The new AI, developed by an international team with contributions from UCL, can translate raw data on brain activity.

The new technique could speed up the search for connections between brain activity and behavior.

The new development was created together with the Kavli Institute for Systems Neurobiology in Trondheim and the Max Planck Institute for Cognitive and Brain Sciences in Leipzig. The authors have created a highly accurate neural network, a special type of deep learning algorithm that can decode many different types of behavior and stimuli from different areas of the brain.

Neuroscientists are recording more and more data from the brain, but understanding the information contained in this data, in other words, reading the neural code, is still a difficult problem.

Marcus Frey, lead author of the study

Frey notes that the team wanted to develop an automated method for analyzing various types of raw neural data without having to manually decode it.

They tested the AI ​​on neural signals from rats and found that their development was able to accurately predict the position, direction of the head and the speed of the animals’ running. Even without manual processing, the results were more accurate than what was obtained using conventional analysis. Also, the AI ​​was able to predict the movements of people’s hands.

Senior author and professor Caswell Barry noted that existing methods for analyzing brain signals miss a lot of potential information in neural recordings. The problem is that we can only decode the elements that we understand, he emphasizes.

However, the new AI can access much more neural code. It decodes neural data more accurately and, importantly, is not limited by existing knowledge.

The authors plan to evolve the design so that it can predict higher-level cognitive processes such as reasoning or problem solving.