Neural network-based muscle torque prediction using mechanomyography during electrically-evoked knee extension and standing in spinal cord injured patients / Muhammad Afiq Dzulkifli

This study sought to design and deploy a torque monitoring system using an artificial neural network (ANN) with mechanomyography (MMG) for situations where torque cannot be independently quantified. The MMG signals from the quadriceps were used to derive muscle torques during prolonged functional electrical stimulation (FES) assisted isometric knee extension contractions and during standing in spinal cord injured (SCI) individuals. Three individuals with motor-complete SCI performed FES-evoked isometric quadriceps contractions on a Biodex dynamometer at 30⁰ knee angle and 100mA stimulation current until the torque declined to a minimum required for ANN model development. Two ANN models were developed based on two different inputs; RMS and RMS-ZC. The performance of the ANN was evaluated by comparing its predicted torque against the actual torque derived from the dynamometer. MMG data from 5 other individuals with SCI who performed FES-evoked standing to fatigue (i.e. until the knee angle reached 30⁰ flexion), were used to test the RMS and RMS-ZC ANN. RMS and RMS-ZC obtained from the FES standing experiments were then provided as inputs to the developed ANN models to determine the predicted torque during the FES-evoked standing. The average correlation between the knee extension predicted torque and the actual torque outputs were 0.87±0.11 for RMS and 0.84± 0.13 for RMSZC. The average accuracies for predicting 50% torque drop for both models were 79±14% for RMS and 86±11% for RMS-ZC. The two models revealed significant trends in torque decrease, both suggesting a critical point at 50% torque drop where there were significant changes observed in RMS and ZC trends. Based on these findings, it can be concluded that both RMS and RMS-ZC models performed similarly well in predicting knee extension torque in this population. However, interference was observed in the ZC values towards the end of the knee buckling. The developed ANN model could be used to predict muscle torque in real-time thereby providing possibly safer automated FES control of standing in persons with motor-complete SCI