In the paper “Synthesising Motion Sensor Data from Biomechanical Simulations to Investigate Motion Sensor Placement and Orientation Variations”, we investigate for the first time effects of sensor positioning and orientation variation using biomechanical simulations and motion data synthesis. Our novel approach enables the synthesis of motion data from hundreds of virtual sensors to investigate how sensor positioning and orientation variation influence the estimation of the clinical Fugl-Meyer-Assessment (FMA) score. In particular, we used walking data from hemiparetic patients after stroke in our biomechanical simulation and derived features from synthesised acceleration data. Subsequently, we investigated the score estimation accuracy using an exhaustive feature selection approach and analysed differences between the generalised linear model and a support vector regression with an RBF kernel in a leave-one-participant-out evaluation.
We propose a motion sensor data synthesis approach to investigate the performance effect of sensor placement and orientation variation on health marker estimation. Using OpenSim we simulate walking motion of patients after stroke while synthesising inertial sensor data. We defined 384 sensor positions with 192 sensors simulated at each leg’s thigh. To demonstrate how synthesised sensor data could be used to analyse the performance of functional ability estimation, we derived scores from Lower-Extremity Fugl-Meyer Assessment (LE-FMA) using regression methods. We evaluated our approach using a public dataset, including 8 stroke patients and showed that LE-FMA scores could be estimated with an error below 0.12 points on average, compared to manually derived scores. We further show that sensors should preferably be placed at the front of the thigh. Our approach demonstrates the potential of combining biomechanical simulations and acceleration synthesis with algorithms for health marker estimation, thus providing rapid insight into sensor positioning and orientation variation.
Visualisation of hemiparetic walking
The video below illustrates the biomechanical walking simulation based on motion-captured data of a patient after a stroke. Virtual sensors, here positioned on both thighs, shins, and feet, are highlighted in blue.
The paper will be presented at the EBMC 2019 conference in Berlin during July 23-27 and published in the Proceedings of the 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’19).
ID: EZ8ISUSI not found
Full texts will soon be available from our publication page.