Brain-IT: Image Reconstruction from fMRI via Brain-Interaction Transformer

Researchers developed Brain-IT, a new AI system using Brain Interaction Transformer technology to reconstruct images from fMRI brain recordings with significantly improved accuracy. The method requires only 1 hour of data versus 40 hours needed by current approaches while surpassing state-of-the-art results.

• →Brain-IT uses functionally-similar brain-voxel clusters shared across subjects as building blocks for brain-computer interface processing.
• →The system predicts both high-level semantic features and low-level structural features to guide diffusion model image reconstruction.
• →Method achieves superior visual and objective metric results compared to current state-of-the-art brain imaging approaches.
• →Training efficiency improved dramatically, requiring only 1 hour of fMRI data versus 40 hours for comparable results.
• →Shared model components across clusters and subjects enable effective training with limited datasets.