HCC: Medium: A novel neural interface for user-driven control of rehabilitation of finger individuation

Following a stroke incident, a majority of stroke survivors lose the ability to use their hand to perform a variety of tasks despite months of therapy. In an effort to restore hand dexterity, advanced assistive devices (e.g., exoskeletons) have been developed. Unfortunately, only few of these novel devices have been used effectively by stroke survivors. One critical factor limiting user acceptance is the lack of reliable method that allows stroke survivors to intuitively control the device. The overarching objective of the project is to combine novel decoding of neurological signals that drive the muscles with a personalized musculoskeletal model of the upper limb to provide intuitive control of an assistive hand exoskeleton. The control strategy will be robust in handling different arm postures and movements. This personalized approach will improve hand functional performance in stroke survivors, with the overall goal of improving their ability to live independently. The computational approaches employed here will also produce a research tool to study human-robot interactions. The researchers will make the computational model available over online repository system, SimTK.org, as a simulation platform for other researcher working on hand function and control of rehabilitative devices. The project will provide educational and training opportunities. The research concepts will be integrated into existing courses. Summer projects incorporating the techniques will be offered to undergraduate and high school students and local school and community college instructors. Outreach programs will be developed to disseminate the proposed research outcomes to underrepresented students.

The goal of this project is to develop a personalized hybrid (neural data-based and model-based) interface that combines the decoded neural command with a musculoskeletal model. The developed interface will be used to control a soft-hard hybrid exoskeleton to enable dexterous finger movements in stroke survivors. The research team will first develop a real-time neural decoding algorithm based on populational firing probability of the motoneurons, extracted from motor unit decomposition of high-density electromyographic (HD-EMG) signals. Through incorporation of binary neuron discharge events, the decoded neural drive signals will be robust to changes in muscle activity features, background noise, and motion artifact. The research team will then employ a personalized musculoskeletal model of the limb, which will be calibrated to the unique musculoskeletal structure and activation parameters of stroke survivors. The model-based controller will be able to compensate for limb posture, movement dynamics, and subject-specific impairments that could otherwise disturb the mapping between user input and desired output. Finally, the research team will evaluate the developed interface for control of an advanced hand exoskeleton, allowing users to control flexion or extension assistance independently for each digit. The assistive forces will reinforce beneficial muscle activation while compensating for abnormal activation patterns. Collectively, the outcomes will restore hand dexterity in stroke survivors, thereby enabling them to perform daily activities and live independently.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.