BrainFound: a generalisable foundation model for brain MRI analysis

A new foundation model, BrainFound, has been developed for brain MRI analysis, outperforming existing methods. BrainFound uses a slice-based learning strategy to process MRI volumes, enabling efficient learning and capturing cross-slice context. It supports multiple MRI modalities and shows strong generalization across diverse tasks. The model's development was supported by various grants and datasets, including ADNI and NACC, highlighting its potential for clinical and research applications.