Edge-boosted graph learning for functional brain connectivity analysis
David Yang, Mostafa Abdelmegeed, John Modl, Minjeong Kim
TL;DR
This paper tackles the inadequacy of node-based functional connectivity for diagnosing neurodegenerative diseases by introducing edge functional connectivity (eFC) derived from edge time signals and a co-embedding mechanism that jointly models node and edge attributes within a GNN. The method computes edge time signals $r_{ij}=z_i z_j$ from $z$-scored ROI time series, builds an edge-Temporal connectivity matrix $E_{FC}$ via $E_{FC}=\frac{E_{TS}^T E_{TS}}{\sqrt{d}\cdot\sqrt{d}^T}$ with $d=\mathrm{diag}(E_{TS}^T E_{TS})$, and updates node representations using $H_i^{(l+1)}=\sigma\left(H^{(l)}\left[ W_0^{(l)}+\Phi(E_{efc}) W_1^{(l)} \right]\right)$. Evaluated on ADNI and PPMI datasets, the approach outperforms CNN, GCN, CRGNN, and MGNN across Accuracy, Precision, and F1, with particularly strong gains on ADNI, indicating that edge-centric representations better capture brain connectivity dynamics relevant to disease states. The work demonstrates that incorporating edge–edge interactions via co-embedding can significantly enhance brain disease classification and offers a principled path toward clinically impactful neuroimaging analytics.
Abstract
Predicting disease states from functional brain connectivity is critical for the early diagnosis of severe neurodegenerative diseases such as Alzheimer's Disease and Parkinson's Disease. Existing studies commonly employ Graph Neural Networks (GNNs) to infer clinical diagnoses from node-based brain connectivity matrices generated through node-to-node similarities of regionally averaged fMRI signals. However, recent neuroscience studies found that such node-based connectivity does not accurately capture ``functional connections" within the brain. This paper proposes a novel approach to brain network analysis that emphasizes edge functional connectivity (eFC), shifting the focus to inter-edge relationships. Additionally, we introduce a co-embedding technique to integrate edge functional connections effectively. Experimental results on the ADNI and PPMI datasets demonstrate that our method significantly outperforms state-of-the-art GNN methods in classifying functional brain networks.