Our research has several main thrusts. These focus on the use of devices that interface between computers and muscles or directly between computers and the brain. We use these powerful interfaces as paradigms to try to improve our understanding of brain function. In addition, our ultimate goal is to use them to improve our treatment of neurologic disorders.
Brain-Machine Interfaces to restore movement
The ultimate goal of this project is to restore hand and arm function to people with paralysis from neurological disorders. Brain machine interfaces (BMIs), which decode electrical signals recorded from the brain, offer one way to bypass the area causing paralysis and intuitively reanimate paralyzed muscles via electrical stimulation. Invasively-obtained signals recorded with electrodes penetrating the brain provide highly detailed information for BMIs. Most of these BMIs have used action potentials, or spikes, from individual neurons. However, with current technologies, spikes can be difficult to record for many years. We have shown that local field potentials (LFPs), which are summed signals from many thousands of neurons, can provide nearly as much information about reaching movements as spikes, even when spikes are not able to be recorded (Flint et al., J. Neural Eng. 2012). We showed for the first time that these LFPs can be used in a BMI to control a cursor with accuracy nearly as good as that of spikes. Further, this performance was stable over a year without having to recalibrate the BMI (Flint et al., J. Neural Eng. 2013). This finding is important because it can enable BMI users to learn a BMI over longer time periods without recalibration. Further, the signals themselves were highly stable over time, particularly in the task-relevant space (Flint et al., J Neurosci 2016).
A surprising amount of information about movement can also be obtained using less-invasive, subdural or epidural signals (recorded below and above, respectively, the tissue covering of the brain; Flint et al., J Neural Eng. 2017).
We are currently investigating the ability to use BMIs controlling haptic devices to drive neural recovery after traumatic brain injury. If successful, this application could greatly broaden the potential population of patients who could benefit from BMIs.
Myoelectric-Computer Interfaces for neurorehabilitation
Stroke remains the leading cause of chronic disability in the U.S., and more than half of stroke survivors have persistent impairment of arm function despite receiving conventional therapy. In these stroke survivors, a significant cause of impaired arm movement is abnormal co-activation between muscles that normally do not activate together. The goal of this research is to develop a new therapy for stroke using a wearable device to improve motor function by decoupling abnormally co-activating muscles. This therapy, a myoelectric computer interface (MCI), maps electrical muscle activity onto movements of a cursor in a computer game. This provides direct, detailed feedback about the co-activation of a pair of muscles to the user. Our early results show that training with the MCI allows stroke survivors to greatly reduce abnormal co-activation in the targeted arm muscle pair and may also improve function (Wright et al., NNR 2014). We are now assessing to what extent this home-based technology can improve function, and the effect of timing of the therapy on functional improvement.
Brain-Machine Interfaces to restore communication
People who become completely paralyzed, or “locked-in,” from diseases such as stroke, ALS or cerebral palsy are unable to speak or communicate in any manner. For these patients, the ability to restore communication is paramount.
BMIs have been able to restore basic communication by allowing patients to slowly spell out individual letters by decoding particular signals, most commonly evoked potentials in response to some stimulus. This paradigm requires intense concentration and a great deal of time (several minutes) to communicate a simple sentence. Thus, any communication restored with this paradigm is inherently limited in information rate. The time delays alone are a critical impediment – consider the frustration engendered by even just a few seconds of delay in conversations.
A far more user-friendly, and far higher throughput, paradigm would be to decode the patient’s intended speech directly from the speech motor cortical signals. That is, the patient would attempt to speak and the BMI would directly decode entire words and sentences, with minimal delays. Words in a language are composed of groups of sounds, or phonemes. Signals recorded from the surface of the motor cortex contain information about phonemes based on the location of phoneme production (e.g., tongue vs. lips). We have shown that we can decode the entire range of American English phonemes from motor cortex using subdural field potentials (Mugler et al., J Neural Eng. 2014). We are now investigating in greater detail how the brain controls speech production. We anticipate that this will ultimately enable us to restore intuitive communication to locked-in patients.