FlatClip: A Geometry-Aware Surface-Level Baseline for fMRI Representation Learning

2026-09-25

AI Summary

This study addresses the trade-off between spatial granularity and computational efficiency in fMRI modeling by repurposing image-pretrained encoders for cortical activity analysis. We propose FlatClip, a baseline that renders cortical activity into geometry-aware surface-level flatmap sequences, serving as an intermediate representation bridging region-of-interest (ROI) and voxel-wise models. A frozen SigLIP2 encoder is then employed for lightweight downstream probing. Experiments on HCP and ADNI benchmarks demonstrate that this approach outperforms ROI-based baselines. Furthermore, anatomically informed spatial arrangements significantly surpass random permutations, confirming the efficacy of reusing pretrained visual features. Overall, this work establishes a practical intermediate-state representation scheme for brain activity decoding.

Abstract

Recent fMRI foundation models differ substantially in the spatial scale at which they represent brain activity. ROI- and connectivity-based models are efficient but coarse, whereas voxel-level models preserve fine-grained spatial structure but require specialized 3D/4D architectures and costly fMRI-specific pretraining. We ask how effectively an image-pretrained encoder can reuse the spatial organization of cortical activity. Motivated by evidence that macroscale brain activity is strongly constrained by brain geometry, we introduce FlatClip, a frozen-encoder surface-level baseline that renders cortical activity as geometry-aware flatmap sequences and reuses a frozen SigLIP2 image encoder with only a lightweight downstream probe. Across resting-state benchmarks, FlatClip serves as a competitive middle-ground representation, outperforming ROI-level baselines on HCP and ADNI tasks while remaining weaker on PPMI and below the strongest voxel-level models overall. On visual-fMRI decoding, restricting the input to visual or NSD-provided task-active cortex improves performance, highlighting the value of task-relevant cortical coverage. Spatial perturbation controls reduce the predictive performance of flatmap features under both retrained and fixed readouts, and anatomy-linked arrangements consistently outperform vertex permutations across three colormaps. Together, these results position surface-level flatmap sequences as a practical middle-ground baseline between ROI and voxel models, and support the utility of anatomy-linked spatial organization for reusing image-pretrained features. Code is available at https://github.com/OneMore1/FlatClip.

Problem

• fMRI representation learning
• spatial scale
• cortical activity
• image-pretrained encoder
• surface-level baseline

Innovation

• fMRI representation learning
• surface-level flatmap
• frozen image encoder
• geometry-aware
• SigLIP2