Inertial sensor-based joint kinematic estimation of the lower limb

A growing interest is visible in small, inexpensive skin-attached inertial measurement units (IMUs) to capture human movements outside a conventional laboratory environment. IMUs do not provide movement information out of the box. Moreover, when attached on the skin their connection with the underlying position of the bones is unknown. In this doctoral project we aimed to combine noisy sensor measurements to estimate the relative movement of skin-attached IMUs during human motion and to identify their relation with the underlying anatomical structures.

A systematic methodological literature search was carried out to summarized the current state-of-the-art in estimating joint kinematics with inertial sensors. The study was written to reach a broad audience from engineers to clinicians and shows that methods for lower limb joint kinematic estimation are inherently application dependent. Sensor restrictions are generally compensated with biomechanically inspired assumptions and prior information. Awareness of the possible adaptations in the IMU-based kinematic estimates by incorporating such prior information and assumptions is necessary, before drawing clinical decisions.

From this literature review a novel joint sensor orientation estimation method was proposed that tightly incorporates the connection between adjacent segments within a sensor fusion algorithm, to obtain drift-free joint kinematics. Drift in the sensor orientations is eliminated solely by utilizing common information in the accelerometer and gyroscope measurements of sensors placed on connecting segments. The method was experimentally validated on a robotic manipulator under varying measurement durations and movement excitations and proven to be applicable in biomechanics, with a prolonged gait trial on healthy subjects. The proposed method does not explicitly model a gravity component as in conventional orientation estimation algorithms but works on the specific force (consisting of both linear acceleration and gravity). This biomechanically inspired constraint replaces magnetometer measurements such that the method can be used in any environment for arbitrary human movements.  

A translation of these relative sensor orientations from skin-attached IMUs, to the underlying anatomical structures is necessary to compare kinematics between subjects or trials. Inertial measurements are often related with a model of the joint movement to overcome possible calibration movements or sensor placement assumptions. However, it is unclear how good such alignment methods can identify the anatomical axes. Therefore, we presented an anatomically correct ground-truth reference from dynamic motions on a cadaveric knee to enable a true evaluation by overcoming common disturbances in the available reference from an optical motion capture system, e.g., soft-tissue movements of the skin and underlying musculature, palpation and orientation errors. The dataset is further used as a ground-truth reference to validated existing inertial-sensor-to-bone-alignment methods. With the currently available model-based alignment methods, a first dominant sagittal rotation axis aligned with the underlying anatomical reference in joints that predominantly behave around in one movement plane. To quantify secondary movements, the inertial-sensor-to-bone alignment model should better match the biomechanics of the joint.

In conclusion, the present dissertation contributed to a better understanding on inertial sensor orientation estimation and inertial-sensor-to-bone alignment strategies for the application of inertial movement analysis.