Toward a Multi-View Brain Network Foundation Model: Cross-View Consistency Learning Across Arbitrary Atlases

Jiaxing Xu; Jingying Ma; Xin Lin; Yuxiao Liu; Kai He; Qika Lin; Yiping Ke; Yang Li

cs.CVMar 20, 2026

Toward a Multi-View Brain Network Foundation Model: Cross-View Consistency Learning Across Arbitrary Atlases

Authors: Jiaxing Xu, Jingying Ma, Xin Lin, Yuxiao Liu, Kai He, Qika Lin, Yiping Ke, Yang Li,

Organizations: Saw Swee Hock School of Public Health at National University of Singapore, Singapore · School of Biomedical Engineering & State Key Laboratory of Advanced Medical Materials and Devices, ShanghaiTech University, Shanghai 201210, China · College of Computing and Data Science, Nanyang Technological University, Singapore · School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, China ·

Abstract

Brain network analysis provides an interpretable framework for characterizing brain organization and has been widely used for neurological disorder identification. Recent advances in self-supervised learning have motivated the development of brain network foundation models. However, existing approaches are often limited by atlas dependency, insufficient exploitation of multiple network views, and weak incorporation of anatomical priors. In this work, we propose MV-BrainFM, a multi-view brain network foundation model designed to learn generalizable and scalable representations from brain networks constructed with arbitrary atlases. MV-BrainFM explicitly incorporates anatomical distance information into Transformer-based modeling to guide inter-regional interactions, and introduces an unsupervised cross-view consistency learning strategy to align representations from multiple atlases of the same subject in a shared latent space. By jointly enforcing within-view robustness and cross-view alignment during pretraining, the model effectively captures complementary information across heterogeneous network views while remaining atlas-aware. In addition, MV-BrainFM adopts a unified multi-view pretraining paradigm that enables simultaneous learning from multiple datasets and atlases, significantly improving computational efficiency compared to conventional sequential training strategies. The proposed framework also demonstrates strong scalability, consistently benefiting from increasing data diversity while maintaining stable performance across unseen atlas configurations. Extensive experiments on more than 20K subjects from 17 fMRI datasets show that MV-BrainFM consistently outperforms 14 existing brain network foundation models and task-specific baselines under both single-atlas and multi-atlas settings.

Figures & tables

Fig. 1: Comparison of representative brain foundation models.

Fig. 2: The architecture of our proposed MV-BrainFM using Schaefer100 and AAL116 as example. (a) Multiple brain network datasets with various atlases are utilized for pretraining. (b) Datasets with arbitrary atlases can be introduced to finetune the model for downstream tasks with a simple prediction head. (c) The proposed Distance-Biased Transformer (DBT) Encoder.

	Datasets	Sex (F/M)	Age (mean $\pm$ std)	Class#	Subjects#	Atlases
Pretrain	HBN [ 2 ]	NA	NA	-	12427	Schaefer200, AAL116, Craddock200
	ADHD [ 12 ]	383/230	12.1 $\pm$ 3.3	-	613	Schaefer100, Schaefer200, Schaefer500
	ADNI [ 16 ]	610/552	71.8 $\pm$ 6.9	-	1162	Schaefer100, Schaefer200, Schaefer500, AAL116
	ABIDE [ 14 ]	122/697	16.9 $\pm$ 7.9	-	819	Schaefer100, Schaefer200, Schaefer500, AAL116
	PPMI [ 43 ]	82/127	62.9 $\pm$ 9.5	-	209	Schaefer100, Schaefer200, Schaefer500, AAL116
	Taowu [ 43 ]	17/23	65.0 $\pm$ 5.0	-	40	Schaefer100, Schaefer200, Schaefer500, AAL116

TABLE I: Statistics and class information of brain network datasets used in this work. Renji, SMHC, DAMOMRI, and Huashan are private datasets, while the rest are publicly available.

Method	Pretrained	Internal Tests				External Tests
Method	Pretrained	ADHD	ADNI	ABIDE	Avg.	HCP-Gender	HCP-Age	Huashan	Avg.
GCN [ 32 ]	✗	54.66 $\pm$ 6.38	66.06 $\pm$ 2.20	62.58 $\pm$ 4.95	61.1	74.13 $\pm$ 4.99	54.43 $\pm$ 8.25	54.73 $\pm$ 3.20	61.1
BrainNetCNN [ 30 ]	✗	55.10 $\pm$ 5.32	68.91 $\pm$ 2.06	67.18 $\pm$ 7.67	63.73	73.35 $\pm$ 14.08	54.81 $\pm$ 8.24	59.20 $\pm$ 4.73	62.46
BrainGNN [ 35 ]	✗	55.86 $\pm$ 5.74	63.23 $\pm$ 3.10	60.41 $\pm$ 4.02	59.83	72.17 $\pm$ 10.30	52.98 $\pm$ 5.41	55.82 $\pm$ 4.85	60.32
BNT [ 29 ]	✗	53.08 $\pm$ 7.21	61.04 $\pm$ 6.78	64.46 $\pm$ 7.22	59.53	76.57 $\pm$ 11.21	51.26 $\pm$ 5.98	59.91 $\pm$ 3.68	62.58
ContrastPool [ 63 ]	✗	59.65 $\pm$ 5.19	64.65 $\pm$ 3.66	68.69 $\pm$ 2.72	64.33	80.00 $\pm$ 5.80	56.30 $\pm$ 5.58	58.35 $\pm$ 6.11	64.88

TABLE II: Results (average AUC

\pm

standard deviation) on 6 datasets by averaging all available atlases. The best result is highlighted in bold while the runner-up is highlighted in underlined .

Panel of Figure 5 — Fig. 3: Results on all 4 atlases. Note that BASC122 is the external atlas, which the models have never seen during pretraining.

Fig. 4: Scaling behavior with increasing pretraining atlases. We progressively increase the number of atlases used during pretraining and evaluate the downstream performance on 6 tasks using the Schaefer100 atlas. The results show that MV-BrainFM consistently benefits from larger and more diverse pretraining data, demonstrating clearer scaling behavior than BrainGFM.

Method	Internal Tests		External Tests
Method	ABIDE	ADNI	HCP-Gender	HCP-Age
MGRL [ 13 ]	66.99 $\pm$ 4.86	59.57 $\pm$ 1.75	76.01 $\pm$ 8.40	56.78 $\pm$ 6.17
METAFormer [ 41 ]	75.87 $\pm$ 3.39	66.40 $\pm$ 1.24	84.17 $\pm$ 4.39	61.40 $\pm$ 4.69
AIDFusion [ 66 ]	75.19 $\pm$ 1.38	66.59 $\pm$ 2.43	85.34 $\pm$ 5.78	59.77 $\pm$ 5.35
MV-BrainFM	77.93 $\pm$ 2.36	72.63 $\pm$ 0.90	86.64 $\pm$ 4.54	61.47 $\pm$ 5.47

TABLE III: Multi-atlas AUC performance comparison using four atlases (Schaefer100, Schaefer200, AAL116, and BASC122). MV-BrainFM achieves superior performance over existing multi-atlas baselines on all datasets.

Fig. 5: AUC performance of MV-BrainFM using the Schaefer100 atlas under different hidden dimensions (0.95M–31.35M parameters). Increasing model capacity improves performance from very small models, but larger models do not consistently yield additional gains, indicating that data scale may be a more critical factor than parameter scale.

Fig. 6: Visualization of the top 20 edges with the largest distance bias for each attention head. Different heads consistently focus on connections at distinct spatial scales.

Panel of Figure 10 — Fig. 7: Visualization of salient brain regions identified by MV-BrainFM across different neurological disorders. The highlighted regions correspond to ROIs with the highest attention weights and represent potential disease-related biomarkers for ADHD, ASD (ABIDE), AD/MCI (ADNI and Huashan).

Models	#Params	Pretrain		Finetune Time (s)
Models	#Params	#Epoch	Time (s)	ADHD	ABIDE	ADNI	HCP-Gender	HCP-Age	Huashan
BrainLM	12.78M	50	3,989	430	609	785	632	605	462
Brain-JEPA	21.69M	50	4,124	1,038	1,573	1,814	1,630	1,275	1,126
PTGB	0.55M	70	1,574	23	15	22	26	30	29
BrainMass	29.35M	100	6,988	74	80	102	102	117	84
BrainGFM	10.31M	20	3,709	46	58	83	63	63	47

TABLE IV: Time cost (seconds) and number of parameters for brain network foundation models on datasets with Schaefer100 atlas. The finetune time reports the 50 epochs running time for a single fold including training, validation, and test.

Atlas Encoder	Distance Bias	$\mathcal{L}_{\text{rec}}$	$\mathcal{L}_{\text{cc}}$	$\mathcal{L}_{\text{ent}}$	ADHD	ABIDE	ADNI	HCP-Gender	HCP-Age	Huashan
✓	✓				54.70 $\pm$ 7.76	69.72 $\pm$ 1.06	64.23 $\pm$ 1.57	83.52 $\pm$ 1.79	58.70 $\pm$ 6.01	60.06 $\pm$ 4.40
		✓	✓	✓	57.26 $\pm$ 6.47	73.38 $\pm$ 1.68	67.74 $\pm$ 1.34	85.54 $\pm$ 2.52	57.99 $\pm$ 5.88	60.71 $\pm$ 6.23
	✓	✓	✓	✓	58.10 $\pm$ 4.97	74.38 $\pm$ 2.49	68.20 $\pm$ 2.10	85.36 $\pm$ 2.67	58.72 $\pm$ 6.22	61.61 $\pm$ 4.48
✓		✓	✓	✓	58.99 $\pm$ 6.45	75.27 $\pm$ 2.49	69.24 $\pm$ 1.63	86.77 $\pm$ 1.77	59.67 $\pm$ 5.42	61.31 $\pm$ 5.08
✓	✓		✓	✓	61.18 $\pm$ 3.54	72.79 $\pm$ 1.48	66.19 $\pm$ 2.79	84.41 $\pm$ 2.05	59.11 $\pm$ 4.21	61.64 $\pm$ 6.43
✓	✓	✓		✓	59.76 $\pm$ 1.85	73.22 $\pm$ 1.96	69.30 $\pm$ 0.89	86.38 $\pm$ 2.90	59.01 $\pm$ 5.15	61.83 $\pm$ 5.22

TABLE V: Ablation study on components in MV-BrainFM on datasets with the Schaefer100 atlas.

Panel of Figure 13 — Fig. 8: Hyperparameter sensitivity analysis of MV-BrainFM with respect to the number of attention heads and prototypes across multiple datasets.

fMRI Foundation Models

Functional magnetic resonance imaging (fMRI) foundation models learn reusable brain representations for prediction, transfer, and neuroscience discovery.

See all 15 papers

Learning Cross-Atlas Consistent Brain Disorder Representations via Disentangled Multi-Atlas Functional Connectivity Learning
May 7, 2026Minheng Chen, Chao Cao, Jing Zhang +2Multi-View Consistency Disentangled Representation Learning
Department of Computer Science and Engineering, University of Texas at Arlington, United States · School of Computing, University of Georgia, United States
BrainNext: A General-Purpose Self-Supervised Foundation Model for Brain MRI Analysis
Jul 20, 2026Moona Mazher, Abdul Qayyum, Steven A. Niederer +1Medical Imaging Foundation Models Self-Supervised Pre-Training
UCL Hawkes Institute, Department of Computer Science, University College London, London, United Kingdom · National Heart & Lung Institute, Imperial College London, London, United Kingdom · Division of Cardiovascular Medicine, Stanford Cardiovascular Institute, Stanford University, California, USA
Scaling Vision Transformers for Functional MRI with Flat Maps
Oct 15, 2025Connor Lane, Mihir Tripathy, Leema Krishna Murali +15fMRI Foundation Models Masked Autoencoders
MedARC · Sophont · Baylor College of Medicine +3

May 7, 2026J/K move · Enter open · S save

Learning Cross-Atlas Consistent Brain Disorder Representations via Disentangled Multi-Atlas Functional Connectivity Learning

Minheng Chen, Chao Cao, Jing Zhang, Tianming Liu+1

Department of Computer Science and Engineering, University of Texas at Arlington, United States · School of Computing, University of Georgia, United States

Multi-View Consistency Disentangled Representation Learning Functional Connectivity Medical Diagnosis Neuroimaging Brain Disorder Classification

Open paper AskPDF arXiv

Functional connectivity (FC) derived from resting-state fMRI is widely used to characterize large-scale brain network alterations in neurological and psychiatric disorders. However, FC construction critically depends on the choice of brain atlas, and different parcellations may emphasize distinct organizational features, leading to heterogeneous and sometimes inconsistent representations. Existing multi-atlas approaches partially alleviate this issue but often fuse atlas-derived features or predictions at a relatively shallow level, while single-atlas disentanglement methods do not explicitly address cross-atlas heterogeneity. We propose Multi-Atlas Disentangled Connectivity LEarning (MADCLE), a multi-branch representation learning framework that jointly encodes FC matrices derived from different brain atlases. Rather than introducing a single explicitly shared latent variable across parcellations, MADCLE learns atlas-wise disease-related representations and encourages them to be cross-atlas consistent through distributional alignment. Meanwhile, covariate-related and atlas-dependent residual factors are modeled separately using covariate similarity supervision, atlas-specific reconstruction, and decorrelation constraints, thereby reducing the leakage of non-disease and parcellation-dependent information into the disease-related embeddings. Experiments on the ADNI and ADHD-200 datasets suggest that MADCLE achieves competitive or improved performance compared with single-atlas baselines, multi-atlas GNN/Transformer models, and recent multi-atlas consistency frameworks. These results support the potential value of structured disentanglement for FC-based disorder identification under heterogeneous parcellation schemes.

Abstract

fMRI Foundation Models

Explore similar work

Learning Cross-Atlas Consistent Brain Disorder Representations via Disentangled Multi-Atlas Functional Connectivity Learning

BrainNext: A General-Purpose Self-Supervised Foundation Model for Brain MRI Analysis

Scaling Vision Transformers for Functional MRI with Flat Maps