Altered brain criticality in Schizophrenia: New insights from MEG

Golnoush Alamian, Tarek Lajnef, Annalisa Pascarella, Jean‐Marc Lina, Laura Knight, James Walters, Krish D. Singh, Karim JerbibioRxiv 2021 · DOI

TL;DR

This study investigates brain dynamics in schizophrenia using magnetoencephalography (MEG) and the concept of "criticality," a state of optimal brain function. They found that schizophrenia patients exhibit altered self-similarity and multifractality in their brain signals, suggesting a less flexible and more rigid neural activity pattern. These alterations were significant enough to allow for accurate classification of patients versus controls using machine learning.

What Problem Does This Paper Address?

Schizophrenia is a complex disorder with a wide range of symptoms and a poorly understood underlying neurobiology. Despite decades of research, a definitive biomarker remains elusive. A key challenge is understanding how the brain's non-linear dynamics – the complex, fluctuating patterns of neural activity – are altered in schizophrenia. The authors propose that the brain, when functioning optimally, operates in a state of "criticality," balancing order and disorder to maximize efficiency, adaptability, and information processing. They hypothesize that disruptions to this critical state could underlie some of the cognitive deficits seen in schizophrenia. This paper aims to characterize these disruptions by examining specific measures of criticality in the resting-state brain activity of schizophrenia patients and healthy controls.

Methods

The core of this study lies in analyzing resting-state MEG data using multifractal analysis. Specifically, they employed the Wavelet p-Leader and Bootstrap based MultiFractal (WLBMF) analysis to quantify two key parameters:

  • C1 (Self-similarity): Reflects how similar a signal is across different scales, indicating memory and regularity. Higher C1 suggests more self-similarity.
  • C2 (Multifractality): Captures the diversity of local fluctuations or "singularities" within a signal, indicating complexity and flexibility. Lower (more negative) C2 suggests higher multifractality.

They also utilized logistic regression with 10-fold cross-validation for machine learning classification to differentiate between patients and controls based on these C1 and C2 values. Standard statistical tests (unpaired t-tests, Pearson correlations) were also used to compare group averages and explore relationships with clinical data.

Datasets

The study used resting-state MEG data from:

  • 25 chronic schizophrenia (SZ) patients
  • 25 age- and sex-matched healthy controls

Demographic and clinical data, including depression scores (BDI-II), mania scores (ASRM), positive and negative symptom scores (SAPS, SANS), and antipsychotic medication dosage, were also collected for the SZ group.

Key Insights and Ideas

  • Opposing Gradients: Healthy brains show a clear rostral-to-caudal gradient for both C1 (increasing from front to back) and C2 (decreasing from front to back). This suggests a structured organization of self-similarity and complexity across the brain.
  • Attenuated Gradients in SZ: While schizophrenia patients also exhibit these gradients, they are less pronounced, indicating a disruption in this organized pattern.
  • Increased Regularity in SZ Frontal/Temporal Regions: Patients showed higher C1 values in fronto-temporal areas, suggesting increased self-similarity and thus more regularity and "memory" in neural activity.
  • Decreased Complexity in SZ Parietal/Occipital Regions: Conversely, patients had less multifractality (higher, less negative C2) in parietal and occipital regions, indicating reduced complexity and flexibility.
  • Machine Learning Success: The C1 and C2 measures were powerful enough to accurately classify SZ patients from controls with up to 77% decoding accuracy, highlighting their potential as biomarkers.
  • Positive C2 Anomaly: A subset of participants, particularly patients, showed positive C2 values, which are theoretically unexpected in multifractal analysis. This suggests that the standard multifractal formalism might not fully capture the complexity of their brain signals, potentially indicating a more rigid or less complex dynamic.

Key Results

  • Classification Accuracy: Machine learning models achieved statistically significant classification of SZ patients and controls with up to 77% decoding accuracy using C1 values and 76% using C2 values.
  • Regional Differences: Specific brain regions, such as the left superior frontal gyrus (for C1) and the right temporal gyrus (for C2), showed the highest classification accuracy.
  • Correlations with Symptoms:
    • Increased C1 values correlated with higher negative symptom scores (SANS) in the inferior frontal gyrus.
  • Increased C2 values showed a trend towards correlation with positive symptom scores (SAPS) in the insula.
  • C1 values also positively correlated with medication dosage, particularly in frontal and temporal regions.

Limitations and Future Work

  • Chronic and Medicated Patients: The study focused on chronic, medicated schizophrenia patients. This limits the generalizability of findings to drug-naïve or first-episode patients, as medication can influence brain dynamics.
  • Age Differences: The authors acknowledge potential age-related differences in scale-free dynamics, noting that their participants were older (mean age 44) than in a similar study by La Rocca et al. (2018) (mean age 22), which might explain differing self-similarity gradient findings.
  • Positive C2 Interpretation: The presence of positive C2 values in some participants, especially patients, suggests limitations of the current multifractal formalism for these specific brain signals. Further exploration with alternative p-leader values (e.g., p=4) showed improved multifractality modeling and reaffirmed diminished multifractality in patients.
  • Future Directions: The authors suggest that these findings provide a new avenue for searching for biomarkers in schizophrenia and could help understand the temporal dynamics of information processing disruptions in the disorder. Further research is needed to explore these criticality alterations in different patient populations and across various cognitive tasks.

Summary updated

Abstract

Abstract Schizophrenia has a complex etiology and symptomatology that is difficult to untangle. After decades of research, important advancements towards a central biomarker are still lacking. One of the missing pieces is a better understanding of how non-linear neural dynamics are altered in this patient population. In this study, the resting-state neuromagnetic signals of schizophrenia patients and healthy controls were analyzed in the framework of criticality. When biological systems like the brain are in a state of criticality, they are thought to be functioning at maximum efficiency (e.g., optimal communication and storage of information) and with maximum adaptability to incoming information. Here, we assessed the self-similarity and multifractality of resting-state brain signals recorded with magnetoencephalography in patients with schizophrenia patients and in matched controls. Our analysis showed a clear ascending, rostral to caudal gradient of self-similarity values in healthy controls, and an opposite gradient for multifractality (descending values, rostral to caudal). Schizophrenia patients had similar, although attenuated, gradients of self-similarity and multifractality values. Statistical tests showed that patients had higher values of self-similarity than controls in fronto-temporal regions, indicative of more regularity and memory in the signal. In contrast, patients had less multifractality than controls in the parietal and occipital regions, indicative of less diverse singularities and reduced variability in the signal. In addition, supervised machine-learning, based on logistic regression, successfully discriminated the two groups using measures of self-similarity and multifractality as features. Our results provide new insights into the baseline cognitive functioning of schizophrenia patients by identifying key alterations of criticality properties in their resting-state brain data.
BibTeX
@misc{alamian2021altered,
  author = {Golnoush Alamian and Tarek Lajnef and Annalisa Pascarella and Jean‐Marc Lina and Laura Knight and James Walters and Krish D. Singh and Karim Jerbi},
  title = {{Altered brain criticality in Schizophrenia: New insights from MEG}},
  year = {2021},
  doi = {10.1101/2021.11.09.467906},
  howpublished = {bioRxiv},
}

Loading related content...

Figure