Accessing Mesoscale Functional Organization from Standard Functional MRI Leveraging AI-Based Super-Resolution

f-LENNS, an AI-based super-resolution method for fMRI magnitude images, substantially improves spatial fidelity of functional activation maps and generalizes without retraining across diverse acquisitions.

• Why it matters: Fine-scale brain functional organization is needed to study cognition and neurological disorders, but conventional fMRI is limited by spatial resolution and contrast-to-noise ratio.
• What they did: The framework was trained only on high-resolution synthetic MRI using extensive domain randomization, with no adversarial or generative components, then tested across multiple imaging paradigms and an amblyopia cohort.
• The result: It substantially improves spatial fidelity versus interpolation-based upsampling and sharpening, curbs hallucination, and enables scalable fine-scale functional brain mapping despite acquisition domain shifts.

AI summary of the abstract · may contain errors · not medical advice · read the paper.