Neuroscientists have uncovered the process through which sensory information is converted into motor actions throughout different brain regions in mice. This study indicates that decision-making is a comprehensive activity involving the entire brain, closely linked with the learning process. These results may benefit the field of artificial intelligence by informing the design of more integrated neural networks.
Neuroscientists have uncovered the mechanism by which sensory information is converted into motor actions across several brain regions in mice. Research carried out at the Sainsbury Wellcome Centre at UCL demonstrates that decision-making functions as a holistic process across the brain, regulated by learning. These discoveries could provide valuable insights for artificial intelligence development, particularly in creating more interconnected neural networks.
Professor Tom Mrsic-Flogel, who leads the Sainsbury Wellcome Centre at UCL and is a key author of the study, noted, “This research brings together ideas previously associated with specific brain regions into a unified model that aligns with overall brain neural activity. We now have a clearer understanding of the processes occurring in the brain as sensory information is transformed through decision-making into action.”
The findings, published today in Nature, describe how the researchers utilized Neuropixels probes, an advanced technology that allows simultaneous recordings from hundreds of neurons across various brain areas, to investigate mice engaged in a decision-making task. This task was created by Dr. Ivana Orsolic at SWC and enabled the team to differentiate between sensory processing and motor response. The impact of learning was also highlighted by comparing trained animals to those with no prior training.
Dr. Michael Lohse, a Sir Henry Wellcome Postdoctoral Fellow at SWC and a lead author of the study, stated, “We frequently make decisions with unclear evidence. For instance, when it begins to rain, you must decide how frequent the raindrops should be before you open your umbrella. We examined this type of ambiguous evidence integration in mice to determine how the brain forms perceptual decisions.”
Mice were conditioned to remain still while observing a shifting visual pattern on a screen. To earn a reward, they needed to lick a spout upon detecting a marked increase in the pattern’s movement speed. The task was crafted to prevent a constant speed, as it fluctuated continually, and the timing of speed variations varied with each trial. Therefore, mice had to stay attentive to the stimulus and integrate the information to conclude if an increase in speed had occurred.
Dr. Andrei Khilkevich, a Senior Research Fellow in Mrsic-Flogel’s lab and another lead author of the study, explained, “By having the mice remain still, our data analysis was significantly clearer, allowing us to examine how neurons respond to random speed changes before the mice take action. We discovered that no single brain region is responsible for integrating sensory information or directing the decision process. Instead, neurons that are sparsely yet widely distributed across the brain connect sensory information with action initiation.”
The research team recorded data from each mouse multiple times, collecting information from over 15,000 cells across 52 brain regions in 15 trained mice. To gain insights into learning, they also compared these findings to data from naïve mice.
Dr. Lohse explained, “We found that when mice were unaware of what the visual cue represented, their brain only processed the information within the visual system and certain midbrain areas. After learning the task, the neurons began integrating evidence from across the brain.”
This study focused solely on naïve animals and those that had completely learned the task; however, future research aims to explore the learning process by monitoring neurons over time to observe changes as mice become adept at the task. The researchers also plan to investigate whether specific brain regions serve as crucial hubs for connecting sensory experiences and actions.
Further questions arising from this study include how the brain anticipates when the speed of the visual pattern will increase, allowing animals to react only to relevant information. The team intends to delve into these questions utilizing the dataset they’ve assembled.
This research was supported by Wellcome grants (217211/Z/19/Z and 224121/Z/21/Z), as well as the Sainsbury Wellcome Centre’s Core Grant from the Gatsby Charitable Foundation (GAT3755) and Wellcome (219627/Z/19/Z).
