Brain-OF：一种用于fMRI、EEG和MEG的全功能基础模型

The study introduces Brain-OF, the first omnifunctional brain foundation model jointly pretrained on fMRI, EEG, and MEG, which is capable of handling both unimodal and multimodal inputs within a unified framework. Existing brain foundation models have made significant strides in various neuroscience tasks, but they are often limited to a single functional modality, hindering their ability to leverage complementary spatiotemporal dynamics and the collective data scale across different imaging techniques. To address this limitation, Brain-OF incorporates the Any-Resolution Neural Signal Sampler, which projects diverse brain signals into a shared semantic space to reconcile heterogeneous spatiotemporal resolutions. Furthermore, innovative methods are employed to manage semantic shifts, enhancing the integration of multimodal data.