Neural network control of neuromuscular stimulation in paraplegics for independent ambulation

The paper discusses an ART-1-based artificial neural network (ANY) adapted to controlling functional electrical stimulation (FES) to facilitate patient-responsive ambulation by paralyzed patients with spinal cord injuries. This network is designed to control FES systems developed by the first author and that is presently in use by over 400 patients worldwide (presently without ANN control) and which is the first and the only FES system approved by FDA. The network that is considered discriminates patterns of above-lesion upper-trunk electromyographic (EMG) time series to map patient's posture for activating standing and walking functions under FES and it controls FES stimuli levels using response-EMG signals to overcome muscle fatigue. The neural network also adaptively controls patient's postural stability via identifying changes in posture through acceleration/gravitational/weight sensors. The network trains itself to adapt to physiological changes of the patient, and it overcomes decision and control errors by simple punishment inputs from a single manual punishment switch. The system thus, is both self adaptive and patient-responsive through a combinations of neural EMG signals and an artificial neural network to achieve patient responsive ambulation using stimulation of the patient's own peripheral motor neurons, namely his own peripheral neural network.