Large-scale Artificial Intelligence for Brain Mapping
Large-scale Artificial Intelligence for Brain Mapping
The team Large-Scale AI for Brain Mapping forms the Young Investigator Group of the Helmholtz AI local unit in Jülich and develops artificial intelligence (AI) methods for large-scale microstructural mapping of the human brain. Embedded in the Big Data Analytics group of the Institute for Neuroscience and Medicine (INM-1), we contribute to the development of a multimodal digital atlas of the human brain at cellular resolution. The focus of our team is on representation learning methods that help to uncover microstructural principles of human brain organization. We specifically aim to develop a large foundation model from petabyte-scale microscopic imaging datasets acquired at INM-1 that capture the spatial distributions of neuronal cells (cytoarchitecture), nerve fiber bundles (fiberarchitecture), neurotransmitter receptors (chemoarchitecture) and their genetic foundation (genomics). The resulting model will provide a holistic data-driven characterization of microstructural brain architecture at the whole-brain level, and enable downstream tasks such classification of brain structures, characterization of structural principles, generative modeling, and cross-modality inference. Driven by continuously increasing image resolutions and data volumes, as well as increasingly more powerful and compute-intensive AI approaches, we approach large-scale AI using the High-Performance Computing (HPC) systems provided by the Jülich Supercomputing Center (JSC), including the European flagship exascale supercomputer JUPITER.
We work with an interdisciplinary network of national and international partners. Our methods are used for the further development of the Julich Brain Atlas and thus contribute to the EBRAINS research infrastructure. Within the Helmholtz International BigBrain Analytics & Learning Laboratory (HIBALL), we collaborate closely with researchers in the fields of neuroscience and machine learning in Montréal, Canada on the development of the BigBrain high-resolution human brain model. In the X-BRAIN project of the Helmholtz Imaging Platform (HIP), we work together with colleagues at HMGU Munich and LMU Munich on AI methods for deepening our understanding of the relationship between cytoarchitecture and the genetic organization in the brain. We contribute to teaching activities of the department „Big Data Analytics for Microscopic Images“ for the programs Computer Science and Artificial Intelligence for Data Science at Heinrich-Heine University (HHU) Düsseldorf.