NeuroAlign: Hierarchical Multimodal Fusion of Dynamic and Structural Neuroimaging for MCI Analysis

2026-05-31

AI Summary

This study addresses the challenge of multimodal fusion in mild cognitive impairment (MCI) analysis arising from heterogeneous feature spaces and misaligned representations between functional MRI (fMRI) and diffusion tensor imaging (DTI). To overcome this, the authors propose a hierarchical multimodal fusion framework incorporating dual-modality hierarchical alignment (DMHA) and dual-domain hierarchical interaction (DDHI) mechanisms. These enable fine-grained alignment and complementary integration of multiscale features—including dynamic and static functional connectivity, amplitude of low-frequency fluctuations (ALFF), and fractional anisotropy (FA)—across both functional–structural and region–connection hierarchies. Additionally, a gradient-free synergistic activation mapping (SAM) module is introduced to support interpretable analysis. Evaluated on the GUTCM, ADNI, and OASIS datasets, the model significantly improves MCI and subjective cognitive decline (SCD) detection performance, demonstrates strong cross-dataset generalizability, and reveals neuroimaging biomarkers with high interpretability.

Abstract

Multimodal neuroimaging fusion of functional MRI (fMRI) and diffusion tensor imaging (DTI) provides complementary information for cognitive impairment analysis, but remains challenged by heterogeneous feature spaces and misaligned representations. We propose \textit{NeuroAlign}, a hierarchical framework for structured multimodal fusion. It introduces (1) \textit{Dual-Modal Hierarchical Alignment} (DMHA), which models multi-scale dynamic connectivity and aligns dynamic-static and functional-structural embeddings; and (2) \textit{Dual-Domain Hierarchical Interaction} (DDHI), which enables fine-grained modulation and global interaction between connectivity- and region-level features. To support feature-level inspection, we design \textit{Synergistic Activation Mapping} (SAM), a gradient-free, marker-oriented attribution method for DFC, SFC, ALFF, and FA. Evaluated on GUTCM, ADNI, and OASIS under five-fold validation, NeuroAlign achieves competitive MCI/SCD detection and preliminary cross-dataset transferability. Attribution analyses reveal modality-specific and partially consistent brain patterns, providing model-derived evidence for multimodal representation analysis.

Problem

• multimodal neuroimaging fusion
• heterogeneous feature spaces
• misaligned representations
• MCI analysis
• functional-structural alignment

Innovation

• Hierarchical Multimodal Fusion
• Dual-Modal Hierarchical Alignment
• Dual-Domain Hierarchical Interaction
• Synergistic Activation Mapping
• Cross-dataset Transferability