Towards a Clinically Viable Myoelectric Control for Transradial Wrist-Hand Prostheses

Interpreting the neurophysiological signals underlying voluntary motor control for driving limb prostheses represents a crucial, yet unsolved challenge in applied neuroscience and rehabilitation engineering. Individuals with a below-elbow (transradial) amputation maintain part of the original musculature that served the digits and wrist. This allows for electromyography (EMG) recorded from extrinsic muscles in the forearm to be used as inputs to control a multi-degree of freedom (DoF) wrist and hand prostheses. The purpose of multi-DoF prostheses is two-fold: to restore the lost functionality of the upper limb, and to reduce unnatural compensatory movements which may result in residual limb pain, secondary musculoskeletal complaints, and overuse syndromes. Ultimately, both of them would strongly increase the patients' quality of life. In clinical practice, the most widespread controller available is substantially a two-state amplitude modulation EMG controller, in which a single pair of agonist/antagonist muscles directly controls either the opening and closing of the different DoFs of the hand or the pronation and supination of the wrist. The different functions, however, are arranged in a sequential scheme, and selected through special switching signals/inputs (e.g. the agonist/antagonist muscles co-contraction), leading to a cumbersome and unnatural control. EMG pattern recognition represents a viable alternative to direct control. It is based on the premise that amputees can voluntarily activate repeatable and distinct muscular contractions for each class of motion and the associated EMG patterns can be identified to send different commands to the prosthesis. Although nowadays there are a few commercial products based on pattern recognition, their long-term stability (and thus their widespread adoption) has been hindered by both the stochastic nature of the EMG signal and the varying environmental conditions. Remarkably, the commonly used steady-state portion of the EMG signal is characterized by an active modification of recruitment and of the firing patterns needed to sustain the contraction. Conversely, the EMG signal associated with the onset of the myoelectric activity (i.e. the transient EMG), shows a clearer temporal structure, likely due to the orderly recruitment of the Motor Units. Starting from these considerations, this thesis proposed a novel EMG pattern recognition control strategy that interprets the transient portion of the EMG signal. In particular, the approach was tested offline and online on both non-amputee and amputee participants. Besides, a case study, with a patient provided with a transradial neuromusculoskeletal prosthesis, provides evidence of the potential clinical viability of the developed control strategy. Finally, this thesis proposed the design of a prosthetic wrist prototype with two degrees of compliance (stiff or compliant) that can be automatically switched by means of an external control input, associated with the closing/opening of the hand prosthesis. The main purpose of the prosthetic wrist, and the related control strategy, was to reduce user compensatory movement, without increasing the complexity of the overall controller. After a brief report on the state of the art which includes the scientific and technological bases of the work carried out, this thesis presents the research rationale and the resulting implementations that were designed and tested.