IMU-Based Lower-limb joint angles: A comparison of methods

Abstract

Inertial measurement units (IMUs) are a popular option for human movement analysis. The untethered, self-contained nature of IMUs overcomes many limitations of conventional measurement systems. The potential of IMU systems makes it worthwhile to pursue clinical and research use. However, IMUs have not proven to be sufficiently reliable or valid. Two barriers facing IMU-based joint kinematics are: (i) the misaligned, unique reference frames of each IMU in the system, hindering joint angle calculation, and (ii) anatomical calibration accuracy and reproducibility, hindering the anatomical relevance of joint angles. A comparison of available methods would help to understand and overcome the current barriers preventing IMU use. The present thesis aimed to provide these comparisons. Several methods have been proposed to align coordinate frames. Three methods were compared mathematically and experimentally. The equivalency of all methods was proved mathematically. Experimentally, all three methods were equivalent (<2° different) in two applications relevant to biomechanics (finding a common IMU reference frame and comparing the IMU orientation to a marker-based orientation). Several methods have also been proposed to find anatomically relevant axes of the lower limb body segments. The joint angles from five methods were compared using the joint angles of a marker-based method as reference. The methods were used for the hip, knee and ankle joint, if they were applicable. The joint angles from three of the methods were similar, while two methods had some joint angles that differed, primarily by a bias. The two dissimilar methods relied on static-normalization, which caused the errors, particularly in the transverse plane angles. Drift (degradation of IMU accuracy over time) between trials was the problem affecting the static-normalization, so it was the IMU sensor fusion and not the method itself that was the cause of dissimilarity. Further research is required to recommend one method for future use. Overall, current methods performed similarly in both methodological options, suggesting that current research is reaching a plateau in improvements. Further research in reliability and agreement is required to understand the strengths, weaknesses and fields of improvement required for research and clinical use of IMUs in human movement analysis.