Detection and prediction of FES-induced fatigue.

The estimation of externally elicited muscle forces is important for the better control of a functional electrical stimulation (FES)-assistive system. Various techniques of signal processing are presented, all with only one aim, to determine the correlation between the decrease of muscle force after continuous stimulation and surface recordings of the evoked potentials. Wrist flexor muscles were stimulated under isometric conditions, and surface electromyography (sEMG) was used to record wrist joint torque in both able-bodied and spinal cord injured volunteers. The joint torque was determined from recordings of the force generated by the wrist flexors, with the forearm immobilized. The sEMG was recorded utilizing a preamplifier with a stimulation artefact suppression circuitry. The signal was processed in the time and frequency domains, and analysed vs time, as well as in the state space formed by the wrist torque and evoked potential. The torque vs sEMG curves were used to establish the relationship that can be used for detection of the decrease of the force associated with FES-induced muscle fatigue. Among seven different techniques of sEMG processing the best correlation was found between the median frequency and force changes. The phase plane plot was fitted with an exponential curve, and the parameters obtained from the fitting were used to determine two events: prediction of the onset of fatigue and detection of fatigue. This suggests that it is possible to use the processed sEMG as a trigger signal to change the pattern of stimulation and allow the muscle to recover while resting, or to inform the user that the muscle force will soon drop rapidly. The recovery of the muscle force and sEMG was also analysed to learn more about the mechanisms that may be responsible for FES-induced fatigue. This technique offers simple on-off type feedback capability for fatigue detection in FES applications.