A Brain Graph Foundation Model: Pre-Training and Prompt-Tuning for Any Atlas and Disorder

2025-05-31

AI Summary

Existing brain graph models lack generalizability across diverse brain atlases and unseen neuropsychiatric disorders. Method: We propose BrainGFM, the first foundation model for multi-atlas, cross-disease brain graphs, trained via graph contrastive learning and graph masked autoencoding on 25,000+ fMRI scans from 27 datasets, covering 25 neuropsychiatric disorders and 8 brain atlases. BrainGFM introduces a novel graph-structured foundation model paradigm enabling zero- or few-shot adaptation to arbitrary atlases and previously unseen diseases. It integrates graph prompts with language prompts and employs meta-learning to optimize graph prompts for language-guided cross-disease generalization. Contribution/Results: BrainGFM achieves significant improvements in zero- and few-shot disease classification across all 25 disorders, attaining an average AUC of >0.82 on novel diseases. The code is publicly available.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Humans and AI: Brain-Sensing and Analysis
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

As large language models (LLMs) continue to revolutionize AI research, there is a growing interest in building large-scale brain foundation models to advance neuroscience. While most existing brain foundation models are pre-trained on time-series signals or region-of-interest (ROI) features, we propose a novel graph-based pre-training paradigm for constructing a brain graph foundation model. In this paper, we introduce the Brain Graph Foundation Model, termed BrainGFM, a unified framework that leverages graph contrastive learning and graph masked autoencoders for large-scale fMRI-based pre-training. BrainGFM is pre-trained on a diverse mixture of brain atlases with varying parcellations, significantly expanding the pre-training corpus and enhancing the model's ability to generalize across heterogeneous fMRI-derived brain representations. To support efficient and versatile downstream transfer, we integrate both graph prompts and language prompts into the model design, enabling BrainGFM to flexibly adapt to a wide range of atlases, neurological and psychiatric disorders, and task settings. Furthermore, we employ meta-learning to optimize the graph prompts, facilitating strong generalization to previously unseen disorders under both few-shot and zero-shot learning conditions via language-guided prompting. BrainGFM is pre-trained on 27 neuroimaging datasets spanning 25 common neurological and psychiatric disorders, encompassing 2 types of brain atlases (functional and anatomical) across 8 widely-used parcellations, and covering over 25,000 subjects, 60,000 fMRI scans, and a total of 400,000 graph samples aggregated across all atlases and parcellations. The code is available at: https://github.com/weixinxu666/BrainGFM

Problem

• Develops a graph-based brain foundation model for neuroscience
• Enables generalization across diverse brain atlases and disorders
• Integrates graph and language prompts for flexible downstream tasks

Innovation

• Graph-based pre-training for brain models
• Uses graph contrastive and masked autoencoders
• Integrates graph and language prompts