Brain-machine interfaces - a moonshot project
Tom Spendlove posted on January 30, 2015 |
USC scientists develop methods to control movement based on brain activity.

Maryam Shanechi and her team at USC work on control systems, neuroscience and signal processing to develop brain machine interfaces. These interfaces are being used to help patients with neuropsychiatric disorders, aid in anesthesia administration, and contribute to the prosthetic limb development.

Brain machine interfaces have traditionally focused on developing motor function for patients with spinal cord injuries. The interfaces take neural activity, translate it into an algorithm, and then control an external device that gives visual feedback to the patient.

The major issues facing the brain machine interface field have been the same for fifteen years. The systems have low performance standards, the control systems are not robust and the applications are required to be very task specific.

Shanechi’s hypothesis was that building a model of the brain using control theory could help to understand that brain’s intention and lead to better control. The brain mainly uses visual feedback to close the feedback loop and the team built a math model to find the cost required for each brain activity.

Changing the decoders to use a millisecond timescale helped the team to achieve a faster response time. Logging each millisecond of brain activity and recording a 1 where brain spikes occurred gave an almost immediate communication link.

 Combining these decoder changes with spinal cord stimulation led Maryam to attempt brain-to-muscle control. In studies one subject was used as the decoder and one was sedated. Subjects were able to accurately control arm movement in two dimensions.

The technology demonstrated in this video is staggering. Using a monkey’s brain waves to control a different monkey definitely feels like a science fiction movie plot. Transferring these discoveries to help patients with neural damage or spinal cord injuries will be an incredible achievement.

(This Solve for X talk was posted in November 2014 and gives a longer discussion of the signal processing aspects of Shanechi’s projects. A three minute overview given at the MIT Technology Review in September 2014 can be found here.)

Recommended For You