Improving Dexterity and Reliability of Restored Hand Movements Using a Brain-Machine Interface and Functional Electrical Stimulation

Abstract

Up to 60 percent of spinal cord injuries occur at the cervical level which often results in a reduction or even complete loss of hand function, severely impacting a person’s ability to interact with the world around them. In the past few decades, significant research has gone into using brain-machine interfaces (BMIs) to help restore independence in more severe cases where rehabilitative therapies are not effective. These interfaces can read out a user’s intentions, allowing them to control devices by thinking about movements rather than relying on residual movements. BMIs have especially gained traction in the last 20 years; clinical studies have used them to control cursors, robotic arms, synthesize speech, type text, and even to restore native limb function by controlling stimulation to paralyzed muscles. Stimulating muscles to restore functional movements is a therapy called functional electrical stimulation (FES), and brain-controlled FES presents a promising but challenging method to restore the full chain of hand control, bypassing damage to the spinal cord. However, this method has seen limited translation to clinical use, chiefly due to a lack of dexterity that it can restore in the hand and reliability with which it can be restored. The aim of this work is to inform how robust BMI are for the type of control required for brain-controlled FES, to assess to what extent FES can restore two-degree-of-freedom movements in the hand and what factors impact reliability, and then to extend FES methods to restoring movements to multiple joints in the hand.

In my first study, I investigated the impact that wrist postures and resistance at the fingers, two task perturbations occurring commonly in acts of daily living, have on a BMI for controlling virtual finger movements in non-human primates (NHP). I found that these changes impact cortical neural activity during the task, resulting in increased prediction error for intended movements and the NHP needing to adjust their control of the BMI. In my second study, I showed that current intramuscular FES methods can achieve graded control of simultaneous finger and wrist flexion as well as how well a BMI can control virtual finger and wrist movements. Additionally, I found that stimulation restored a large range of movements in both the wrist and the fingers, but the range of movements is significantly impacted by muscle fatigue and interactions between stimulation-evoked movements for each degree-of-freedom. In my third study, I present a proof-of-concept intramuscular FES implant method targeting multiple nerve entry points in each muscle. Using this method, stimulation evoked more discrete finger movements with individual electrodes, theoretically increasing the available hand postures that can be restored.

The results of this work demonstrate that BMI can be used to infer intended finger and wrist movements in real-time and FES can be used to control graded movements of the wrist and fingers. We found reduced efficacy in both BMI decoding accuracy and stimulated range of movements due to interactions controlling both the wrist and fingers. As future work continues to increase the degrees-of-freedom that are simultaneously controlled, there is a need to design stimulation protocols that account for interactions with stimulation on more electrodes and design BMI algorithms that are intentionally trained to generalize well, for example with variable postures and force requirements.