A continually expandable foundation model for brain MRI
Organizations: 1*Cancer Research UK Cambridge Institute, University of Cambridge Li Ka Shing Centre, Robinson Way, Cambridge, CB2 0RE, Cambridgeshire, United Kingdom. · Department of Computer Science and Technology, University of Cambridge, 15 JJ Thomson Ave, Cambridge, CB3 0FD, Cambridgeshire, United Kingdom. · Department of Psychiatry, University of Cambridge, Hills Road, Cambridge, CB2 2QQ, Cambridgeshire, United Kingdom. · Department of Psychology, University of Cambridge, Downing Pl, Aug Cambridge, CB2 3EB, Cambridgeshire, United Kingdom. · Department of Communication Engineering E.T.S. Ingenier´ıa de Telecomunicaci´on, University of M´alaga, Blvd. Louis Pasteur 35, Malaga, 29010, Spain. · Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Wilberforce Rd, Cambridge, CB3 0WA, Cambridgeshire, United Kingdom. · Department of Oncology, University of Cambridge, Cambridge Biomedical Campus, Cambridge, CB2 0SP, Cambridgeshire, United Kingdom. · Department of Clinical Neuroscience, University of Cambridge, Cambridge Biomedical Campus, Cambridge, CB2 0SP, Cambridgeshire, United Kingdom. · School of Data Science, University of Virginia, Charlottesville, VA, United States of America . · DaSCI Andalusian Institute of Data Science and Computational Intelligence,University of Granada, Av. del Conocimiento 37, Granada, 18016, Granada, Spain.
Abstract
Brain magnetic resonance imaging (MRI) is central to neuroscience and clinical assessment, but models are commonly developed for individual diseases, populations or imaging protocols. Foundation models promise more general representations, yet they are usually pretrained once and can lose earlier capabilities when updated with new data. Here we show that Alcmaeon, a three-dimensional brain MRI foundation model pretrained without manual labels on more than 425,000 volumes and derived imaging maps, can be expanded sequentially across clinical domains. Alcmaeon combines volumetric encoding and latent diffusion generation with Graph-Blueprint Pruning (GBP), which protects network modules important to earlier domains while leaving the remaining capacity trainable. Across expansion from healthy ageing and neurodegeneration to developmental, psychiatric and tumour imaging, GBP showed less forgetting than sequential adaptation and elastic weight consolidation across voxel-level reconstruction measures, with its largest advantage after adaptation to tumour imaging. The blueprints provided an inspectable record of how model capacity was protected and reused. Representations from different model levels supported image synthesis, disease classification, survival modelling and postoperative prediction, although no single representation was optimal for every task. These findings provide a route towards brain MRI foundation models that can grow with emerging data while retaining earlier capabilities.