Scaling Vision Transformers for Functional MRI with Flat Maps
Researchers adapted Vision Transformers (ViT) to functional MRI data using a 'flat map' projection to create CortexMAE. This method tests if this intermediate representation balances lossy and dense data well for brain analysis. The study found that the flat map representation is best for decoding cognitive states, suggesting it's a versatile way to use ViTs in neuroscience.