New FM-fMRI Model Synthesizes Task fMRI from Rest Data

Researchers have developed FM-fMRI, a novel flow-matching model designed to synthesize task-based fMRI time series from resting-state fMRI data. This method conditions the synthesis on specific task events, enabling the generation of realistic neural dynamics. Evaluations on public and internal datasets show FM-fMRI outperforms existing methods like conditional diffusion, GANs, and VAEs in spectral and connectivity consistency. The model's utility is further demonstrated by its ability to augment limited clinical data, improving downstream diagnostic classification for autism. AI

IMPACT This model could enable more efficient and scalable fMRI studies, potentially accelerating research in neuroscience and clinical diagnostics by synthesizing task-specific data from readily available resting-state scans.

RANK_REASON This is a research paper detailing a new model for fMRI data synthesis. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• BioPoint
• FM-fMRI
• Human Connectome Project

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →