Knee kinesiography is a technology developed at the Laboratoire d’Innovation Ouverte (LIO) to identify and quantify dynamic biomechanical markers of three-dimensional knee motion during gait. The technology is built around the KneeKG system (Emovi Inc., Canada), a device that reduces soft-tissue artifacts by up to six-fold compared to skin-mounted markers, paired with an optoelectronic motion capture system and a functional calibration procedure. While optoelectronic systems remain the gold standard for lower limb kinematic analysis, their cost, infrastructure requirements and limited portability hinder widespread clinical use. Magnetic and inertial measurement units (MIMUs) represent a promising alternative, offering affordable cost, high portability and a broader acquisition field. However, their clinical use remains limited, primarily due to the lack of a standardized calibration approach and insufficient accuracy in the frontal and transverse planes. The goal of this thesis is to develop and validate a MIMU-based calibration method for three-dimensional knee kinematics assessment during gait using the KneeKG system.
The first study constituted the analytical starting point of the research. Prior to developing a new calibration method, we examined an existing double-pose calibration method in depth to identify its limitations. Specifically, we examined the sensitivity of three-dimensional knee kinematics to variations in lower limb positioning during the second posture of a double-pose calibration method, in ten healthy participants. A leg deviation from the sagittal plane when performing the second calibration posture was found to affect knee kinematics primarily in the frontal and transverse planes, with adduction/abduction being the most sensitive. Importantly, larger deviations of the leg from the sagittal plane during the second calibration posture did not systematically produce larger errors, owing to inter-individual variability in tibia-femur relative orientation across conditions. These findings identified the crosstalk phenomenon, induced by an imprecise definition of the sagittal plane, as a partial source of the frontal and transverse plane errors reported in the literature, and directed subsequent studies toward a functional calibration approach. Rather than relying on static postures, the approach developed in this thesis constructs femoral and tibial longitudinal axes from the estimated positions of the hip, knee and ankle joint centers, obtained from functional movements.
Accurate joint center estimation is therefore a prerequisite. The second study focused on identifying the optimal MIMU configuration and functional movement for hip joint center (HJC) estimation, as the first building block of the femoral anatomical coordinate system. Twenty healthy participants performed five functional movements drawn from the literature, with varying numbers of MIMUs per segment (one or two) and with or without inclusion of pelvic motion alongside femoral motion. The optimal configuration relied on a single femoral MIMU during a circumduction movement, yielding the most accurate HJC estimates and demonstrating that minimal sensor instrumentation can achieve clinically acceptable joint center localization.
Before the final calibration method could be implemented, an intermediate methodological issue required attention: a misalignment between the individual reference frames in which each MIMU expresses its orientation. Left uncorrected, this misalignment propagates into joint angle estimates. The issue was resolved by re-estimating MIMU orientations from raw inertial data using an Attitude and Heading Reference System (AHRS) filter, tuned on a carefully selected reference trial and applied consistently across all sensors and sessions.
The third and final study brought these contributions together into a complete sensor-tosegment calibration (SSC) procedure for three-dimensional knee kinematics assessment during gait. The method estimates hip, knee and ankle joint centers to define femoral and tibial longitudinal axes, as well as a neutral knee posture through functional movements. Twenty healthy participants were instrumented with the KneeKG equipped with five MIMUs. For each joint center, the optimal movement and minimal sensor configuration were identified. The resulting two-MIMU setup, consisting of one femoral and one tibial MIMU, was then used to assess knee kinematics during treadmill walking against the optoelectronic reference. Root mean square errors (RMSE) remained below the 5° clinical threshold across all three anatomical planes, and fell below 2.5° after removal of the mean inter-system offset, demonstrating a clinically acceptable level of accuracy.
Together, these studies contribute to the development of a clinically applicable inertial calibration method for knee motion analysis. In the long term, such an approach would enable ambulatory gait assessment in real-world settings, facilitating early detection of knee pathologies and optimization of their rehabilitation management. It would also contribute to reducing the instrumentation and infrastructure requirements of knee kinesiography, supporting its broader adoption by the clinical community and ultimately improving the quality of care for patients with knee conditions such as osteoarthritis.
Di Falco, C. (Author),
Hagemeister (Supervisor) &
Aissaoui (Co-supervisor),
1 Aug 2026Student thesis: Doctoral thesis › Doctorate in Engineering: Engineering