High-dosage rehabilitation therapy enhances neuroplasticity and motor recovery after neurologic injuries such as stroke and spinal cord injury. The optimal exercise dosage necessary to promote upper extremity (UE) recovery is unknown. However, occupational and physical therapy sessions are currently orders of magnitude too low to optimally drive recovery. Taking therapy outside of the clinic and into the living environment using sensing and computer technologies is attractive because it could result in a more cost efficient and effective way to extend therapy dosage. This dissertation developed innovative wearable sensing algorithms and a novel robotic system to enhance hand rehabilitation. We used these technologies to provide on-demand exercise in the living environment in ways not previously achieved, as well as to gain new insights into UE use and recovery after neurologic injuries.
Currently, the standard-of-practice for wearable sensing of UE movement after stroke is bimanual wrist accelerometry. While this approach has been validated as a way to monitor amount of UE activity, and has been shown to be correlated with clinical assessments, it is unclear what new information can be obtained with it. We developed two new kinematic metrics of movement quality obtainable from bimanual wrist accelerometry. Using data from stroke survivors, we applied principal component analysis to show that these metrics encode unique information compared to that typically carried by conventional clinical assessments. We presented these results in a new graphical format that facilitates the identification of limb use asymmetries.
Wrist accelerometry has the limitation that it cannot isolate functional use of the hand. Previously, we had developed a sensing system, the Manumeter, that quantifies finger movement by sensing magnetic field changes induced by movement of a ring worn on the finger, using a magnetometer array worn at the wrist. We developed, optimized, and validated a calibration-free algorithm, the “HAND” algorithm, for real-time counting of isolated, functional hand movements with the Manumeter. Using data from a robotic wrist simulator, unimpaired volunteers and stroke survivors, we showed that HAND counted movements with ~85% accuracy, missing mainly smaller, slower movements. We also showed that HAND counts correlated strongly with clinical assessments of hand function, indicating validity across a range of hand impairment levels.
To date, there have been few attempts to increase hand use and recovery of individuals with a stroke by providing real-time feedback from wearable sensors. We used HAND and the Manumeter to perform a first-of-its-kind randomized controlled trial of the effect of real-time hand movement feedback on hand use and recovery after chronic stroke. We found that real-time feedback on hand movement was ineffective in increasing hand use intensity and improving hand function. We also showed for the first time the non-linear relationship between hand capacity, measured in the laboratory, and actual hand use, measured at-home. Even people with a moderate level of clinical hand function exhibit very low hand use at home.
Finally, the challenge of improving hand function for people with moderate to severe injuries highlights the need for novel approaches to rehabilitation. One emerging technique is regenerative rehabilitation, in which regenerative therapies, such as stem cell engraftment, are coupled with intensive rehabilitation. In collaboration with the Department of Veteran Affairs Gordon Mansfield Spinal Cord Injury Translational Collaborative Consortium, we developed a robot for promoting on-demand, hand rehabilitation in a non-human primate model of hemiparetic spinal cord injury that is being used to synergize hand rehabilitation with novel regenerative therapies. Using an innovative bimanual manipulation paradigm, we show that subjects engaged with the device at a similar rate before and after injury across a range of hand impairment severity. We also demonstrate that we could shape relative use of the arm and increase the number of exercise repetitions per reward by changing parameters of the robot. We then evaluated how the peak grip force that the subjects applied to the robot decreased after SCI, demonstrating that it can serve as a potential marker of recovery.
These developments provide a foundation for future work in technologies for therapeutic movement rehabilitation in the living environment by establishing: 1) new metrics of upper extremity movement quality; 2) a validated algorithm for achieving a “pedometer for the hand” using wearable magnetometry; 3) a negative clinical trial result on the therapeutic effect of real-time hand feedback after stroke, which begs the question of what can be improved in future trials; 4) the nonlinear relationship between hand movement ability and at-home use, supporting the concept of learned non-use; and 5) the first example of robotic regenerative rehabilitation.