subcortical-shape-cognition-aging - SKILL.md Agent Skill

name: subcortical-shape-cognition-aging description: "Subcortical shape variations and their associations with cognition across the 8th decade of life. Longitudinal study using neuroimaging and cognitive data from Lothian Birth Cohort 1936. Analyzes heterogeneous morphological trajectories in hippocampus, thalami, globus pallidi, and ventral DC. Uses ANCOVA and mixed linear model analyses to investigate vertex displacement patterns associated with cognitive aging. Use when studying brain morphology changes, subcortical shape analysis, cognitive aging, longitudinal neuroimaging, or vertex-based morphometry." license: Complete terms in LICENSE.txt metadata: arxiv_id: "2605.29703" published: "2026-05-28" authors: "Maria del C. Valdes-Hernandez, Wonjung Park, Joanna Moodie, Susana Muñoz Maniega, Janie Corley, Fraser N. Sneden, Mark E. Bastin, Joanna M. Wardlaw, Simon R. Cox, Jinah Park" tags: [subcortical, morphology, shape-analysis, cognitive-aging, longitudinal, vertex-displacement, hippocampus, thalamus, neuroimaging]

Subcortical Shape Variations and Cognition Across Aging

Methodology from "Subcortical Shape Variations and Their Associations with Cognition Across the 8th Decade of Life. A Study in the Lothian Birth Cohort 1936" (arXiv:2605.29703).

Overview

This methodology provides longitudinal shape analysis of subcortical brain structures across aging, revealing:

Heterogeneous atrophy patterns: Different subcortical structures show distinct morphological deformation trajectories
Left-right asymmetry: Hippocampus and ventral DC show hemisphere-specific patterns
Cognition-morphology association: Vertex displacements correlate with cognitive changes

Key insight: Shape changes capture functionally-relevant brain aging beyond gross volumetry.

Study Design

Lothian Birth Cohort 1936

Unique longitudinal dataset:

Community-dwelling individuals (age ~70-80 across study)
Multiple neuroimaging timepoints
Detailed cognitive assessments
Demographic and lifestyle data

Advantages:

Same birth year → Controls for developmental cohort effects
High retention rate → Longitudinal tracking feasible
Rich phenotyping → Multivariate analysis possible

Analysis Approach

Two-stage methodology:

Shape quantification: Vertex-based morphometry of subcortical structures
Association testing: ANCOVA and mixed linear models linking shape changes to cognition

Core Methodology

Vertex-Based Morphometry

Shape representation:

Surface meshes for each subcortical structure
Vertex coordinates define shape
Vertex displacement: Movement of surface points from baseline

Key structures analyzed:

Hippocampus (memory encoding)
Thalamus (sensory relay, consciousness)
Globus pallidus (motor control)
Ventral DC (dorsal striatum, motor/cognitive)

Measurement Pipeline

Workflow:

# Conceptual framework for shape analysis

def longitudinal_shape_analysis(scan_baseline, scan_followup, structure_name):
    """
    Analyze shape changes in subcortical structure across timepoints
    """
    # 1. Segmentation
    structure_baseline = segment_subcortical(scan_baseline, structure_name)
    structure_followup = segment_subcortical(scan_followup, structure_name)
    
    # 2. Surface extraction
    mesh_baseline = extract_surface(structure_baseline)
    mesh_followup = extract_surface(structure_followup)
    
    # 3. Vertex correspondence
    # Match vertices across timepoints (same anatomical location)
    matched_vertices = establish_correspondence(mesh_baseline, mesh_followup)
    
    # 4. Displacement calculation
    # Vector from baseline to followup position
    displacements = calculate_vertex_displacement(matched_vertices)
    
    # 5. Shape statistics
    # Magnitude, direction, regional patterns
    displacement_stats = analyze_displacement_patterns(displacements)
    
    return displacements, displacement_stats

Statistical Analysis

ANCOVA framework:

Shape_change ~ Cognitive_score + Age + Sex + Education + covariates

Mixed linear models:

Vertex_displacement ~ Time + Cognitive_trajectory + Random(Subject)

Key comparisons:

Between-subject: Individual differences in shape trajectories
Within-subject: Time-dependent shape changes
Structure-specific: Different patterns for hippocampus vs thalamus vs globus pallidus

Key Findings

Heterogeneous Atrophy Patterns

Structure-specific deformations:

Hippocampus and ventral DC:
- Varied morphological deformations
- Left-right asymmetry: Different patterns in each hemisphere
- Region-specific inward/outward movements
Thalami and globus pallidi:
- More uniform volume contraction
- Symmetrical: Nearly identical patterns in left and right
- Global shrinkage rather than focal deformation

Implication: Not all subcortical structures age uniformly—shape analysis reveals hidden heterogeneity.

Cognitive Associations

Vertex-cognition links:

General cognition: Associated with inward/outward vertex displacements between timepoints
Direction matters: Expansion vs contraction linked to different cognitive domains
Regional specificity: Certain vertex clusters show stronger cognitive correlations

Interpretation:

Shape changes capture functionally-relevant aging
Morphological trajectories may predict cognitive decline
Targeted analysis of deformation patterns improves understanding beyond volume alone

Implementation Guide

Step 1: Image Processing

Required tools:

Segmentation: FSL FIRST, FreeSurfer, or manual delineation
Surface extraction: Mesh generation from binary masks
Correspondence: Point-set registration algorithms

Quality control:

Visual inspection of segmentation accuracy
Consistency check across timepoints (same structure labeled)
Outlier detection for extreme displacements

Step 2: Shape Quantification

Vertex displacement calculation:

import numpy as np

def calculate_displacement(mesh_baseline, mesh_followup):
    """
    Calculate vertex displacement vectors
    """
    n_vertices = len(mesh_baseline.vertices)
    
    # Initialize displacement array
    displacements = np.zeros((n_vertices, 3))  # x, y, z components
    
    for i in range(n_vertices):
        # Baseline position
        p_baseline = mesh_baseline.vertices[i]
        
        # Followup position (same anatomical location)
        p_followup = mesh_followup.vertices[i]  # Assumes correspondence
        
        # Displacement vector
        displacements[i] = p_followup - p_baseline
    
    return displacements

def displacement_statistics(displacements):
    """
    Calculate shape change metrics
    """
    magnitudes = np.linalg.norm(displacements, axis=1)
    
    # Global metrics
    mean_displacement = np.mean(magnitudes)
    std_displacement = np.std(magnitudes)
    
    # Regional clustering (group nearby vertices)
    # Apply k-means or anatomical parcellation
    
    return mean_displacement, std_displacement

Step 3: Association Testing

Cognitive correlation:

from statsmodels.formula.api import mixedlm

def shape_cognition_association(displacements, cognitive_scores, covariates):
    """
    Test association between vertex displacement and cognition
    """
    # Magnitude of displacement
    displacement_magnitude = np.linalg.norm(displacements, axis=1)
    
    # Mixed linear model
    # Random intercept for subject (repeated measures)
    model = mixedlm(
        "displacement_magnitude ~ cognitive_score + age + sex",
        data=dataframe,
        groups=dataframe['subject_id']
    )
    
    result = model.fit()
    
    return result

Step 4: Visualization

Shape deformation maps:

def visualize_shape_changes(mesh, displacements):
    """
    Plot displacement vectors on surface mesh
    """
    # Color vertices by displacement magnitude
    magnitudes = np.linalg.norm(displacements, axis=1)
    
    # Plot surface with displacement arrows
    plot_mesh_with_vectors(mesh, displacements, color_by=magnitudes)
    
    # Show inwards (contraction) vs outwards (expansion)
    inward_mask = magnitudes < threshold
    outward_mask = magnitudes > threshold
    
    return visualization

Applications

Aging Research

Longitudinal tracking:

Monitor subcortical shape trajectories across decades
Detect early morphological changes before clinical symptoms
Identify individuals at risk for cognitive decline

Clinical translation:

Shape markers as early biomarkers for Alzheimer's, Parkinson's
Target regions for intervention (e.g., hippocampal preservation)

Neuroimaging Methodology

Beyond volumetry:

Gross volume measures miss regional heterogeneity
Shape analysis captures nuanced morphological changes
Vertex-based approach enables high-resolution spatial mapping

Other applications:

Developmental studies (brain shape maturation)
Disease progression tracking
Treatment response monitoring

Pitfalls

Methodological Challenges

Vertex correspondence: Establishing consistent vertex matching across timepoints is critical
- Solution: Use surface registration algorithms (SPHARM-PDM, spherical harmonics)
- Alternative: Statistical shape models (SSM) with built-in correspondence
Segmentation accuracy: Errors propagate to shape analysis
- Solution: Visual QC, multi-atlas segmentation, manual correction where needed
Multiple comparisons: Thousands of vertices → multiple testing burden
- Solution: Cluster-based correction, ROI-focused analysis, permutation testing
Confounding factors: Age, sex, education, vascular health affect shape
- Solution: Include as covariates in statistical models

Interpretation Issues

Direction meaning: Inward displacement ≠ atrophy necessarily (could be displacement from surrounding tissue)
- Solution: Consider global volume change alongside local displacement
Asymmetry significance: Left-right differences may reflect functional specialization or noise
- Solution: Test statistical significance of asymmetry metrics
Causality: Shape changes and cognition association is correlational
- Solution: Longitudinal design helps infer temporal precedence

Statistical Power

Sample size considerations:

Lothian Birth Cohort: N ~ 100s with multiple timepoints
Vertex-level analysis: Requires sufficient power for thousands of tests
Cluster-based approach: Reduces multiple testing penalty

Effect sizes:

Subcortical shape changes are subtle (mm-level displacements)
Cognitive associations moderate (partial correlations ~0.2-0.4)
Longitudinal tracking reveals cumulative effects

Cross-References

brain-mri-foundation-clinical: Clinical deployment of brain MRI foundation models
geosae-brain-mri-sae: Interpretable brain MRI foundation models
brain-segmentation-active-learning: Active learning for brain segmentation

References

Valdes-Hernandez, M. C., Park, W., Moodie, J., et al. (2026). Subcortical Shape Variations and Their Associations with Cognition Across the 8th Decade of Life. arXiv:2605.29703
FSL FIRST: https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST
FreeSurfer Subcortical Segmentation: https://surfer.nmr.mgh.harvard.edu/

Activation Keywords

subcortical shape
vertex morphometry
longitudinal neuroimaging
cognitive aging
hippocampus morphology
thalamus shape
brain aging
vertex displacement
shape trajectory