Predicting Shoulder Joint Reaction Forces From 3D Body Kinematics: A Convolutional Neural Network Approach

This study aimed to develop a convolutional neural network (CNN) model to predict shoulder joint reaction forces from 3D body kinematics. Results showed a good convergence between CNN model prediction and musculoskeletal model estimation for six novel tasks. Therefore, a CNN-based deep learning model can be used as a simple and relatively less time- and labor-intensive method to identify unsafe shoulder exertions in order to prevent the incidence of shoulder injuries or pathologies in occupational settings.