There are over 33,000 people in the United States living with complete tetraplegia due to traumatic spinal cord injury (SCI). These individuals rely heavily on family and caregivers as they are unable to perform many activities of daily living.People with complete tetraplegia rank restoration of hand and arm function as their highest priority, as it would offer greater independence and improved quality of life. In this study, we show that subjects with chronic (>1-year post-injury) C5/C6-level, motor-complete SCI are able to control a brain computer interface-functional electrical stimulation (BCI-FES) system to perform a hand grasp and release task.
Electroencephalographic (EEG) signals were acquired using a 20-channel wireless EEG system and input to a BCI, which enabled autonomous control over FES of paralyzed muscles for hand grasp and release. A novel stimulation configuration and control paradigms were developed in order to provide reliable activation of the muscles responsible for hand movements. Input features and decoding strategies were evaluated from subjects with SCI, as well as uninjured, control subjects.
After optimization of the BCI-FES system and experimental paradigm, 5 subjects with C5/C6, motor complete spinal cord injury and 5 uninjured, control subjects participated in 6 sessions of closed-oop BCI-FES. Subjects were asked to imagine opening and closing their right hand during the trials for motor imagery. Average power in 5 Hz bins(5-35 Hz) was extracted from C3, C1, Cz, C2, and C4 electrodes and input as features to a Support Vector Machine classification algorithm.
When “movement intention” was classified correctly from the motor imagery period, a custom stimulation sequence was delivered to the forearm muscles via surface electrodes to enable opening and closing of the hand for grasp and release. Spinal cord injured subjects produced an average of 21.0% ± 3.9% event-related desynchronization and control subjects averaged 13.5% ± 3.2%. Average decoding accuracy was similar, at 73.3% ± 5.6% in the spinal cord injury group and 73.6% ± 3.8% in the control group.
Over the course of experiments, average event-related desynchronization increased significantly in the SCI group and decoding accuracy improved. This study demonstrates that subjects with motor complete, cervical SCI were able to control a BCI-FES system with performance levels as high as healthy controls with minimal training. Non-invasive BCI-FES systems may have the potential to restore hand function in people with motor-complete SCI, which would increase independence and improve quality of life.
Subject Recruitment and Screening:
Subjects with SCI were recruited from The Miami Project to Cure Paralysis database. To be considered for participation in any phase of the study, a potential SCI subject had to meet the following inclusion criteria: 18-50 years old, chronic injury (longer than 1 year) but no more than 15 years post-injury, C5 or C6-level motor complete SCI as classified by ISNCSCI standards, and no extensive denervation of target muscles. The excitability and strength of paralyzed hand muscles were assessed during the first visit, by applying surface stimulation.
Neural Data Acquisition:
Two different wireless EEG systems were used to acquire EEG signals from the subjects. Either a 9-channel (Fz F3, F4, Cz, C3, C4, P3, P4, POz, X10 headset) or 20-channel (Fz, F1, F2, F3, F4, Cz, C1, C2, C3, C4, CPz, Pz, P1, P2, P3, P4, POz, Oz, O1, O2, X24 headset) EEG system (256 Hz sampling rate, 16-bit resolution, Advanced Brain Monitoring, Carlsbad, CA) with linked-mastoid reference electrodes was fitted to the subject’s head.
For the preliminary BCI-FES experiments, a Bioness H200 neuroprosthetic wrist-hand orthosis (Bioness Inc, Valencia, CA) was used to deliver electrical stimulation to the muscles controlling movements of the hand. During the first visit, a neuroprosthetic wrist-hand orthosis (NESS H200, Bioness Inc, Valencia, CA) was fitted to the right hand of the subject. FES was delivered to the extensor (extensor digitorum communis and extensor pollicis brevis) and flexor (flexor pollicis longus and thenar) muscle groups alternately to produce opening and closing movements of the fingers and hand. Stimulation intensity was set by holding the pulse duration (300μs) and frequency (35Hz) constant, while slowly increasing the current amplitude.
BCI studies differ from basic neuroscience studies, which are typically observational. In basic science research, neural signals are recorded in a passive, open-loop way, with no feedback from the recording system. BCIs record activity from the nervous system and produce an output that is given to the user as feedback in real time. In this way, a BCI can affect the signals being recorded from the nervous system.
PRELIMINARY BCI-FES EXPERIMENTS
The experimental setup. The ABM X10 headset was used to record neural signals from subjects during these experiments. All commands and feedback were displayed on an Arduino UNO microcontroller board with a displayshield (1.8” 18-bit Color TFT Shield with microSD and Joystick, Adafruit.com), which was interfaced to the system through a serial port of the ESU. The Bioness wrist-hand orthosis was used to deliver stimulation to the extensor muscles responsible for opening the hand. No stimulation was delivered for hand closing, as the Bioness was not capable of delivering this type of stimulation in solation.
CUSTOM FES CONTROL PARADIGM
Aim 1 of this project was to investigate stimulation paradigms for electrical activation of hand muscles for a functional task. In order to accomplish this, two main objectives were established: 1) design a control paradigm for delivering stimulation, and 2) test efficacy and safety of stimulation protocol.
Control Paradigm for Delivering Stimulation:
In order to perform the grasp and release task, a custom stimulation configuration and control paradigm were developed for delivering stimulation to the muscles using a Digitimer stimulator. Custom stimulation protocols were programmed in the Arduino integrated development environment (IDE 1.0.5) software and then uploaded to an Arduino UNO microcontroller board. The Arduino code is provided in the Appendix material. Trigger inputs (20 μs pulse duration) were sent from the Arduino to the Digitimer, through custom cables (Cooner Wire),at a frequency of 35 Hz.
Safety and Efficacy of Stimulation Protocol:
The safety and effectiveness of the stimulation was assessed before use with human subjects. The Digitimer stimulator/PCB/Arduino configuration was bench tested in an electronics lab and output characteristics were assessed on an oscilloscope. A simulated body impedance circuit was designed to test the output parameters of the flexor and extensor circuits. A resistive-capacitive load made up of a 1kΩ resistor in parallel with a .047μF capacitor was used for voltage output measurements. A 1kΩ shunt resistor was used for current measurements.
FEATURES FOR BCI-FES CONTROL
Aim 2 of this project was to enable closed-loop brain control of FES to paralyzed muscles in order to perform a functional task. In order to accomplish this, two main objectives were established: 1) determine neural features that are optimal for triggering muscle stimulation sequences synchronously, and 2) develop an experimental paradigm for brain-controlled activation of paralyzed musclesto perform a functional task.
Since event-related desynchronization was not clear in earlier experiments, when only 1 second was given for subjects to imagine either opening or closing of the hand, it was determined that a longer period of time should be allowed. In order to verify the presence ofERDduring actual and imagined movements of the hand, control subjects were instructed to spend 10 seconds idling/resting, followed by 10 seconds moving their right hand (cycling slowly between opening and closing the hand) 10 seconds idling/resting, and then another 10 seconds of imagining moving the right hand.
ASSESSMENT OF FUNCTION
Aim 3 of this project was to assemblea battery of rehab metrics to quantify improvements in hand function over time. During preliminary closed-loop experiments, function was not assessed. Moving forward, grip strength will be assessed at the beginning of each testing session using a microFET4 grip and pinch force gauge dynamometer. Testing of the grip strength will be evaluated here in three uninjured, control subjects. Functional performance of the task will also be evaluated in the same group of subjects by the Grasp and Release Test (GRT).
BCI-FES FOR GRASP AND RELEASE
During each session, subjects performed 120 trials of BCI training with FES of the right hand, which was the dominant hand for all participants. Each subject completed a total of 720 BCI-FES trials over the 6 days of training. Each training session lasted 2-3 hours. Subjects were seated comfortably either in their wheelchair (SCI subjects) or a stationary chair (control subjects) in front of a computer monitor. A trial began with a fixation cross, which was displayed for 1 second to minimize eye movements.
In this study, we developed a BCI-FES system and demonstrated effective control of a grasp and release task in 5 subjects with chronic (>1 year post-injury), complete, cervical SCI and compared results with 5 uninjured, control subjects (see supplementary videos). Subjects with SCI were able to control the BCI-FES system just as well as uninjured subjects despite living with a complete SCI for at least 1 and up to 15 years post-injury. We also evaluated the functional aspects of the FES by testing grip strength as well as performance of a grasp and release task.
Source: University of Miami
Author: Katie Gant