A new AI-assisted brain atlas that helps visualize the human brain in unprecedented detail has been developed by UCL researchers, marking a major step forward in neuroscience and neuroimaging.
The human brain is made up of hundreds of interconnected areas that drive our thoughts, emotions, and actions. Existing brain atlases can identify key structures such as the hippocampus, which support memory and learning, in MRI scans, but its more subtle subregions remain difficult to detect. These distinctions are important because subregions of regions such as the hippocampus, for example, are affected differently during the progression of Alzheimer’s disease.
Examination of the brain at the cellular level can be done using microscopy (histology), but it cannot be done on living individuals, which limits the possibility of understanding how the human brain changes during development, aging, and disease processes.
Published in naturea new study introduces NextBrain, an atlas of the entire adult brain that can be used to analyze MRI scans of living patients in minutes and at a level of detail previously impossible.
The authors of the freely available atlas hope it will ultimately help accelerate discoveries in brain science and apply them to better diagnosis and treatment of conditions such as Alzheimer’s disease.
How was the AI-assisted brain atlas developed?
The atlas took a team of researchers six years to build, a painstaking process similar to completing a jigsaw puzzle. However, it was not made using post-mortem tissue from five human brains.
Each brain was painstakingly dissected, divided into 10,000 pieces, stained to identify brain structures, photographed under a microscope, and reassembled into a 3D digital model. Before starting this process, the researchers conducted an MRI scan of the brain to see how they could put the brain back together, similar to the photo on the front of the jigsaw puzzle box.
AI was used to help align the microscopic images and MRI scans, taking into account the differences between the two techniques and ensuring there were no overlapping parts or gaps.
A total of 333 brain regions were then labeled on digital 3D models of each of the five brains. This process has been greatly accelerated by AI. Researchers say this would take decades if done manually.
NextBrain is the culmination of years of efforts to bridge the gap between microscopic imaging and MRI. By combining high-resolution tissue data with advanced AI techniques, we have created a tool that allows researchers to analyze brain scans at a level of detail previously unattainable. This opens new possibilities for the study of neurodegenerative diseases and aging. ”
Dr Juan Eugenio Iglesias, lead study author from UCL Medical Physics and Biomedical Engineering and Massachusetts General Hospital/Harvard Medical School
The resulting atlas is an “average” of the five brain models and can be generalized to all adults. This means it can be used to automatically infer details from MRI scans of living or dead subjects.
Brain Atlas accuracy tested over thousands of scans
NextBrain has been successfully tested on thousands of MRI datasets and has been demonstrated to reliably identify brain regions across a variety of imaging conditions and scanner types.
In one experiment, the team used the atlas to automatically label brain regions in publicly available ultra-high resolution MRI scans. This closely matched manually labeled regions, even for small regions such as subregions of the hippocampus.
In another experiment, the researchers applied NextBrain to more than 3,000 MRI scans of living individuals to examine age-related changes in brain volume. The atlas allows for a more detailed analysis of aging patterns than is possible using existing tools.
Dr Zane Jaunmukhtan, study author from the UCL Queen Square Neurological Institute and the Queen Square Neurological Disorders Brain Bank, said: “Our goal in building this atlas was to preserve the fine-grained anatomical accuracy of the microscopy data, while still allowing researchers to “NextBrain’s level of anatomical detail is astonishing, and its public availability means researchers around the world can immediately benefit from it.”
“NextBrain provides an unparalleled map of the cellular structure of the brain. The infrastructure built into the Atlas enables rapid, accurate and accessible analysis of brain images in living individuals, opening the door to detecting early signs of neurological diseases such as Alzheimer’s disease long before symptoms appear, improving our ability to understand, monitor and ultimately prevent these devastating diseases.”
All underlying data, tools, and annotations used in NextBrain, along with visualization tools and educational resources, are openly released through the FreeSurfer neuroimaging platform.
This research was supported by the European Research Council, the Alzheimer’s Association, the Lundbeck Foundation, and the National Institutes of Health (USA).
sauce:
university college london
Reference magazines:
Casamigiana, A. others. (2025). A probabilistic histological atlas of the human brain for MRI segmentation. nature. doi.org/10.1038/s41586-025-09708-2
