Synthetic Neural Decoding Accelerates Scalable Brain Computer Interface Testing

Fingers typing on a keyboard. A bladder that knows when to hold and when to let go. Though they seem like worlds apart, both depend on complex communication networks. Wallace H. Coulter Department of Biomedical Engineering Associate Professor Zachary Danziger studies those hidden signals that allow the nervous system to control the body and what happens when those signals break down.  

Danziger’s research focuses on developing new models to serve as surrogates for the human body. These models give researchers less-invasive ways to study how the body’s systems interact and communicate, while finding potential therapies that could help people regain lost functionality.  

One area of focus is brain-computer interfaces. Brain-computer interfaces help people with paralysis to send commands to devices like power wheelchairs, robotic arms, or other assistive technologies to turn thoughts into actions. The challenge is decoding those intentions from the brain’s electrical signals.  

That’s a difficult thing to do, from recording someone’s brain signals to figuring out what they want, because we don’t speak the language of the brain.”

Zachary Danziger, Associate Professor, Georgia Institute of Technology

Brain-computer interface research typically relies on neural signals recorded from electrodes implanted in the brain. But in Danziger’s lab, participants move their hands and fingers in a virtual environment, while artificial intelligence (AI) generates simulated neural signals based on their movements. With this approach, researchers can evaluate brain-computer interface algorithms across large groups of participants that would not be possible through clinical studies.  

“In the history of brain-computer interfaces ... there’s something like 40 to 70 total people who have ever had these systems implanted,” explains Danziger. “Our last study had 100 participants. In the in one study, we can get more [participants] than in entire history of this field.”  

The next step is to determine how well those algorithms perform in human studies. Through a collaboration with Coulter BME Associate Professor Chethan Pandarinath and Emory University’s BrainGate clinical trial team, Danziger’s lab is sharing the most promising findings for evaluation in future studies involving participants with implanted electrodes.  

Understanding how neural signals produce movement is only one part of Danziger’s research. His lab also investigates how the nervous system controls another essential bodily function: bladder control.

According to data from the National Institutes of Health, 20 million Americans are affected by underactive bladder, a condition that makes it difficult to fully empty the bladder. The condition can lead to other health problems like urinary tract infections. Finding out why the bladder is not emptying is a crucial step towards developing treatment.  

Danziger’s team is looking at the role of the urethra, and how its neurons impact bladder contraction. Their research suggests that may be possible to boost neural activity to reduce the symptoms.  

The questions Danziger studies are simple: How does the body know what to do, and what happens when those signals fail? Answering them requires a team of experts from AI, mathematics, computer science, physiology, and biomedical engineering. Bringing these disciplines together has become an important part of Danziger’s work. He often takes on a role as a translator, helping experts from different fields work towards understanding how the nervous system controls the body.  

It can be challenging, but as Danziger says “when it works, it’s great because it’s a much bigger project than anyone could do on their own.”