Multimodal Graph Neural Networks for Prognostic Modeling of Brain Network Reorganization

TL;DR

The paper presents a Multimodal Graph Neural Network (MGNN) framework to prognosticate brain network reorganization using longitudinal, multimodal neuroimaging data. By integrating structural MRI, DTI, fMRI, PET, CSF biomarkers, and clinical scores within a graph-based, fractional stochastic dynamics model and cross-modality attention, it yields interpretable biomarkers such as network energy, graph curvature, and diffusion centrality. Empirically, MGNN outperforms unimodal and classical baselines in predicting reorganization risk and provides mechanistic insights, supporting its potential for clinically meaningful prognostic biomarkers. The work emphasizes rigorous, interpretable graph-based modeling as a path toward translational neuroscience and decision-support tools.

Abstract

Understanding the dynamic reorganization of brain networks is critical for predicting cognitive decline, neurological progression, and individual variability in clinical outcomes. This work proposes a multimodal graph neural network framework that integrates structural MRI, diffusion tensor imaging, and functional MRI to model spatiotemporal brain network reorganization. Brain regions are represented as nodes and structural and functional connectivity as edges, forming longitudinal brain graphs for each subject. Temporal evolution is captured via fractional stochastic differential operators embedded within graph-based recurrent networks, enabling the modeling of long-term dependencies and stochastic fluctuations in network dynamics. Attention mechanisms fuse multimodal information and generate interpretable biomarkers, including network energy entropy, graph curvature, fractional memory indices, and modality-specific attention scores. These biomarkers are combined into a composite prognostic index to quantify individual risk of network instability or cognitive decline. Experiments on longitudinal neuroimaging datasets demonstrate both predictive accuracy and interpretability. The results highlight the potential of mathematically rigorous, multimodal graph-based approaches for deriving clinically meaningful biomarkers from existing imaging data without requiring new data collection.

Figures (2)

• Figure 1: System Architecture of the proposed Multimodal Graph Neural Network framework for brain network reorganization modeling.
• Figure 2: Heatmap of cross-modality attention weights over brain regions for a representative subject.