Prediction of the 3d ground reaction force during rollator supported and unsupported gait in old persons using artificial neural networks

During the last years, in-field motion analysis has gained more and more importance. The development of inertial sensors for motion analysis enables the measurement of motion kinematics in any environment. Although being of high importance for clinical and sports related analyses, the analysis of kinetic motion parameters is still challenging. Therefore, this study aims to apply a feedforward neural network to estimate the 3D ground reaction force using the acceleration and angular rate of both feet as input data. The data was collected from healthy elderly as well as geriatric inpatients with and without the support of a rollator. The results show a high accuracy in predicting the vertical and anterior-posterior ground reaction force components, but a lower accuracy in the medio-lateral component. Nevertheless, clinically relevant conclusions on the loading during gait can be drawn.