Using AI to understand brain function
The European Research Council (ERC) has awarded a Consolidator Grant to the computer scientist and neuroscientist Professor Fabian Sinz at the University of Göttingen. The project “Vision2Action: A data-driven computational framework to discover how behavior impacts processing in mouse visual cortex” has been awarded around two million Euro for five years. In the project, Sinz and his team plan to develop new machine learning methods to understand how motor movement influences visual processing in the brain.
Prof. Dr. Fabian Sinz, Photo: Emmanouil Froudarakis
Bernstein members involved: Fabian Sinz
“Our movements influence how we perceive the world,” explains Sinz. “However, the interaction between movement and visual perception in natural environments is so complex that we need machine learning approaches to capture and understand these computational processes in the brain.” This effect is particularly pronounced in the visual cortex of rodents, where movement influences even very early processing stages of the visual pathway. However, the underlying organizational principles in the brain are currently not well understood.
In order to approach this question, the research team will use a computer-assisted approach. They will use deep learning models to create a “digital twin” of the visual system of freely moving mice. “This digital twin enables us to use computer simulations to analyse the complex dynamics between movement, brain states, and visual processing,” explains Sinz. “The simulations make it possible for us to explore different hypotheses that would be difficult to test otherwise.” The predictions obtained from the AI models can then be tested in subsequent experiments.
The research combines state-of-the-art machine learning methods with experimental neuroscience and is carried out in close collaboration with the Universities of Bonn in Germany and Stanford in the US. Sinz has been Professor of Machine Learning at the Institute of Computer Science and a board member of the Campus Institute Data Science (CIDAS) at the University of Göttingen since 2021. He previously worked as a group leader at the University of Tübingen and as an assistant professor at Baylor College of Medicine in Houston (USA), where he already successfully developed deep learning models for analysing brain activity.
The project promises new insights into the organization of the visual system, as well as the development of innovative AI methods for neuroscience. These computational approaches could also be ground-breaking for other areas of brain research in the future, as they provide new ways of using machine learning to better understand complex biological systems in natural environments.
