In the space of barely over ten years, Brain Computer Interfaces (BCIs) used to restore movement have developed from the stuff of science fiction to clinically relevant devices. However, most existing BCIs, while technically remarkable, require the user to be wired to stationary equipment, and allow only intermittent control of a computer cursor or disembodied robotic limb. They require that the algorithm linking brain activity to the restored movement be frequently recalibrated. We have developed a wireless BCI that will operate 24 hours a day, restoring voluntarily movement to monkeys despite paralysis of their hand, for a broad range of their normal motor behaviors, such as foraging, feeding, or playing with enrichment toys. By using ?autoencoding neural networks? we will be able to greatly extend the period over which the BCI will work without recalibration. We have developed a unique model of spinal cord injury (SCI) using a chronically implanted infusion pump that delivers a potent drug (tetrodotoxin) to cuffs placed around two key nerves in the arm. The drug causes a nerve block that produces the acute effects of spinal cord injury for indefinite periods of time, yet with full recovery within a day of stopping the drug. Prior to the nerve block, we will record wirelessly not only neural signals from the brain, but also electromyograms (EMGs) from a large number of muscles in the arm and hand. We will make these recordings not only during typical, constrained motor behaviors in the lab, but also during completely unconstrained behaviors while the monkey is in its home cage. We will use the data to develop algorithms (?decoders?) that transform the neural signals into predicted EMG signals. Following the onset of paralysis, our BCI will use these EMG predictions as control signals for Functional Electrical Stimulation (FES), causing contractions of the paralyzed muscles that the monkey can control voluntarily through the computer interface. We will study the gradually changing brain activity as the monkeys learn to use this FES BMI. In addition, we will attempt to augment the monkey's performance by developing ?adaptive? decoders that improve their performance in parallel with the monkey's own adaptation, as well as ?teacher? decoders that coach the monkeys, pushing them toward desired control strategies and away from counterproductive ones. This technology gives us the ability to study the brain's representation of movement across a range of motor behaviors that has never been possible before. During paralysis, it will allow us to study motor learning and adaption without the limitations imposed by the intermittent availability of current BCIs. Finally, it provides a platform close to that necessary for clinical translation, with which we will be able to study the limits of current decoders and to develop nonlinear and adaptive decoders designed to assist the monkey's own adaptive processes. While this application is focused on restoration of grasp, its general principles will extend to the control of reaching, lower limb function, and even prosthetic limbs. Ultimately, this work will develop the interface, decoder, and control technology that will be necessary to move BCIs from the lab to the clinic.
Brain computer interfaces (BCIs) offer remarkable opportunities to study how the brain learns and to restore function to paralyzed patients, but existing BCIs are usable only intermittently, in highly constrained lab settings. We have developed a novel BCI that will restore hand use to monkeys with nerve-block paralysis, 24 hours a day, in their home cage as well as in the lab. This BCI offers a novel tool to study motor learning, and is a critical bridge toward clinical translation of this technology to human patients.
|Pandarinath, Chethan; Ames, K Cora; Russo, Abigail A et al. (2018) Latent Factors and Dynamics in Motor Cortex and Their Application to Brain-Machine Interfaces. J Neurosci 38:9390-9401|
|Gallego, Juan A; Perich, Matthew G; Naufel, Stephanie N et al. (2018) Cortical population activity within a preserved neural manifold underlies multiple motor behaviors. Nat Commun 9:4233|
|Benjamin, Ari S; Fernandes, Hugo L; Tomlinson, Tucker et al. (2018) Modern Machine Learning as a Benchmark for Fitting Neural Responses. Front Comput Neurosci 12:56|
|Perich, Matthew G; Gallego, Juan A; Miller, Lee E (2018) A Neural Population Mechanism for Rapid Learning. Neuron 100:964-976.e7|
|Brill, N A; Naufel, S N; Polasek, K et al. (2018) Evaluation of high-density, multi-contact nerve cuffs for activation of grasp muscles in monkeys. J Neural Eng 15:036003|
|Perich, Matthew G; Miller, Lee E (2017) Altered tuning in primary motor cortex does not account for behavioral adaptation during force field learning. Exp Brain Res 235:2689-2704|
|Gallego, Juan A; Perich, Matthew G; Miller, Lee E et al. (2017) Neural Manifolds for the Control of Movement. Neuron 94:978-984|
|Brill, Natalie A; Tyler, Dustin J (2017) Quantification of human upper extremity nerves and fascicular anatomy. Muscle Nerve 56:463-471|
|Ethier, Christian; Acuna, Daniel; Solla, Sara A et al. (2016) Adaptive neuron-to-EMG decoder training for FES neuroprostheses. J Neural Eng 13:046009|
|Sachs, Nicholas A; Ruiz-Torres, Ricardo; Perreault, Eric J et al. (2016) Brain-state classification and a dual-state decoder dramatically improve the control of cursor movement through a brain-machine interface. J Neural Eng 13:016009|
Showing the most recent 10 out of 32 publications