Reproduce “CATD: Unified Representation Learning for EEG-to-fMRI Cross-Modal Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Weiheng Yao, Zhihan Lyu, Mufti Mahmud, Ning Zhong, Baiying Lei, Shuqiang Wang
The Condition-Aligned Temporal Diffusion (CATD) framework generates high-quality functional MRI BOLD signals directly from more accessible EEG data using a diffusion-driven approach. This method also enhances the temporal resolution of the synthesized BOLD signals, demonstrating improved diagnostic accuracy for neurological disorders and the ability to identify abnormal brain regions.
Key insights
- Diffusion models provide a stable and effective approach for generating high-dimensional, heterogeneous functional neuroimages, outperforming prior Autoencoder-based methods for cross-modal synthesis.
- Aligning diverse spatiotemporal characteristics of EEG and BOLD signals within a unified latent space is crucial for accurate and constrained generative synthesis.
- The high temporal fidelity of EEG can be exploited to overcome the inherent temporal resolution limitations of fMRI BOLD signals, enabling the capture of finer brain dynamics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.