Machine Learning for Personalization of Musculoskeletal Models, Movement Analysis, and Movement Predictions (C01)

我们的目标是研究步态模拟是否可以有效地使用基于运动数据的神经网络进行个性化。我们首先研究了车身参数对仿真的影响。一个初始的个性化网络是在模拟的运动数据上训练的，因为一个基本的事实是已知的。然后，我们研究了无梯度方法，利用实验运动数据进一步训练网络。所得到的网络通过磁共振成像、肌电图和体内变量进行验证。

We aim to research if gait simulations can effectively be personalized using neural networks based on movement data. We investigate first the effect of body parameters on the simulations. An initial personalization network is trained on simulated movement data because a ground truth is known. Then we research gradient-free methods to further train the network with experimental movement data. The resulting network is validated with magnetic resonance imaging, electromyography, and internal body variables.