4D DYNAMIC SPATIAL BRAIN NETWORKS AT REST LINKED TO COGNITION SHOW ATYPICAL VARIABILITY AND COUPLING IN SCHIZOPHRENIA

Krishna Pusuluri, Zening Fu, Robyn L. Miller, Godfrey D. Pearlson, Peter Kochunov, Theo G.m. van Erp, Armin Iraji, Vince D. CalhounbioRxiv 2023 · DOI

TL;DR

This paper introduces a novel method to analyze the dynamic, 4D nature of brain networks by measuring their volumetric changes over time. They found that these dynamic network features are linked to cognitive performance and are significantly altered in individuals with schizophrenia, suggesting a new avenue for understanding brain disorders.

What Problem Does This Paper Address?

Traditional functional magnetic resonance imaging (fMRI) studies often treat brain networks as static entities, analyzing their connectivity patterns at a fixed spatial configuration. However, the brain is a dynamic system, and its functional networks likely change in size and shape over time. This "4D" aspect – three spatial dimensions plus time – is crucial for understanding brain function and dysfunction. Previous research has explored temporal fluctuations in connectivity (dynamic functional connectivity), but this paper focuses on the spatial dynamics of these networks, specifically how their volumes change and how these changes are coupled between different networks. The authors aim to bridge this gap by developing a method to quantify these spatial dynamics and investigate their relationship with cognitive abilities and their alterations in schizophrenia, a disorder known for cognitive impairments.

Methods

The study employs a multi-step analysis pipeline:

  1. Group-Level Spatial ICA: Standard spatially constrained independent component analysis (sICA) is used on resting-state fMRI (rsfMRI) data to identify 14 key intrinsic connectivity networks (ICNs) across a large group of subjects. This step establishes a common set of networks to track.
  2. Subject-Level MOO-ICAR: A sliding-window approach combined with multi-objective optimization ICA with reference (MOO-ICAR) is applied to each subject's data. This method ensures that the identified ICNs remain consistent across different time windows and subjects, allowing for the tracking of dynamic changes within each network.
  3. Volumetric Analysis: For each time window, the "volume" of each network is calculated. This is defined as the number of voxels within the network that exhibit a z-scored BOLD signal above a certain threshold (Vth). This allows for the quantification of how much of the network is "active" at any given time.
  4. Volumetric Coupling (VC): The temporal correlation between the volumes of pairs of networks is computed. A high positive correlation indicates that two networks tend to grow and shrink together (synchronized), while a high negative correlation suggests they tend to do the opposite (one grows as the other shrinks). This creates a "volumetric coupling" matrix for each subject.
  5. Hypothesis Testing:
    • 1-sample t-tests: Used to identify significant volumetric coupling between network pairs within each group (controls and schizophrenia patients).
  6. 2-sample t-tests: Used to compare mean network volumes and volumetric coupling between network pairs between the control and schizophrenia groups.
  7. Robust Linear Regression: Used to assess the association between dynamic spatial brain network features (mean/std network volumes, VC) and cognitive scores, symptom severity, and drug dosage.

The novelty lies in the focus on voxel-wise spatial dynamics (volume changes) of networks and their inter-network volumetric coupling, rather than just connectivity strength.

Datasets

The study analyzed rsfMRI data from 508 subjects collected across three existing studies:

  • The Functional Imaging Biomedical Informatics Research Network (FBIRN)
  • The Center for Biomedical Research Excellence (COBRE)
  • A Maryland Psychiatric Research Center (MPRC) study

This cohort comprised 315 healthy controls (CN) and 193 individuals with schizophrenia (SZ). The paper notes that subjects with genomic data were selected for inclusion.

Key Insights and Ideas

  • Brain Networks are Dynamic Sculptures: The study moves beyond static network definitions, revealing that brain networks actively "breathe" – they shrink and grow over time. This volumetric dynamism is a fundamental aspect of brain function.
  • Synchronized Network Behavior: Different brain networks don't act in isolation; they exhibit synchronized volumetric changes. This "volumetric coupling" suggests coordinated activity across distributed brain regions.
  • Cognition is Tied to Dynamic Network Features: The degree of dynamic variability within networks and the strength of volumetric coupling between network pairs are positively associated with cognitive performance. This implies that a more dynamic and coordinated brain is a more cognitively adept brain.
  • Schizophrenia Disrupts Dynamic Network Architecture: Individuals with schizophrenia show significant alterations in these dynamic network features. Specifically, they exhibit lower spatial dynamical variability and reduced volumetric coupling between network pairs. This suggests that the breakdown in cognitive function in schizophrenia may be partly due to a less dynamic and less coordinated brain network system.
  • Focus of Activity Matters: The study highlights that the focus of a network's activity (i.e., whether it's more concentrated or widespread) also impacts cognitive scores, further emphasizing the importance of spatial dynamics.

Key Results

  • 14 Key Brain Networks Identified: The analysis successfully identified 14 relevant brain networks, including visual, somatomotor, attention, frontal, default mode, salience, and temporal networks.
  • Significant Volumetric Coupling: Across groups, 62 out of 91 possible network pairs showed significant volumetric coupling at a specific threshold (Vth=2), with 34 being positively coupled and 28 negatively coupled.
  • Schizophrenia Shows Reduced Coupling: In individuals with schizophrenia, the number of network pairs with significant volumetric coupling was reduced (45 pairs), with fewer positive (30) and negative (15) couplings compared to controls.
  • Altered Network Dynamics in SZ:
    • Several networks showed significantly different mean volumes in SZ compared to controls, with some showing lower volumes (e.g., SUB, CER, FPN-R) and others higher volumes (e.g., ATN, pDMN) at certain thresholds.
  • The variability (standard deviation) of network volumes was generally lower in SZ patients across multiple networks, indicating reduced spatial dynamism.
  • Association with Cognition: Robust linear regression revealed significant associations between dynamic network features and cognitive scores. For instance, more focused activity in certain networks was linked to better visual learning, while higher dynamic variability and stronger volumetric coupling were generally associated with improved cognitive performance.

Limitations and Future Work

  • Voxel-wise Thresholding: The choice of volume threshold (Vth) influences the results, and exploring a wider range of thresholds is important for a comprehensive understanding. The paper acknowledges that different Vths reveal networks at different levels of granularity.
  • Medication Effects: While the study controlled for CPZ drug scores, the full impact of medication on dynamic network features warrants further investigation.
  • Specific Network Dysfunctions: The paper suggests future work should delve deeper into specific network dysfunctions, such as the visual system and the default mode network (DMN), and their potential links to observed changes in SZ.
  • Mechanisms of SZ: The authors propose that reduced network-level spatial dynamics and inter-network VC in SZ might be a mechanism underlying cognitive deficits. Future research could explore this further.
  • Other Dynamic Metrics: The paper suggests exploring other metrics for dynamic spatial brain network coupling beyond volumetric measures, and adapting techniques like dynamic mode decomposition.
  • Sliding Window Length: The impact of different sliding window lengths on the observed dynamic patterns is also a direction for future research.

Summary updated

Abstract

ABSTRACT Despite increasing interest in the dynamics of functional brain networks, most studies focus on the changing relationships over time between spatially static networks or regions. Here we propose an approach to study dynamic spatial brain net-works in human resting state functional magnetic resonance imaging (rsfMRI) data and evaluate the temporal changes in the volumes of these 4D networks. Our results show significant volumetric coupling (i.e., synchronized shrinkage and growth) between networks during the scan. We find that several features of such dynamic spatial brain networks are associated with cognition, with higher dynamic variability in these networks and higher volumetric coupling between network pairs positively associated with cognitive performance. We show that these networks are modulated differently in individuals with schizophrenia versus typical controls, resulting in network growth or shrinkage, as well as altered focus of activity within a network. Schizophrenia also shows lower spatial dynamical variability in several networks, and lower volumetric coupling between pairs of networks, thus upholding the role of dynamic spatial brain networks in cognitive impairment seen in schizophrenia. Our data show evidence for the importance of studying the typically overlooked voxelwise changes within and between brain networks.
BibTeX
@misc{pusuluri2023d,
  author = {Krishna Pusuluri and Zening Fu and Robyn L. Miller and Godfrey D. Pearlson and Peter Kochunov and Theo G.M. van Erp and Armin Iraji and Vince D. Calhoun},
  title = {{4D DYNAMIC SPATIAL BRAIN NETWORKS AT REST LINKED TO COGNITION SHOW ATYPICAL VARIABILITY AND COUPLING IN SCHIZOPHRENIA}},
  year = {2023},
  doi = {10.1101/2023.09.18.558295},
  howpublished = {bioRxiv},
}
Figure