🔍 Literature Spotlight: Diffusion Models & Brain Imaging

[Sarah0ravari]

Hi everyone,

As I continue preparing my GSoC 2025 proposal for the project "Advancing Brain Decoding and Cognitive Analysis", I came across a relevant paper that explores the use of latent diffusion models for reconstructing visual experiences from fMRI data.

📄 Paper Title: High-Resolution Image Reconstruction with Latent Diffusion Models from Human Brain Activity
🔗 Link: ResearchGate – Paper Link

📌 Key Takeaway:
This study introduces a method for reconstructing visual images from fMRI signals using a latent diffusion model (Stable Diffusion). It involves mapping brain activity to the latent space of the model, allowing high-resolution image reconstruction that reflects the subject's visual experience. The results highlight how diffusion models can capture rich, complex patterns in brain signals.

I believe this methodology could help shape the model architecture or training approach for our project—particularly in how we handle spatiotemporal patterns and condition on cognitive states.

I’d love to hear your thoughts on whether this direction aligns with the goals of the NISYSLAB project or if there are related papers/approaches you'd recommend exploring.

Best regards,
Sadaf (Sarah) Draper
GitHub: @Sarah0ravari