Biomechanical Insights into Neural Control of Movement

Lower-limb Joint Torque Prediction using LSTM Neural Networks and Transfer Learning

Estimation of joint torque during movement provides important information in several settings, such as effect of athletes' training or of a medical intervention, or analysis of the remaining muscle strength in a wearer of an assistive device. The ability to estimate joint torque during daily activities using wearable sensors is increasingly relevant in such settings. In this study, lower limb joint torques during ten daily activities were predicted by long short-term memory (LSTM) neural networks and transfer learning. LSTM models were trained with muscle electromyography signals and lower limb joint angles. Hip flexion/extension, hip abduction/adduction, knee flexion/extension and ankle dorsiflexion/plantarflexion torques were predicted. The LSTM models' performance in predicting torque was investigated in both intra-subject and inter-subject scenarios. Each scenario was further divided into intra-task and inter-task tests. We observed that LSTM models could predict lower limb joint torques during various activities accurately with relatively low error (root mean square error ≤ 0.14 Nm/kg, normalized root mean square error ≤8.7%) either through a uniform model or through ten separate models in intra-subject tests. Furthermore, a transfer learning technique was adopted in the inter-task and inter-subject tests to further improve the generalizability of LSTM models by pre-training a model on multiple subjects and/or tasks and transferring the learned knowledge to a target task/subject. Particularly in the inter-subject tests, we could predict joint torques accurately in several movements after training from only a few movements from new subjects.

Kinematic and EMG determinants in quadrupedal locomotion of a non-human primate (Rhesus)

We hypothesized that the activation patterns of flexor and extensor muscles and the resulting kinematics of the forelimbs and hindlimbs during locomotion in the Rhesus would have unique characteristics relative to other quadrupedal mammals. Adaptations of limb movements and in motor pool recruitment patterns in accommodating a range of treadmill speeds similar to other terrestrial animals in both the hindlimb and forelimb were observed. Flexor and extensor motor neurons from motor pools in the lumbar segments, however, were more highly coordinated than in the cervical segments. Unlike the lateral sequence characterizing subprimate quadrupedal locomotion, non-human primates use diagonal coordination between the hindlimbs and forelimbs, similar to that observed in humans between the legs and arms. Although there was a high level of coordination between hind- and forelimb locomotion kinematics, limb-specific neural control strategies were evident in the intersegmental coordination patterns and limb endpoint trajectories. Based on limb kinematics and muscle recruitment patterns, it appears that the hindlimbs, and notably the distal extremities, contribute more to body propulsion than the forelimbs. Furthermore, we found adaptive changes in the recruitment patterns of distal muscles in the hind- and forelimb with increased treadmill speed that likely correlate with the anatomical and functional evolution of hand and foot digits in monkeys. Changes in the properties of both the spinal and supraspinal circuitry related to stepping, probably account for the peculiarities in the kinematic and EMG properties during non-human primate locomotion. We suggest that such adaptive changes may have facilitated evolution toward bipedal locomotion.