New algorithm disentangles intrinsic brain patterns from sensory inputs

Maryam Shanechi and her team developed a new machine learning method that reveals surprisingly consistent intrinsic brain patterns across different subjects by dissociating these patterns from the effect of visual inputs. The work was published in the Proceedings of the National Academy of Sciences.

When performing various everyday movement behaviors, such as searching for a book, our brains must absorb information, often in the form of visual input, for example to see where the book is. Our brain must then process this information internally to coordinate the activity of our muscles and carry out the movement.

But how do the millions of neurons in our brain accomplish such a task? Answering this question requires studying the collective activity patterns of neurons, but this while dissociating the effect of the input from the intrinsic (i.e. internal) processes of the neurons, whether they are or not related to movement.

That’s what Shanechi, his Ph.D., thinks. student Parsa Vahidi and a research associate in her lab, Omid Sani, did this by developing a new machine learning method that models neural activity while taking into account both movement behavior and inputs. sensory.

“Previous methods of analyzing brain data either considered neural activity and input but not behavior, or neural activity and behavior but not input,” Shanechi said.

“We developed a method that can take into account all three signals (neural activity, behavior and input) when extracting hidden brain patterns. This allowed us not only to disentangle intrinsic and input-related neural patterns, but also to separate intrinsic patterns that were related to movement behavior and those that were not.

Shanechi and his team used this method to study three publicly available datasets, in which three different subjects performed one of two separate motion tasks, involving moving a cursor on a computer screen across a grid or move it sequentially to random locations.

“Using methods that did not take into account all three signals, the patterns found in the neuronal activity of these three subjects appeared different,” Vahidi said. But when the team used the new method to consider all three signals, a remarkably consistent hidden pattern emerged from the three subjects’ neural activity, relevant to movement. This similarity occurred despite the fact that the tasks performed by the three subjects were also different.

“In addition to revealing this new consistent pattern, the method also improved the prediction of neural activity and behavior compared to when all three signals were not considered during machine learning, such as during previous work,” Sani said. “The new method allows researchers to more precisely model neural and behavioral data by taking into account various inputs measured in the brain, such as sensory inputs as in this work, electrical or optogenetic stimulation, or even inputs from different areas of the brain.”

This method and the discovered model can help us understand how our brain makes movements, guided by the information we receive from the outside world. Furthermore, by modeling the effect of inputs and separating intrinsic patterns relevant to behavior, this method can help develop future brain-computer interfaces that regulate abnormal brain patterns in disorders such as major depression by optimizing the external inputs such as deep brain stimulation therapy. .

“We are excited about how this algorithm could facilitate both scientific discoveries and the development of future neurotechnologies for millions of patients suffering from neurological or neuropsychiatric disorders,” Shanechi said.