Training, retaining and transferring novel myoelectric skills for prosthetic hand control

Abstract

Recent advancements in robotics have led to the creation of highly dexterous multi-articulating hands that can mimic human capabilities. Despite this achievement, high rates of prosthesis abandonment persist in the field of upper-limb prosthetics, with control issues and limited functionality often cited as key reasons. Although various solutions have been attempted, there remains a critical gap between the capabilities of modern artificial hands and the means to effectively operate them. Until prosthetic systems can perfectly interpret user intent, prosthesis control will always involve an aspect of motor-learning, as users naturally adapt their motor behaviour when they encounter an error. Various studies have inadvertently provided evidence that human motor learning actively compensates for inaccuracies in prosthetic systems. However, serious exploration of the impact of the human element within the control loop has only recently begun. This thesis investigates the integration of neuroscience and motor learning principles to enhance prosthesis control and system robustness. Moreover, it emphasises the often-overlooked role of user learning in optimising pre-device training for improved control experiences. Four studies form the core contributions of this work. Two multi-day studies, one lab-based and one home-based, comprise the first two studies that analyse the effects of different feedback mechanisms on the permanency of learned myoelectric skills. In total, ∼35,000 trials were collected, yielding one of the largest myoelectric datasets in the field. The findings highlight the importance of utilising appropriate feedback mechanisms during user learning and provides a novel method of myoelectric training that leads to improved skill permanency in the absence of artificial feedback. The third study focused on the transferability of these findings to actual prosthesis use, which showed improved prosthesis control following myoelectric training with delayed feedback, highlighting the method’s efficacy in pre-device rehabilitation. The fourth study examined the impact of arm position changes on muscle activity, showcasing the benefits of delayed feedback training in enhancing muscle activation consistency which generalised to untrained positions. These findings offer the field a novel tool for combatting the limb position effect. Collectively, they underscore the potential of human motor learning to optimise rehabilitation protocols and enhance prosthesis control performance.