BDSB: Brain Disk Schrödinger Bridge for Enhancing 3T BOLD fMRI using Unpaired 7T Data for Visual Retinotopic Decoding

ICLR 2026 Conference Submission20620 Authors

Keywords: BOLD fMRI, Signal Enhancement, Schrödinger Bridge, Unsupervised Learning, Retinotopic Mapping, Visual Decoding

TL;DR: Enhances 3T fMRI using unpaired 7T data via a diffusion model in a shared brain disk space to improve visual decoding.

Abstract: Brain–computer interfaces increasingly rely on retinotopic mapping and visual decoding to reconstruct perceptual experiences from brain activity. High spatial and temporal resolution, coupled with a strong signal-to-noise ratio (SNR), has made 7-Tesla (7T) blood-oxygenation-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) an invaluable tool for understanding how the brain processes visual stimuli. However, the limited availability of 7T MRI systems means that most research relies on 3-Tesla (3T) scans, which offer lower spatial and temporal resolution and SNR. This naturally raises the question: Can we enhance the spatiotemporal resolution and SNR of 3T BOLD fMRI data to approximate 7T quality? In this study, we propose a novel framework that aligns 7T and 3T fMRI data from different subjects and datasets in a shared parametric domain. We then apply an unpaired Brain Disk Schrödinger Bridge (BDSB) diffusion model to enhance the spatiotemporal resolution and SNR of the 3T data. Our approach addresses the challenge of limited 7T data by improving the 3T scan quality. We demonstrate its effectiveness by testing it on three distinct public fMRI retinotopy datasets (one 7T, one 3T, and one paired 3T/7T), as well as synthetic data. The results show that our method significantly improves the SNR and goodness-of-fit of the population receptive field (pRF) retinotopic decoding in the enhanced 3T data, making it comparable to 7T quality. The codes will be available at Github.

Primary Area: applications to neuroscience & cognitive science

Submission Number: 20620