Cognitive Implications of Correlated Structural Network Changes in Schizophrenia

Dawn Jensen, Elaheh Zendrehrouh, Vince D. Calhoun, Jessica A. TurnerFrontiers in Integrative Neuroscience 2022 · DOI

TL;DR

This study uses a multimodal neuroimaging approach (pICA) to investigate how correlated changes in gray matter volume and white matter integrity are linked to cognitive deficits in schizophrenia. They found three distinct structural networks showing significant differences between healthy controls and patients, with two of these networks directly relating to cognitive performance.

What Problem Does This Paper Address?

Schizophrenia is characterized by widespread structural brain changes (in gray matter volume and white matter integrity) and significant cognitive impairments. While previous research has examined these aspects separately, understanding how the interplay between gray and white matter structural changes contributes to cognitive deficits has been challenging. This paper aims to bridge this gap by using a multimodal fusion technique to explore these interconnected relationships and their implications for cognitive function in schizophrenia.

Methods

The researchers employed Parallel Independent Component Analysis (pICA), a data-driven technique that allows for the simultaneous analysis of multiple imaging modalities (in this case, structural MRI for gray matter volume and diffusion tensor imaging for white matter integrity, specifically fractional anisotropy or FA). pICA identifies independent components that represent spatial patterns of gray matter and white matter. By analyzing these components, the study could:

  1. Identify correlated patterns: Find pairs of gray matter and white matter components that covary across individuals.
  2. Detect group differences: Determine which of these correlated patterns showed significant differences between individuals with schizophrenia (SZ) and healthy controls (HC).
  3. Assess cognitive links: Use MANCOVA (Multivariate Analysis of Covariance) to examine how these identified structural network differences relate to global cognitive performance, while accounting for age and gender.

This approach is novel in its direct fusion of GM and FA data to identify intermodal covariance patterns, rather than analyzing them in isolation.

Datasets

The study utilized data from the COBRE (Center for Biomedical Research Excellence) dataset. This dataset included:

  • 73 subjects with schizophrenia (SZ)
  • 82 healthy controls (HC)

The data was collected on a Siemens 3T Trio TIM scanner.

Key Insights and Ideas

  • Interconnectedness of Brain Structure: The study strongly suggests that gray matter and white matter changes in schizophrenia are not isolated but rather occur in interconnected networks. The multimodal approach highlights this crucial link.
  • Structural Networks as Biomarkers: The identified structural networks, showing differences between SZ and HC, could serve as potential biomarkers for schizophrenia.
  • Cognition is Tied to Specific Structural Networks: Two of the identified structural networks, specifically those related to gray matter changes, were directly linked to differences in global cognitive performance. This implies that specific structural alterations have a more direct impact on cognitive function than others.
  • Frontoparietal Control Networks and DMN Involvement: The identified networks show similarities to known functional networks, including frontoparietal control networks (involved in cognitive control and flexibility) and subnetworks of the Default Mode Network (DMN) (associated with memory and metacognition). This suggests that disruptions in these fundamental brain systems underlie cognitive deficits in schizophrenia.

Key Results

  • Three Significant Component Pairs: The pICA analysis identified three pairs of correlated gray matter (GM) and white matter (FA) structural networks that showed significant differences between healthy controls (HC) and patients with schizophrenia (SZ). In all three pairs, HC showed greater GM and FA values than SZ.
  • Cognitive Links in Two Networks: Two of these three component pairs revealed that the gray matter patterns within these networks were significantly associated with global cognitive performance. Specifically, higher loading coefficients (indicating greater GM differences) in these networks were linked to poorer cognitive performance in patients with schizophrenia.
  • Broad Network Involvement: The identified structural networks are not localized to single brain regions but pull from a wide range of areas, including the thalamus, lateral occipital cortex, middle temporal gyrus, precuneus cortex, postcentral gyrus, cingulate gyrus/cingulum, lingual gyrus, and brain stem.
  • Global Cognitive Deficit: A PCA of the cognitive battery confirmed that individuals with schizophrenia performed significantly worse on cognitive tests compared to healthy controls.

Limitations and Future Work

  • Cross-Sectional Design: The study is cross-sectional, meaning it captures a single point in time. This limits the ability to establish causal relationships between structural changes and cognitive decline.
  • Summary Measure of Cognition: A global cognitive score (g) was used to reduce statistical tests. While justified by the generalized nature of cognitive impairment in schizophrenia, this approach may mask specific cognitive domain deficits.
  • Medication Effects: All patients were taking antipsychotic medication, which could potentially influence brain structure and cognitive function.
  • Future Directions:
    • Tractography Studies: The authors suggest tractography could provide more detailed insights into the connectivity within these identified structural networks.
  • Longitudinal Studies: A longitudinal approach, tracking individuals from the prodromal stage through chronic illness, could better elucidate the causal relationships between structural network changes and cognitive decline.
  • Specific Cognitive Domains: Future research could explore the relationships between these structural networks and specific cognitive domains, rather than a global measure.

Summary updated

Abstract

Background Schizophrenia is a brain disorder characterized by diffuse, diverse, and wide-spread changes in gray matter volume (GM) and white matter structure (fractional anisotropy, FA), as well as cognitive impairments that greatly impact an individual’s quality of life. While the relationship of each of these image modalities and their links to schizophrenia status and cognitive impairment has been investigated separately, a multimodal fusion via parallel independent component analysis (pICA) affords the opportunity to explore the relationships between the changes in GM and FA, and the implications these network changes have on cognitive performance. Methods Images from 73 subjects with schizophrenia (SZ) and 82 healthy controls (HC) were drawn from an existing dataset. We investigated 12 components from each feature (FA and GM). Loading coefficients from the images were used to identify pairs of features that were significantly correlated and showed significant group differences between HC and SZ. MANCOVA analysis uncovered the relationships the identified spatial maps had with age, gender, and a global cognitive performance score. Results Three component pairs showed significant group differences (HC > SZ) in both gray and white matter measurements. Two of the component pairs identified networks of gray matter that drove significant relationships with cognition (HC > SZ) after accounting for age and gender. The gray and white matter structural networks identified in these three component pairs pull broadly from many regions, including the right and left thalamus, lateral occipital cortex, multiple regions of the middle temporal gyrus, precuneus cortex, postcentral gyrus, cingulate gyrus/cingulum, lingual gyrus, and brain stem. Conclusion The results of this multimodal analysis adds to our understanding of how the relationship between GM, FA, and cognition differs between HC and SZ by highlighting the correlated intermodal covariance of these structural networks and their differential relationships with cognitive performance. Previous unimodal research has found similar areas of GM and FA differences between these groups, and the cognitive deficits associated with SZ have been well documented. This study allowed us to evaluate the intercorrelated covariance of these structural networks and how these networks are involved the differences in cognitive performance between HC and SZ.
BibTeX
@inproceedings{jensen2022cognitive,
  author = {Dawn Jensen and Elaheh Zendrehrouh and Vince D. Calhoun and Jessica A. Turner},
  title = {{Cognitive Implications of Correlated Structural Network Changes in Schizophrenia}},
  booktitle = {Frontiers in Integrative Neuroscience},
  year = {2022},
  doi = {10.3389/fnint.2021.755069},
}
Figure