Prediction of wrist angle from sonomyography signals with artificial neural networks technique

Surface electromyography (SEMG) is widely used for the functional assessment of skeletal muscles, while sonography has been commonly used to detect its morphological information. We defined the signal about the continuous change of the morphological parameters of muscles detected by ultrasound as sonomyography (SMG). In this study, we continuously sampled the ultrasound image, SEMG signals on the extensor carpi radialis muscle together with the wrist angle simultaneously during the whole process of wrist extension and flexion from 7 normal subjects. A three-layer feed-forward artificial neural network with BP learning algorithm was used to predict the wrist angle with the muscle deformation SMG and root mean square of SEMG signals as inputs. The overall mean R value was 0.96 +/- 0.02, the mean standard root mean square error was 7.26 +/- 1.98, and the mean relative root mean square errors was 0.160 +/- 0.037. The results demonstrated that the wrist angle could be well predicted by combining the SMG and SEMG signals with ANN. Our result suggested that the combination of the information of SMG and SEMG could provide more comprehensive assessment of the skeletal muscle.