Neuroscientists have uncovered the process by which sensory information is converted into motor actions in various regions of the mouse brain. Their study indicates that decision-making is a comprehensive process influenced by learning throughout the brain. These insights could potentially enhance artificial intelligence efforts by informing the development of more decentralized neural networks.
Recent research conducted at UCL’s Sainsbury Wellcome Centre has shed light on how sensory input transforms into motor actions across different brain regions in mice. The findings demonstrate that decision-making operates as a widespread brain process, guided by learning, and may help advance artificial intelligence by offering valuable information for designing more distributed neural networks.
“This research brings together various concepts previously identified in individual brain areas into a unified perspective that reflects brain-wide neural activity. We now have a holistic understanding of how sensory input is processed through decision-making to result in actions,” stated Professor Tom Mrsic-Flogel, Director of the Sainsbury Wellcome Centre at UCL and lead author of the study.
Published in Nature, the study describes how researchers utilized Neuropixels probes—cutting-edge technology that allows for simultaneous recordings from numerous neurons across different brain areas—while studying mice engaged in a decision-making task designed by Dr. Ivana Orsolic at SWC. This task enabled the team to separate sensory processing from motor control. The researchers were also able to assess the role of learning by comparing trained mice with those that had not undergone training.
“We often face decisions based on unclear evidence. For instance, when it begins to rain, you must determine how often raindrops need to fall before you decide to open your umbrella. We examined this type of ambiguous decision-making in mice to gain insights into how the brain handles perceptual decisions,” said Dr. Michael Lohse, Sir Henry Wellcome Postdoctoral Fellow at SWC and co-first author of the study.
In the experiment, mice were trained to remain still while observing a moving visual pattern on a screen. They had to lick a spout to receive a reward when they noticed a consistent increase in the speed of the moving pattern. The task was designed with speed fluctuations, so the timing of speed increases varied with each trial, preventing the mice from relying on memory alone. Consequently, the mice needed to stay focused on the stimulus and process information to determine if the speed increase had occurred.
“By having the mice stand still, we were able to conduct a more straightforward data analysis and investigate how neurons respond to random speed changes before mice take action. In the trained mice, we discovered that there is no single brain region responsible for integrating sensory information or coordinating the decision-making process. Instead, we identified neurons that are distributed sparsely yet broadly throughout the brain, linking sensory evidence to action initiation,” explained Dr. Andrei Khilkevich, Senior Research Fellow in the Mrsic-Flogel lab and co-first author of the paper.
The researchers recorded data multiple times from each mouse, gathering information from over 15,000 cells across 52 brain regions in 15 trained mice. To study learning, they also compared these results with recordings from naïve mice.
“We found that when mice lack understanding of the visual stimulus, the information is only represented within the visual system and a few midbrain regions. Once they learn the task, cells throughout the brain begin to integrate this evidence,” Dr. Lohse explained.
While this study exclusively focused on naïve mice and those who had completely learned the task, the researchers aim to explore the learning process further by monitoring neurons over time to observe how they adapt as mice come to understand the task. They also intend to investigate whether specific brain regions serve as central hubs that establish connections between sensations and actions.
The research has raised several additional questions, such as how the brain anticipates when the speed of the visual pattern will increase, prompting animals to respond only to relevant information. The team plans to further explore these topics using the dataset collected.
This study received funding from Wellcome awards (217211/Z/19/Z and 224121/Z/21/Z) as well as from the Sainsbury Wellcome Centre’s Core Grant provided by the Gatsby Charitable Foundation (GAT3755) and Wellcome (219627/Z/19/Z).
