Multi-spatial-scale dynamic interactions between functional sources reveal sex-specific changes in schizophrenia

Armin Iraji, Ashkan Faghiri, Zening Fu, Srinivas Rachakonda, Peter Kochunov, Ayşenil Belger, Judy M. Ford, Sarah McEwen, Daniel H. Mathalon, Bryon A. Mueller, Godfrey D. Pearlson, Steven G. Potkin, Adrian Preda, Jessica A. Turner, Theodorus G.m. van Erp, Vince D. CalhounNetwork 2021 · DOI

TL;DR

This paper introduces "multiscale ICA" (msICA), a novel method to analyze brain connectivity at multiple spatial scales simultaneously. Using msICA, they discovered significant sex-specific differences in dynamic brain network interactions in individuals with schizophrenia, suggesting that these multiscale patterns could be crucial for understanding the disorder and developing sex-specific biomarkers.

What Problem Does This Paper Address?

Understanding how different brain regions communicate (functional connectivity) is key to understanding brain function and dysfunction. Traditional methods often analyze connectivity at a single "spatial scale" (e.g., looking at large brain networks or small, localized regions) or use fixed anatomical regions. This approach can miss crucial information because brain networks exist and interact across multiple scales, and their organization isn't uniform. Furthermore, brain activity is dynamic, changing over time, and these dynamic interactions also vary across spatial scales. Existing methods struggle to capture this multi-scale dynamic interplay, especially in complex disorders like schizophrenia, where sex differences in presentation and progression are known but poorly understood at the neural level. The authors aim to bridge this gap by developing a method that can simultaneously analyze dynamic functional connectivity across multiple spatial scales and apply it to investigate sex-specific alterations in schizophrenia.

Methods

The core innovation is multiscale ICA (msICA). Standard ICA decomposes brain signals into independent components (ICNs) representing functional sources, but typically at a single "model order" (which influences the scale/granularity of the components). msICA addresses this by running ICA with multiple model orders simultaneously. This allows the estimation of functional sources at various spatial scales without pre-defining their size or number.

To capture the dynamic nature of these interactions, the authors employed dynamic functional network connectivity (dFNC). This involves analyzing short segments of brain activity over time to see how connectivity patterns change. They combined msICA with dFNC to analyze both static (overall) and dynamic functional interactions within and between these multiscale functional sources.

For statistical analysis, they used a general linear model (GLM) to compare groups (controls vs. schizophrenia, males vs. females) while accounting for covariates like age and acquisition site. They also employed a robust permutation testing approach to assess sex-specific differences in how schizophrenia affects connectivity, comparing the magnitude of these effects between males and females.

Datasets

The study utilized resting-state fMRI (rsfMRI) data from 827 individuals, comprising 477 typical controls and 350 individuals with schizophrenia. This data was pooled from three large projects:

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

The paper does not introduce new benchmarks but rather applies its novel methodology to existing, well-characterized datasets.

Key Insights and Ideas

  • Brain connectivity is inherently multiscale: The study strongly supports the idea that brain function isn't just about large networks or small regions, but the interplay between these different levels of organization. msICA provides a powerful way to visualize this.
  • Schizophrenia shows sex-specific disruptions across scales: The most significant differences in brain connectivity related to schizophrenia were not uniform but varied depending on sex and the spatial scale of analysis. This highlights the importance of considering both factors.
  • Dynamic states are key to sex differences: Many sex-specific differences in schizophrenia were most prominent in specific dynamic states of brain activity (dFNC states), particularly in State 4, which involves strong interactions between visual, somatomotor, and temporal domains.
  • Multiscale analysis is crucial for biomarker discovery: The correlation between multiscale connectivity patterns and symptom severity, especially in males, suggests that these complex, multiscale interactions could be valuable for identifying biomarkers for schizophrenia.

Key Results

  • msICA effectively segregates brain networks: The method successfully identified functional sources (ICNs) at multiple spatial scales, showing how brain regions can be organized differently depending on the chosen resolution.
  • Sex-specific differences in schizophrenia are widespread: Significant differences in functional connectivity were observed between male and female individuals with schizophrenia, particularly in between-scale interactions. These differences were most pronounced in dynamic states.
  • Specific domains show sex-specific alterations: Sex-specific changes were noted within the subcortical, somatomotor, cerebellum, temporal, and visual domains, often in dynamic states.
  • Opposite patterns observed: In some instances, the study found opposite patterns of connectivity alteration in males versus females with schizophrenia, suggesting distinct underlying neural mechanisms.
  • Correlations with symptoms: Multiscale connectivity features showed significant correlations with symptom scores in males, but not females, further emphasizing sex-specific relationships.

Limitations and Future Work

  • Unbalanced sample sizes: The study acknowledges the unbalanced number of male and female participants and suggests future studies should aim for larger, more balanced datasets.
  • Medication and treatment effects: The authors note that long-term effects of medication and treatment could not be fully accounted for, which might influence observed differences.
  • Age of onset: Future longitudinal studies are recommended to investigate the role of age of onset in sex-specific differences.
  • Further exploration of spatial scales: The authors suggest increasing the range of model orders in future msICA studies to capture a broader spectrum of spatial scales.
  • Dynamic functional segregation: Future work should also explore how functional sources themselves change dynamically over time.
  • Multi-temporal scale analysis: Extending the multiscale approach to include temporal dynamics and frequencies could provide a more comprehensive understanding.
  • Clinical validation: While promising, these findings require further validation and investigation for their potential as diagnostic or predictive biomarkers.

Summary updated

Abstract

We introduce an extension of independent component analysis (ICA), called multiscale ICA, and design an approach to capture dynamic functional source interactions within and between multiple spatial scales. Multiscale ICA estimates functional sources at multiple spatial scales without imposing direct constraints on the size of functional sources, overcomes the limitation of using fixed anatomical locations, and eliminates the need for model-order selection in ICA analysis. We leveraged this approach to study sex-specific and sex-common connectivity patterns in schizophrenia. Results show dynamic reconfiguration and interaction within and between multi-spatial scales. Sex-specific differences occur (a) within the subcortical domain, (b) between the somatomotor and cerebellum domains, and (c) between the temporal domain and several others, including the subcortical, visual, and default mode domains. Most of the sex-specific differences belong to between-spatial-scale functional interactions and are associated with a dynamic state with strong functional interactions between the visual, somatomotor, and temporal domains and their anticorrelation patterns with the rest of the brain. We observed significant correlations between multi-spatial-scale functional interactions and symptom scores, highlighting the importance of multiscale analyses to identify potential biomarkers for schizophrenia. As such, we recommend such analyses as an important option for future functional connectivity studies.
BibTeX
@inproceedings{iraji2021multi,
  author = {Armin Iraji and Ashkan Faghiri and Zening Fu and Srinivas Rachakonda and Peter Kochunov and Ayşenil Belger and Judy M. Ford and Sarah McEwen and Daniel H. Mathalon and Bryon A. Mueller and Godfrey D. Pearlson and Steven G. Potkin and Adrian Preda and Jessica A. Turner and Theodorus G.M. van Erp and Vince D. Calhoun},
  title = {{Multi-spatial-scale dynamic interactions between functional sources reveal sex-specific changes in schizophrenia}},
  booktitle = {Network},
  year = {2021},
  doi = {10.1162/netn\_a\_00196},
}
Figure