Long-term morphometric similarity gradients relate to cortical hierarchy and psychiatric symptoms in schizophrenia

This longitudinal study of 193 healthy controls and 350 individuals with schizophrenia spectrum disorders reveals that long-term morphometric similarity gradients, which align with cortical hierarchy and disease epicenters, are significantly associated with treatment duration, medication, and psychiatric symptoms, thereby elucidating how brain organization and clinical interventions shape disease progression.

The Gist

Imagine the human brain not just as a collection of organs, but as a bustling, high-tech city. In this city, different neighborhoods (brain regions) have different jobs. Some are busy industrial zones handling basic tasks like seeing and moving (the "sensorimotor" areas), while others are high-rise corporate headquarters managing complex thoughts, planning, and social interactions (the "association" areas).

This paper is like a long-term urban planning study investigating what happens to the roads and connections between these neighborhoods in people with Schizophrenia Spectrum Disorders (SSD).

Here is the breakdown of the study using simple analogies:

1. The Tool: "MIND" (The City's Blueprint)

Usually, to see how neighborhoods are connected, scientists need expensive, complex maps (like Diffusion MRI). But this study used a clever new tool called MIND (Morphometric INverse Divergence).

Think of MIND as a "Style Matcher." Instead of looking at the physical roads, it looks at the architecture of the buildings in different neighborhoods. If the buildings in Neighborhood A and Neighborhood B look very similar (same height, same style, same layout), MIND says, "These two places are probably talking to each other a lot." It creates a map of structural similarity.

2. The City's Hierarchy: The "Gradient"

The researchers noticed that the brain isn't just a flat map; it has a hierarchy, like a ladder.

• Bottom of the ladder: Basic sensory areas (seeing, hearing, moving).
• Top of the ladder: Complex thinking areas (planning, self-awareness).

They used math to find "Gradients" (sliding scales) that show how the brain moves from the bottom to the top.

• Gradient 1 (G1): A path from the emotional/attention centers to the visual/parietal centers.
• Gradient 2 (G2): The main "Ladder" moving from Basic Senses $\to$ Complex Thinking.

3. What They Found: The "City" in Schizophrenia

The team followed 350 people with SSD and 193 healthy people for up to 20 years. Here is what they discovered:

A. The "Epicenter" of the Problem

In a healthy city, the high-rise corporate districts (complex thinking areas) are very distinct from the industrial zones.

• The Finding: In people with SSD, the "Style Matcher" (MIND) showed that these distinct neighborhoods started to look more similar to each other in a bad way. It's like the unique architecture of the corporate towers got blurred, making them look more like the factories.
• The Location: This "blurring" happened most in the evolutionarily newest parts of the brain (the human-specific high-rises). These are the areas that make us human, and they are the most vulnerable to the disease.

B. The Effect of Time and Medicine

This is the most interesting part. They didn't just take a snapshot; they watched the city change over 20 years.

• Time: As time went on, the connections in the healthy brain naturally changed. In the SSD brain, the connections in the complex areas tended to get weaker over time.
• Medication (Antipsychotics): Think of medication as construction crews trying to fix the roads.
  • The study found that taking medication was associated with strengthening the connections (increasing the "similarity" between regions).
  • However, there was a catch. If the medication dose was too high for too long, the "construction crew" might actually start to cause problems, reversing the benefits and making symptoms worse. It's like over-repairing a road until it becomes a traffic jam.

C. The Symptoms

Finally, they checked the "citizens' mood" (psychiatric symptoms).

• The Rule: When the brain's "Style Matching" (structural similarity) was better and the "Ladder" (gradients) was clearer, the citizens felt better.
• The Result: As the brain structure improved (or stayed stable) and medication was managed correctly, the psychiatric symptoms (like hallucinations or confusion) went down. But if the brain structure got too "blurred" or the medication dose was too high for too long, symptoms got worse.

The Big Picture Takeaway

This study tells us that Schizophrenia isn't just random chaos in the brain. It follows a specific blueprint:

1. It attacks the most complex, human-specific parts of the brain first.
2. It messes up the hierarchy, making complex areas lose their unique identity.
3. Medication can help rebuild these connections, but it needs to be the right amount for the right amount of time. Too little, and the city falls apart; too much for too long, and the city gets clogged.

In short: The brain is a complex city. Schizophrenia blurs the lines between the neighborhoods. Treatment is like a renovation project that can restore the city's structure, but it requires a skilled architect (the doctor) to get the dosage and timing just right to help the "citizens" (the patients) live better lives.

Technical Summary

1. Problem Statement

Schizophrenia spectrum disorders (SSD) are characterized by widespread structural brain alterations and dysconnectivity. While cross-sectional studies have identified volume and thickness reductions, the longitudinal evolution of these changes remains unclear. Key challenges include:

• Disentangling Causes: Difficulty separating disease progression from the effects of antipsychotic medication.
• Hierarchical Organization: Lack of understanding regarding how brain structural similarity relates to the cortical hierarchy (the organization from primary sensory to transmodal association areas) over time.
• Clinical Correlation: The need to link specific structural connectivity patterns to long-term psychiatric symptom trajectories.
• Methodological Gaps: Traditional diffusion MRI is time-consuming and not routine; there is a need for robust structural similarity metrics derived from standard T1-weighted MRI that can capture connectome-like properties.

2. Methodology

Study Cohort:

• Participants: 193 Healthy Controls (HC) and 350 individuals with SSD (including first-episode psychosis).
• Longitudinal Design: Follow-up assessments at 1, 3, 5, 10, 15, and 20 years (up to 20 years total).
• Data: 1,293 high-quality structural MRI scans (T1-weighted).

Image Processing & Feature Extraction:

• Parcellation: Cortical regions were parcellated using a symmetric, equally-sized sub-parcellation of the Desikan-Killiany atlas (318 regions). 14 subcortical structures were also analyzed.
• MIND (Morphometric INverse Divergence): A novel method was used to estimate structural similarity between regions based on the Kullback-Leibler divergence of multivariate MRI features.
  • Cortical features: Thickness, mean curvature, sulcal depth, surface area, gray matter volume.
  • Subcortical features: Surface area and radial distance.
• Network Construction: Subject-specific MIND networks (318x318 cortical, 14x14 subcortical) were generated.
• Gradient Decomposition: Principal gradients (G1 and G2) were derived from cortical MIND networks to map hierarchical organization (reducing high-dimensional connectivity to low-dimensional representations).

Statistical Modeling:

• Baseline Analysis: Multiple regression to assess differences between SSD and HC.
• Longitudinal Analysis: Linear Mixed Models (LMM) and Generalized Linear Mixed Models (GLMM) to evaluate the effects of:
  • Diagnosis (SSD vs. HC)
  • Time (treatment duration)
  • Medication (antipsychotic dose in chlorpromazine equivalents)
  • Interactions (Diagnosis $\times$ Time, Medication $\times$ Time).
• Co-localization Analysis: Spatial correlations (using spin tests to preserve spatial autocorrelation) between MIND features and:
  • Evolutionary expansion maps (Human vs. Macaque).
  • Functional gradients (Sensorimotor-to-Association axis).
  • Schizophrenia epicenters (regions most vulnerable to disease-related alterations).
• Symptom Correlation: Associations between MIND features and Brief Psychiatric Rating Scale (BPRS) scores.

3. Key Contributions

1. Longitudinal MIND Mapping: First study to characterize how cortical and subcortical morphometric similarity (MIND) evolves over a 20-year period in SSD.
2. Hierarchical Vulnerability: Demonstrated that structural alterations in SSD are not random but are co-localized with the brain's evolutionary expansion and hierarchical organization (specifically affecting higher-order association areas).
3. Medication vs. Disease Effects: Provided evidence that while disease progression generally reduces structural similarity, antipsychotic medication is associated with an increase in structural similarity (degree), suggesting a potential restorative or stabilizing effect on brain architecture.
4. Gradient Compression: Identified a "compression" of the sensorimotor-to-association (S-A) gradient in SSD, indicating a loss of differentiation between primary and transmodal regions.
5. Clinical Relevance: Linked specific structural connectivity patterns (degree and gradients) to the trajectory of psychiatric symptoms, showing that symptom improvement correlates with increased structural similarity.

4. Key Results

Baseline Alterations:

• Degree (Connectivity): SSD patients showed significantly reduced MIND degree in the left superior temporal gyrus, nucleus accumbens, and pallidum. Conversely, the lingual area showed increased degree.
• Gradients: SSD showed increased G1 in the lingual/cuneus regions and reduced G1 in the middle temporal area.
• Co-localization: Baseline reductions in degree and gradients were negatively correlated with evolutionary expansion and the S-A ranking, meaning the most affected areas are those with the highest evolutionary expansion and highest hierarchical status (transmodal/DMN regions).

Longitudinal Effects (Time & Medication):

• Time: In HC, connectivity generally decreased over time in frontal/parietal/temporal regions. In SSD, this decline was attenuated (interaction effect), suggesting a different trajectory.
• Medication: Antipsychotic use was positively associated with MIND degree, suggesting medication increases inter-regional structural similarity.
• Gradient Evolution: Over time, G2 (S-A axis) decreased in frontal/cingulate/temporal regions but increased in parietal/occipital regions, reflecting a compression of the hierarchy. Medication partially reversed these time-dependent gradient changes.
• Epicenters: Regions identified as "epicenters" of schizophrenia pathology showed the strongest increases in connectivity over time in SSD, particularly in functional epicenters.

Symptom Associations:

• Baseline: Lower MIND degree in frontal/temporal/cingulate regions correlated with higher symptom severity (BPRS).
• Longitudinal: Symptom severity decreased over time. This improvement was associated with increased structural similarity (degree) and specific gradient changes.
• Medication Dose: While higher doses initially reduced symptoms, long-term exposure to high doses was associated with more severe symptoms, potentially due to neurocognitive side effects or tolerance.

5. Significance

• Mechanistic Insight: The study supports the hypothesis that schizophrenia preferentially targets evolutionarily recent, high-order association networks, disrupting the brain's hierarchical organization.
• Biomarker Potential: MIND-derived gradients and degree offer potential biomarkers for tracking disease progression and treatment response over decades, utilizing standard clinical MRI.
• Treatment Implications: The finding that medication increases structural similarity suggests antipsychotics may have a neuroprotective or restorative effect on brain architecture, though high long-term doses may have diminishing returns on symptom control.
• Clinical Translation: By linking structural gradients to symptom trajectories, the study advances the understanding of how brain organization shapes clinical outcomes, potentially guiding personalized treatment strategies based on an individual's specific connectivity profile.

Limitations Noted: High attrition rates over 20 years, inability to include an unmedicated control group (ethical constraints), and the confounding nature of medication dosage (often higher in sicker patients).