Scaling Vision Transformers for Functional MRI with Flat Maps
📰 ArXiv cs.AI
Learn how to scale Vision Transformers for functional MRI analysis using flat maps and self-supervised foundation models
Action Steps
- Convert 3D fMRI volumes to 2D flat maps using spatial transformations
- Train a Vision Transformer model using the masked autoencoder framework on large fMRI datasets
- Evaluate the performance of the model using the Brainmarks open evaluation suite
- Fine-tune the model for specific tasks such as functional connectivity analysis or brain decoding
- Apply the trained model to new fMRI datasets for analysis and visualization
Who Needs to Know This
Neuroimaging researchers and AI engineers can benefit from this technique to improve functional MRI analysis, and develop more accurate self-supervised foundation models
Key Insight
💡 Converting 3D fMRI volumes to 2D flat maps enables the application of Vision Transformers to functional MRI analysis
Share This
🧠💻 Scaling Vision Transformers for functional MRI analysis with flat maps and self-supervised learning! #fMRI #AI #Neuroimaging
Key Takeaways
Learn how to scale Vision Transformers for functional MRI analysis using flat maps and self-supervised foundation models
Full Article
Title: Scaling Vision Transformers for Functional MRI with Flat Maps
Abstract:
arXiv:2510.13768v2 Announce Type: replace-cross Abstract: We study the problem of training self-supervised foundation models for functional MRI. Our main contributions are: (1) we introduce a new model family (CortexMAE) trained using the masked autoencoder framework on 2.1K hours of open fMRI data, and (2) we release the first open evaluation suite (Brainmarks) for fMRI foundation models. Our core innovation is simple: we adapt the Vision Transformer to fMRI by first converting each 3D fMRI vol
Abstract:
arXiv:2510.13768v2 Announce Type: replace-cross Abstract: We study the problem of training self-supervised foundation models for functional MRI. Our main contributions are: (1) we introduce a new model family (CortexMAE) trained using the masked autoencoder framework on 2.1K hours of open fMRI data, and (2) we release the first open evaluation suite (Brainmarks) for fMRI foundation models. Our core innovation is simple: we adapt the Vision Transformer to fMRI by first converting each 3D fMRI vol
DeepCamp AI