Brain imaging technologies capture the neurons and brain circuitry with different modalities, including structural and functional management resonance imaging (MRIs). We aim to build a large-scale foundation model from healthy brain MRIs and then use it for various downstream applications, such as understanding the brain development process during early ages of life, decoding brain aging mechanisms, and identifying the pathology of different neurodegenerative or neuropsychiatric disorders. There are many public datasets; however, they often introduce different protocols and image acquisition procedures. These differences create distribution shifts, which call for developing harmonization methods in addition to training FMs. When trained, our Brain Foundation Model (BFM) can contribute to building a better understanding of the human brain while informing different clinical disease understanding applications.
