged 27–67) during flexion–extension, lateral bending, and axial rotation tasks. Inverse kinematics analyses were performed on subject-specific models with 17 thoracolumbar joints allowing 51 rotational DOF (51DOF) and corresponding models including seven sets of kinematic constraints that limited spine motion from 3 to 9DOF. Outcomes included: (1) root mean square (RMS) error of spine markers (measured vs. model); (2) lag-one autocorrelation coefficients to assess smoothness of angular motions; (3) maximum range of motion (ROM) of intervertebral joints in three directions of motion (FE, LB, AR) to assess whether they are physiologically reasonable; and (4) segmental spine angles in static ROM trials. We found that RMS error of spine markers was higher with constraints than without (p < 0.0001) but did not notably improve kinematic constraints above 6DOF. Compared to segmental angles calculated directly from spine markers, models with kinematic constraints had moderate to good intraclass correlation coefficients (ICCs) for flexion–extension and lateral bending, though weak to moderate ICCs for axial rotation. Adding more DOF to kinematic constraints did not improve performance in matching segmental angles. Kinematic constraints with 4–6DOF produced similar levels of smoothness across all tasks and generally improved smoothness compared to 9DOF or unconstrained (51DOF) models. Our results also revealed that the maximum joint ROMs predicted using 4–6DOF constraints were largely within physiologically acceptable ranges throughout the spine and in all directions of motions. We conclude that a kinematic constraint with 5DOF can produce smooth spine motions with physiologically reasonable joint ROMs and relatively low marker error.
Keywords
