Comparison of knee loading during walking via musculoskeletal modelling using marker-based and IMU-based approaches

The current thesis is the result of the candidate's work over a six-month period with the assistance of the supervisor and co-supervisors, thanks to the collaboration between the Human Movement Bioengineering Laboratory Research group at the University of Padova (Italy) and the Human Movement Biomechanics Research group at KU Leuven (Belgium). Gait analysis, at a clinical level, is a diagnostic test with multiple potentials, in particular in identifying functional limitations related to a pathological path. Three-dimensional motion capture is now consolidated as an approach for human movement research studies and consists of a set of very precise measurements, the latter are processed by biomechanical models, and curves relating to the kinematics and indirect dynamics, i.e., the joint angles and the relative forces and moments, can be obtained. These results are considered fully reliable and based on these curves it is decided how to intervene on the specific subject to make the path as less pathological as possible. However, the use of wearable sensors (IMUs) consisting of accelerometers, gyroscopes, and magnetic sensors for gait analysis, has increased in the last decade due to the low production costs, portability, and small size that have allowed for studies in everyday life conditions. Inertial capture (InCap) systems have become an appealing alternative to 3D Motion Capture (MoCap) systems due to the ability of inertial measurement units (IMUs) to estimate the orientation of 3D sensors and segments. Musculoskeletal modelling and simulation provide the ideal framework to examine quantities in silico that cannot be measured in vivo, such as musculoskeletal loading, muscle forces and joint contact forces. The specific software used in this study is Opensim: an open-source software that allows modelling, analysis, and simulation of the musculoskeletal system. The aim of this thesis is to compare a marker-based musculoskeletal modelling approach with an IMUs-based one, in terms of kinematics, dynamics, and muscle activations. In particular, the project will focus on knee loading, using an existing musculoskeletal model of the lower limb. The current project was organized as follows: first, the results for the MoCap approach were obtained, following a specific workflow that used the COMAK IK tool and the COMAK algorithm to get the secondary knee kinematics, muscle activations, and knee contact forces. Where COMAK is a modified static optimization algorithm that solves for muscle activations and secondary kinematics to obtain measured primary DOF accelerations while minimizing muscle activation. Then these results were used to make a comparison with those obtained by the inertial-based approach, with the attempt to use as little information as possible from markers while estimating kinematics from IMU data using an OpenSim toolbox called OpenSense. Afterward, in order to promote an approach more independent from the constraints of a laboratory, the Zero Moment Point (ZMP) method was used to estimate the center of pressure position of the measured ground reaction forces (GRFs), and a specific Matlab code was implemented to improve this estimation. Using the measured GRFs with the new CoPs, the results of Inverse Dynamics, muscle activations, and finally knee loading were calculated and compared to the MoCap results. The final step was to conduct a statistical analysis to compare the two approaches and emphasize the importance of using IMUs for gait analysis, particularly to study knee mechanics.