Dynamic fusion of structural and functional connectivity via joint connectivity matrix ICA

Lei Wu, Marlena Duda, Armin Iraji, Vince D. CalhounbioRxiv 2025 · DOI

TL;DR

This paper introduces "dynamic fusion," a novel framework that integrates structural and functional brain connectivity data over time. By analyzing these dynamic connections, the study reveals how brain organization changes across different states and successfully identifies differences between healthy controls and individuals with schizophrenia, offering new insights into neurological disorders.

What Problem Does This Paper Address?

Neuroimaging techniques like fMRI and dMRI provide complementary views of brain activity: fMRI captures functional connectivity (how brain regions communicate), while dMRI maps structural connectivity (the physical white matter pathways). Integrating these two modalities offers a more complete picture of brain function. However, combining them, especially when considering their dynamic, moment-to-moment changes, is challenging due to their vastly different data characteristics. Existing methods often treat connectivity as static, missing crucial temporal dynamics. This paper aims to bridge this gap by developing a method to fuse dynamic structural and functional connectivity, revealing how these relationships evolve and differ across brain states and populations.

Methods

The core of this work is the "dynamic fusion" framework, an extension of joint component analysis (cmICA). The approach involves:

  • Joint Decomposition: Applying cmICA to simultaneously decompose both static structural connectivity (SC) and dynamic functional connectivity (FC) from fMRI and dMRI data.
  • Dynamic FC Extraction: Instead of using a single static FC measure for the entire scan, the authors divide the fMRI data into segments (e.g., first and last thirds of the time series) to capture dynamic FC states.
  • Component Matching and Stability Analysis: Two separate runs of joint cmICA are performed using the same SC but different dynamic FC segments. A similarity measure is then used to match the resulting components across these runs and assess their stability (how consistent they are across states) or dynamism (how much they change).
  • Multiscale Similarity Measure: A 3D multiscale similarity measure is adapted to quantify the resemblance between components, allowing for the identification of stable and dynamic multimodal components.
  • Group Comparison: A two-step method is used to compare similarity measures between healthy controls (HC) and individuals with schizophrenia (SZ) to identify group-specific differences in stable and dynamic connectivity.

Datasets

The study utilized fMRI and dMRI data from:

  • 57 individuals diagnosed with schizophrenia.
  • 60 typical control subjects.

These datasets were previously used in other studies by the authors for unimodal SC and FC analyses.

Key Insights and Ideas

  • Brain Connectivity is Both Stable and Dynamic: The framework reveals that brain connectivity isn't monolithic; some connections are remarkably stable across different brain states, while others exhibit significant dynamism, reflecting the brain's adaptability.
  • Dynamic Fusion Uncovers Hidden Differences: By considering dynamic connectivity, the method can detect structural-functional connectivity differences between groups (like HC vs. SZ) that might be missed by static analyses.
  • Schizophrenia Linked to Reduced Network Flexibility: The findings suggest that individuals with schizophrenia may have reduced network flexibility, potentially impacting their adaptive responses to cognitive demands.
  • Structural Connectivity Can Be Dynamic: Contrary to expectations, the study shows that structural connectivity patterns can also change over time, influenced by the functional states they are fused with.

Key Results

  • The dynamic fusion framework successfully identified both stable and dynamic multimodal components.
  • Most of the 30 joint source components showed high similarity (above 0.6) across different states, indicating overall stability.
  • The analysis revealed significant differences in dynamic connectivity between healthy controls and individuals with schizophrenia, particularly in structural connectivity (SC) maps.
  • Specifically, two dynamic components showed substantial group dissimilarity in both SC and FC maps, suggesting disruptions in adaptive neural responses in schizophrenia.

Limitations and Future Work

  • Reliance on Two Runs: The study used two separate runs of joint cmICA, which limits the analysis of continuous dynamic changes and may over-rely on segment assignments. Future work should explore integrating findings from multiple runs into a more cohesive dynamic cmICA framework.
  • Dissimilarity Measure Properties: The adapted 3D multiscale similarity measure used for group comparisons does not strictly adhere to mathematical properties of a valid distance function, potentially complicating the interpretation of group differences. Future studies could employ metrics that satisfy these properties.
  • Further Examination of State-wise Differences: State-wise differences at the individual subject level require further investigation.
  • Clinical Implications: The findings highlight the importance of assessing both stable and dynamic connectivity for understanding neurological disorders like schizophrenia and could inform targeted interventions. Future research should continue to explore this interplay in clinical contexts.

Summary updated

Abstract

Abstract The integration of multimodal MRI images, including functional MRI (fMRI) and diffusion MRI (dMRI), depicts a key advancement in neuroimaging, since it offers a more comprehensive and better understanding of brain function and connections. FMRI captures brain functional activity while dMRI reveals structural connectivity via white matter bundles, each of which provide unique yet complementary insights; however combing these two modalities, particularly at a dynamic level, is challenging due to their drastically different data characteristics. This study introduces a novel framework named “dynamic fusion,” which extends joint component analysis (cmICA) to integrate static structural connectivity (SC) with dynamic functional connectivity (FC). Our approach gauges the relationship between these joint components across various temporal states, aiming to discover both static and dynamic features of brain connectivity. We applied this approach to fMRI and dMRI data from the same set of control subjects, which we had previously studied to estimate joint parcellation and their structural and functional connections using a static model only, and also included a comparable number of individuals with schizophrenia from the same study. Our results reveal that dynamic fusion not only highlights diverse temporal dynamics in FC but, more importantly, also shows how SC patterns differ across dynamic functional states at different time frames, providing new insights into brain organization. Furthermore, it successfully detects joint structural-functional connectivity differences between individuals with schizophrenia and controls, demonstrating its potential for detecting hidden group differences. Overall, this study establishes our dynamic fusion as a powerful tool for integrating structural and dynamic functional connectivity data, enhancing our understanding of brain connectivity and offering new perspectives for studying neurological disorders.
BibTeX
@misc{wu2025dynamic,
  author = {Lei Wu and Marlena Duda and Armin Iraji and Vince D. Calhoun},
  title = {{Dynamic fusion of structural and functional connectivity via joint connectivity matrix ICA}},
  year = {2025},
  doi = {10.1101/2025.05.15.653851},
  howpublished = {bioRxiv},
}
Figure