Development and validation of a novel adaptive assist-as-needed controller for robot-assisted rehabilitation of the upper-extremity following stroke
University of New Brunswick
Rehabilitation plays an important role in regaining strength and functionality following neurological injuries such as stroke. Robot-assisted therapies can be an effective and efficient mode of rehab delivery; however, robots are susceptible to helping too much or too little, both of which may impede the motor recovery process. To overcome this deficiency, the main objectives of this thesis were: 1) to develop and implement a novel adaptive assist-as-needed (aAAN) algorithm to control a haptic robot for patient-specific upper-extremity training, and 2) to establish proof-of-concept that the novel controller has the potential to improve motor recovery in people with stroke. The real-time aAAN controller gathers user performance data (EMG and velocity) and adapts the robot motors to either assist or challenge the participant, depending on their training progress over time. In this research, the effectiveness of the training program was quantified in two ways. First, the interaction between user and robot for a single (naïve) training session in five participants with stroke was studied, and their performance and brain activity measures (EEG) were validated against their sensorimotor score (Fugl-Myer score). Second, performance and brain activity measures were evaluated in two participants with stroke - one high functioning and one low functioning - over repeated sessions (2x/week) for one month. The cross-sectional results showed strong second-order relationships between Fugl-Myer score and outcome measures. Performance metrics (path length and accuracy) were sensitive to change in participants with lower sensorimotor score. In comparison, speed and electrophysiological metrics (EMG and EEG) were more susceptible to change in participants with higher sensorimotor status. The longitudinal case results showed that changes over time, although slight, were consistent with the sensitivity analysis, allowing us to conclude that the training paradigm had an observable effect on motor learning processes and could be attributed to the aAAN controller's ability to adapt according to the participant's progress.