CFBrainAug: Modularity-Induced Counterfactual Brain Network Augmentation for Interpretable fMRI-based Autism Diagnosis

Yifan Qiao; Sheng Li

cs.LGOct 23, 2026

CFBrainAug: Modularity-Induced Counterfactual Brain Network Augmentation for Interpretable fMRI-based Autism Diagnosis

Authors: Yifan Qiao, Sheng Li

Organizations: University of Virginia, Charlottesville, VA, USA

Abstract

Machine learning on resting-state functional magnetic resonance imaging (rs-fMRI) brain networks is promising for diagnosing Autism Spectrum Disorder (ASD), but existing methods are prone to overfitting on limited training data, and their post-hoc statistical attribution offers limited clinical interpretability. We propose CFBrainAug, a counterfactual brain network learning framework that unifies data augmentation and counterfactual explanation by learning label-altered brain networks. Unlike conventional label-preserving augmentation, CFBrainAug first retrieves a label-opposite yet topologically similar candidate and then refines it under a module-number constraint. Jointly training on factual and counterfactual networks enriches the training distribution around the decision boundary, enabling a lightweight feature-selection-based SVM to perform competitively with recent graph deep learning methods. Meanwhile, the counterfactuals provide intrinsic explanations by identifying the connectivity differences between each subject and its counterfactual. Experiments on the ABIDE I dataset show that CFBrainAug reaches an accuracy of 74.1% on the NYU site, suggesting that counterfactual augmentation can complement more complex models in small-sample brain network analysis.

Figures & tables

Fig. 1: Comparison between conventional learning and CFBrainAug. Unlike conventional methods that separately optimize generalization and interpretability, CFBrainAug unifies both through counterfactual brain networks.

Fig. 2: Overview of CFBrainAug. (A) Original brain network construction from rs-fMRI. (B) Candidate counterfactual brain networks are retrieved and refined through multi-constraint optimization. (C) Original and counterfactual brain networks are jointly used for RBF-SVM diagnosis, and the most altered connections between factual and counterfactual networks serve as counterfactual explanations.

Site	Category	Subject	M	F	Age
NYU	ASD	79	68	11	14.5 (6.9)
NYU	NC	105	79	26	15.8 (6.2)
UM	ASD	68	58	10	13.1 (2.4)
UM	NC	77	59	18	14.7 (3.5)
LEUVEN	ASD	29	26	3	17.7 (5.0)
LEUVEN	NC	35	30	5	18.1 (4.9)

TABLE I: Demographics of the ABIDE I subjects. Age: mean (standard deviation); M/F: male/female.

Method	Site NYU				Site UM				Site LEUVEN
Method	ACC (%)	AUC (%)	SEN (%)	SPE (%)	ACC (%)	AUC (%)	SEN (%)	SPE (%)	ACC (%)	AUC (%)	SEN (%)	SPE (%)
BrainGNN [ 3 ]	63.9 (5.0) †	68.1 (5.7) †	49.0 (8.7) †	72.0 (9.1)	59.4 (12.8) †	73.6 (9.8)	58.9 (12.3)	62.6 (14.9)	63.5 (3.7) †	69.3 (4.5) †	66.6 (10.6)	61.7 (9.7) †
STGCN [ 25 ]	70.4 (4.8)	74.3 (4.5)	67.8 (7.0)	73.7 (6.3)	65.3 (3.9) †	66.7 (3.6) †	42.4 (4.2) †	84.4 (8.7)	51.4 (2.0) †	58.3 (7.1) †	57.9 (13.3)	45.2 (14.9) †
GoFormer [ 26 ]	72.5 (4.6)	76.7 (5.1)	63.9 (6.5)	79.4 (6.6)	62.3 (4.7) †	72.7 (7.1)	52.5 (13.9)	73.8 (7.8)	63.7 (5.4) †	60.4 (6.7) †	50.6 (7.3)	74.5 (6.6)
Miniformer [ 27 ]	62.7 (2.6) †	66.8 (4.6) †	54.9 (6.4) †	64.1 (4.1) †	67.7 (1.8) †	69.3 (1.3) †	60.7 (2.7) †	74.6 (4.9)	66.9 (2.6)	69.4 (6.2)	58.9 (8.7)	78.6 (7.6)
WindowSlicing [ 13 ]	69.2 (4.4)	74.9 (5.3)	68.8 (9.0)	70.1 (7.9)	72.9 (2.6)	78.7 (2.7)	70.8 (2.3)	75.1 (5.6)	69.8 (6.2)	74.6 (7.7)	63.9 (14.7)	76.2 (10.9)

TABLE II: Performance comparison on the ABIDE I NYU, UM, and LEUVEN sites, reported as mean (standard deviation) over 10 repetitions of 5-fold cross-validation (best in bold; † : significantly worse than CFBrainAug,

p<0.05

Method	ACC (%)	AUC (%)	SEN (%)	SPE (%)
Full CFBrainAug (Ours)	74.1 (2.6)	79.2 (4.7)	66.7 (6.2)	79.5 (6.8)
w/o ModularConstraint	66.8 (10.6)	71.0 (14.4)	65.7 (10.6)	67.6 (6.2)
w/o Refine	67.4 (9.8)	71.7 (9.6)	59.5 (8.7)	73.3 (12.4)
w/o CounterfactualAug	66.3 (10.7)	74.0 (10.2)	65.8 (12.7)	66.7 (13.6)

TABLE III: Ablation study on the ABIDE I NYU site (best in bold).

Fig. 3: Group-level (top) and subject-specific (middle and bottom) counterfactual explanations on the NYU site. The group-level map averages $|\Delta A|$ over all test subjects; the subject-specific maps correspond to two randomly selected correctly classified subjects. Warmer colors indicate larger counterfactual perturbation strength $s_{i}$ (Sec. IV-F ).

Fig. 4: Top-10 connections from the group-level counterfactual explanation on the NYU site. Thicker edges indicate larger differences.

Fig. 5: Sensitivity to the number of modules $k$ (left) and the counterfactual weight $\lambda$ (right) on the NYU site.

Counterfactual Data Augmentation

Counterfactual data augmentation creates plausible examples by changing a chosen feature while preserving relevant context, helping models learn robust distinctions.

See all 38 papers

MCDA: Multi-View Graph Contrastive Learning with Dynamic Topology Augmentation for Functional MRI Analysis
Oct 19, 2026Hao Zhang, Xiaoyun Liu, Shuo Huang +4Graph Contrastive Learning Functional Magnetic Resonance Imaging
College of Biomedical Engineering, Fudan University, Shanghai 200433, China · College of Information Science and Technology & Artificial Intelligence, Nanjing Forestry University, Nanjing 210037, China · Department of Psychosomatics and Psychiatry, Zhongda Hospital, School of Medicine, Jiangsu Provincial Key Laboratory of Brain Science and Medicine, Southeast University, Nanjing 210009, China +3
Beyond Anatomy: Explainable ASD Classification from rs-fMRI via Functional Parcellation and Graph Attention Networks
Mar 3, 2026Syeda Hareem Madani, Noureen Bibi, Adam Rafiq Jeraj +3Autism Spectrum Disorder Neuroimaging
Salim Habib University · VentureDive · National University of Computer and Emerging Sciences
Continual Learning for fMRI-Based Brain Disorder Diagnosis via Functional Connectivity Matrices Generative Replay
Apr 15, 2026Qianyu Chen, Shujian YuContinual Learning Replay-Based Continual Learning
†Nanyang Technological University · ‡VU Amsterdam · §UiT The Arctic University of Norway

Oct 19, 2026J/K move · Enter open · S save

MCDA: Multi-View Graph Contrastive Learning with Dynamic Topology Augmentation for Functional MRI Analysis

Hao Zhang, Xiaoyun Liu, Shuo Huang, Yue Ma+3

College of Biomedical Engineering, Fudan University, Shanghai 200433, China · College of Information Science and Technology & Artificial Intelligence, Nanjing Forestry University, Nanjing 210037, China · Department of Psychosomatics and Psychiatry, Zhongda Hospital, School of Medicine, Jiangsu Provincial Key Laboratory of Brain Science and Medicine, Southeast University, Nanjing 210009, China+3

Graph Contrastive Learning Functional Magnetic Resonance Imaging Graph Data Augmentation Functional Connectivity

Open paper AskPDF arXiv

Resting-state fMRI (rs-fMRI) is a vital non-invasive tool for diagnosing brain disorders and uncovering pathological mechanisms. However, existing graph contrastive learning (GCL) methods often underexploit augmentation histories across training epochs and the relationships among different augmented views. In this paper, we introduce a topology-guided Multi-view graph Contrastive learning framework with Dynamic graph topology Augmentation (MCDA) for fMRI analysis. Unlike traditional GCL that uses fixed noise or edge-dropping, MCDA employs a Dynamic Augmentation Update (DAU) module that iteratively refines the augmented topology. Specifically, we incorporate self-aware and cross-sample augmentation strategies to simultaneously capture individual-specific functional signatures and inter-subject heterogeneity. We then design a topology-guided multi-view graph contrastive learning scheme that explores interrelationships not only between the original network and its augmented views, but also across different augmented views to maximize structural information preservation. We characterize the score-level optimization directions and establish a lower bound for the contrastive objective. The framework combines classification and contrastive losses for brain disorder diagnosis and examines discriminative functional connectivity patterns as potential biomarkers. Experiments on two public rs-fMRI datasets comprising 1,236 subjects demonstrate that MCDA outperforms the state-of-the-art methods. The code has been made publicly available at: https://github.com/goodcodingagoodboy/MCDA.

Abstract

Counterfactual Data Augmentation

Explore similar work

MCDA: Multi-View Graph Contrastive Learning with Dynamic Topology Augmentation for Functional MRI Analysis

Beyond Anatomy: Explainable ASD Classification from rs-fMRI via Functional Parcellation and Graph Attention Networks

Continual Learning for fMRI-Based Brain Disorder Diagnosis via Functional Connectivity Matrices Generative Replay