CoHyFuse: Condition-wise Hypergraph Fusion with Global Connectome in Task-fMRI
Abstract
Task-fMRI connectomes reveal state-dependent neural reconfigurations, yet conventional methods marginalize these signals by aggregating distinct conditions into static pairwise graphs, thereby obscuring condition-specific multi-ROI organization. We introduce CoHyFuse, a condition-aware ROI-centered hypergraph framework that constructs a task-state-specific incidence matrix from condition-wise functional connectivity (FC)-profile embeddings, allowing the same ROI to form different multi-ROI hyperedges across task phases. Condition-specific neighborhood sizes Kq further adapt the hyperedge scale to each task state, and the resulting condition embeddings are fused with a complementary whole-session FC branch for prediction. In the AABC cohort (N=1,074), CoHyFuse achieved the best mean out-of-fold predictive performance among evaluated baselines on FACENAME Fluid Cognition Composite (FCC) prediction (7.83±0.10 MAE, 0.439±0.026 R2) and VISMOTOR age prediction (7.52±0.37 MAE, 0.592±0.022 R2). In an auxiliary CMI-HBN attention-deficit/hyperactivity disorder (ADHD) classification benchmark (N=223), CoHyFuse obtained 72.0±2.1\% macro-AUC and 74.2±2.9\% accuracy. Ablation studies support the contributions of condition-wise incidence construction and dual-view fusion, suggesting that state-resolved ROI-set structure provides complementary predictive information beyond whole-session FC alone. Occlusion analysis identifies the Distraction condition as the primary driver of model prediction, pointing toward the Salience/Ventral Attention Network (SAN)--FrontoParietal Network (FPN) and within-SAN hyperedge-defined ROI-set motifs as candidate model-relevant patterns. This framework provides an interpretable, state-resolved view of the connectome for downstream cohort analysis.
Related
- FroHyNet: frozen temporal priors and trainable hypergraph fusion for brain time-series classification.
- Connectome modulates critical brain dynamics across local and global scales
- RepGene: Auditing Condition-Specific Fusion Biases for Multi-View Gene Representations
- Satellite-gene abundance fusion reveals global hotspots of ocean ecosystem services
- Modular core network constructed from Escherichia coli transcriptome datasets using a hypergraph-based pan-network approach
BibTeX
@misc{kim2026cohyfuse,
title = {{CoHyFuse: Condition-wise Hypergraph Fusion with Global Connectome in Task-fMRI}},
author = {Boseong Kim and Haejun Chung and Ikbeom Jang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.05913},
eprint = {2610.05913},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.05913},
}
RIS
TY - UNPB
TI - CoHyFuse: Condition-wise Hypergraph Fusion with Global Connectome in Task-fMRI
AU - Boseong Kim
AU - Haejun Chung
AU - Ikbeom Jang
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2610.05913
UR - https://arxiv.org/abs/2610.05913
AB - Task-fMRI connectomes reveal state-dependent neural reconfigurations, yet conventional methods marginalize these signals by aggregating distinct conditions into static pairwise graphs, thereby obscuring condition-specific multi-ROI organization. We introduce CoHyFuse, a condition-aware ROI-centered hypergraph framework that constructs a task-state-specific incidence matrix from condition-wise functional connectivity (FC)-profile embeddings, allowing the same ROI to form different multi-ROI hyperedges across task phases. Condition-specific neighborhood sizes K_q further adapt the hyperedge scale to each task state, and the resulting condition embeddings are fused with a complementary whole-session FC branch for prediction. In the AABC cohort (N=1,074), CoHyFuse achieved the best mean out-of-fold predictive performance among evaluated baselines on FACENAME Fluid Cognition Composite (FCC) prediction (7.83±0.10 MAE, 0.439±0.026 R^2) and VISMOTOR age prediction (7.52±0.37 MAE, 0.592±0.022 R^2). In an auxiliary CMI-HBN attention-deficit/hyperactivity disorder (ADHD) classification benchmark (N=223), CoHyFuse obtained 72.0±2.1\% macro-AUC and 74.2±2.9\% accuracy. Ablation studies support the contributions of condition-wise incidence construction and dual-view fusion, suggesting that state-resolved ROI-set structure provides complementary predictive information beyond whole-session FC alone. Occlusion analysis identifies the Distraction condition as the primary driver of model prediction, pointing toward the Salience/Ventral Attention Network (SAN)--FrontoParietal Network (FPN) and within-SAN hyperedge-defined ROI-set motifs as candidate model-relevant patterns. This framework provides an interpretable, state-resolved view of the connectome for downstream cohort analysis.
ER -