An international research team has uncovered a new method that can analyze the dynamics of brain activity to assess how we control our everyday behaviors.
Researchers from Monash University, the University of Southern California, and New York University created a new “preferential subspace identification” (PSID) algorithm that can extract behaviorally-relevant dynamics from brain signals by learning relationships between brain signals and observed behaviors.
A central challenge in understanding how the brain works is finding a link between the dynamics of our brain activity and behaviors it controls, such as moving your arm and using fingers to grasp items.
Research co-author Dr. Yan Wong from Monash University’s Department of Electrical and Computer Systems Engineering, and the Monash Biomedicine Discovery Institute, said, in the future, this algorithm will unlock the potential to decode brain signals and allow patients to control therapeutic devices with their minds.
The study, led by Assistant Professor Maryam Shanechi at the University of Southern California, was published in Nature Neuroscience. Dr Wong said:
“Modelling neural dynamics is essential to investigate or decode behaviourally-measurable brain functions, such as movement planning, initiation and execution; speech and language; mood; decision-making; as well as neurological dysfunctions like movement tremors. We can also use this to measure internal states of the brain, such as thirst or hunger.
“In the future, we can use this algorithm to help improve the performance of brain machine interfaces for quadriplegics, as an example.
“If we can better understand how the brain represents complex behaviours like reaching then we can build better algorithms to extract information from the brain.”
The research team applied PSID to neural activity in two subjects performing 3D reach-and-grasps, and uncovered new features for neural dynamics. Findings showed that PSID revealed behaviourally relevant dynamics to be significantly lower-dimensional than otherwise implied.
The algorithm also discovered distinct rotational dynamics that were more predictive of behavior, and it more accurately learned the behaviorally relevant dynamics for each joint and recording channel.
According to Dr. Wong, these findings showed promise for the monitoring and assessment of brain patterns in people with quadriplegia and neurological diseases, and could foreshadow clinical advances in neuroscience. Dr. Wong said:
“As opposed to neural dynamic modelling (NDM), our method combines neural activity and behaviour to identify a subspace between the two and project, through learning, the extent of the likely behavioural outcome.
“Compared with NDM, we found PSID more accurately learned behaviourally relevant neural dynamics for almost 27 arm and finger joint angles, for 3D end-point kinematics, and for almost all individual channels across the multi-regional recordings.”
Provided by: Monash University [Note: Materials may be edited for content and length.]