Structural Connectome Harmonization Using Deep Learning: The Strength of Graph Neural Networks

2025-07-18

AI Summary

Multi-site neuroimaging studies suffer from site-related biases induced by scanner heterogeneity. Conventional harmonization methods—such as ComBat, linear regression, or generic deep learning approaches—typically require extensive metadata or traveling subjects and neglect the intrinsic graph-topological structure of structural connectomes (SCs). To address this, we propose a site-conditioned graph autoencoder framework that requires neither metadata nor traveling subjects; to our knowledge, this is the first work embedding graph neural networks into SC harmonization. Our method preserves individual connectomic fingerprints and network topology while enabling cross-site data integration. Experiments demonstrate that our approach significantly outperforms fully connected and convolutional autoencoders, as well as linear regression baselines, in both topological fidelity and subject identification accuracy. It further ensures statistical robustness and clinical generalizability, offering a novel paradigm for biomarker discovery in neuropsychiatric disorders with limited sample sizes.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Search and Optimization: Metareasoning and Metaheuristics
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Small sample sizes in neuroimaging in general, and in structural connectome (SC) studies in particular limit the development of reliable biomarkers for neurological and psychiatric disorders - such as Alzheimer's disease and schizophrenia - by reducing statistical power, reliability, and generalizability. Large-scale multi-site studies have exist, but they have acquisition-related biases due to scanner heterogeneity, compromising imaging consistency and downstream analyses. While existing SC harmonization methods - such as linear regression (LR), ComBat, and deep learning techniques - mitigate these biases, they often rely on detailed metadata, traveling subjects (TS), or overlook the graph-topology of SCs. To address these limitations, we propose a site-conditioned deep harmonization framework that harmonizes SCs across diverse acquisition sites without requiring metadata or TS that we test in a simulated scenario based on the Human Connectome Dataset. Within this framework, we benchmark three deep architectures - a fully connected autoencoder (AE), a convolutional AE, and a graph convolutional AE - against a top-performing LR baseline. While non-graph models excel in edge-weight prediction and edge existence detection, the graph AE demonstrates superior preservation of topological structure and subject-level individuality, as reflected by graph metrics and fingerprinting accuracy, respectively. Although the LR baseline achieves the highest numerical performance by explicitly modeling acquisition parameters, it lacks applicability to real-world multi-site use cases as detailed acquisition metadata is often unavailable. Our results highlight the critical role of model architecture in SC harmonization performance and demonstrate that graph-based approaches are particularly well-suited for structure-aware, domain-generalizable SC harmonization in large-scale multi-site SC studies.

Problem

• Small sample sizes limit reliable neurological disorder biomarkers
• Multi-site studies suffer from scanner-related biases and inconsistencies
• Existing harmonization methods overlook SC graph-topology or need metadata

Innovation

• Site-conditioned deep harmonization framework for SCs
• Graph AE preserves topological structure effectively
• No metadata or traveling subjects required