[论文解读] Aberrant High-Order Dependencies in Schizophrenia Resting-State Functional MRI Networks

Qiang Li, Vince D. Calhoun+4 more|arXiv (Cornell University)|Oct 26, 2023

Functional Brain Connectivity Studies被引用 7

一句话总结

The paper uses high-order ICA and multivariate information-theoretic metrics (O-information and S-information) to reveal that aberrant high-order dependencies in resting-state fMRI networks differentiate schizophrenia patients from controls, highlighting synergy-dominated patterns as potential biomarkers.

ABSTRACT

The human brain has a complex, intricate functional architecture. While many studies primarily emphasize pairwise interactions, delving into high-order associations is crucial for a comprehensive understanding of how functional brain networks intricately interact beyond simple pairwise connections. Analyzing high-order statistics allows us to explore the nuanced and complex relationships across the brain, unraveling the heterogeneity and uncovering patterns of multilevel overlap on the psychosis continuum. Here, we employed high-order independent component analysis (ICA) plus multivariate information-theoretical metrics ($O$-information and $S$-information) to estimate high-order interaction to examine schizophrenia using resting-state fMRI. The results show that multiple brain regions networks may be altered in schizophrenia, such as temporal, subcortical, and higher-cognitive brain regions, and meanwhile, it also shows that revealed synergy gives more information than redundancy in diagnosing schizophrenia. All in all, we showed that high-order dependencies were altered in schizophrenia. Identification of these aberrant patterns will give us a new window to diagnose schizophrenia.

研究动机与目标

• Investigate whether high-order brain network interactions are altered in schizophrenia beyond pairwise connectivity.
• Apply high-model-order independent component analysis (ICA) to rsfMRI data to extract intrinsic connectivity networks (ICNs).
• Use information-theoretic measures (O-information and S-information) to quantify high-order redundancy and synergy among ICNs.
• Determine whether high-order interactions reveal patterns not captured by traditional low-order connectivity.
• Assess potential brain regions and network groups implicated in schizophrenia through high-order metrics.

提出的方法

• Perform group ICA on two datasets to extract 105 ICNs per subject and categorize them into VI, CB, TP, SC, SM, HC.
• Match ICNs to a predefined NeuroMark template to classify components into six network groups.
• Model fMRI signals as Gaussian and compute high-order information measures (TC, DTC, O-information, S-information) for interaction orders 3, 4, 5.
• Use Gaussian entropy and joint entropy formulas to estimate TC and related metrics for multivariate Gaussian variables.
• Compare low-order (pairwise) FC to high-order connectivity (HOFC) and analyze redundancy vs synergy across groups.
• Highlight that HOFC captures more aberrant networks and that synergy information may be a stronger discriminator for schizophrenia.

Figure 1: Flowchart of capturing high-order dependencies among brain connectivity states. A. Group ICA is performed on two independent healthy control and schizophrenia datasets, and corresponding 105 spatial network maps and time courses are extracted from each subject. B. The estimated independent

实验结果

研究问题

• RQ1 Do aberrant high-order dependencies exist in schizophrenia that are not captured by pairwise functional connectivity?
• RQ2 Can high-order ICA combined with multivariate information theory reveal synergy-driven patterns in rsfMRI that distinguish schizophrenia from controls?
• RQ3 Which brain network groups (VI, CB, TP, SC, SM, HC) show altered high-order interactions in schizophrenia?
• RQ4 Is synergy information more informative than redundancy for diagnosing schizophrenia using HOFC?

主要发现

• HOFC reveals aberrant interactions involving visual, somatomotor, temporal, subcortical, and higher cognitive regions in schizophrenia.
• Synergy information contributes more than redundancy in distinguishing schizophrenia from controls.
• Redundancy information is mainly distributed in SC and HC regions and becomes more prominent with higher interaction order.
• High-order interactions capture more aberrant brain networks (TP, SC, HC) than low-order analyses.
• The approach suggests high-order dependencies altered in schizophrenia offer a new biomarker window for diagnosis.

Figure 2: Comparison of low-order functional connectivity (LOFC) and high-order functional connectivity (HOFC) between HC subjects and SC patients. The top row shows that LOFC and the main difference present a pair-wise interaction between VI-SM and TP-HC. The bottom row presents the HOFC with redun