[Paper Review] Community-Level Modeling of Gyral Folding Patterns for Robust and Anatomically Informed Individualized Brain Mapping

Minheng Chen, Tong Chen+6 more|arXiv (Cornell University)|Feb 1, 2026

Functional Brain Connectivity Studies0 citations

TL;DR

The paper presents a spectral graph learning framework that models gyral folding at the community level (3HG folding communities) to achieve robust, anatomically grounded individualized brain mapping and cross-subject correspondence.

ABSTRACT

Cortical folding exhibits substantial inter-individual variability while preserving stable anatomical landmarks that enable fine-scale characterization of cortical organization. Among these, the three-hinge gyrus (3HG) serves as a key folding primitive, showing consistent topology yet meaningful variations in morphology, connectivity, and function. Existing landmark-based methods typically model each 3HG independently, ignoring that 3HGs form higher-order folding communities that capture mesoscale structure. This simplification weakens anatomical representation and makes one-to-one matching sensitive to positional variability and noise. We propose a spectral graph representation learning framework that models community-level folding units rather than isolated landmarks. Each 3HG is encoded using a dual-profile representation combining surface topology and structural connectivity. Subject-specific spectral clustering identifies coherent folding communities, followed by topological refinement to preserve anatomical continuity. For cross-subject correspondence, we introduce Joint Morphological-Geometric Matching, jointly optimizing geometric and morphometric similarity. Across over 1000 Human Connectome Project subjects, the resulting communities show reduced morphometric variance, stronger modular organization, improved hemispheric consistency, and superior alignment compared with atlas-based and landmark-based or embedding-based baselines. These findings demonstrate that community-level modeling provides a robust and anatomically grounded framework for individualized cortical characterization and reliable cross-subject correspondence.

Motivation & Objective

• Address substantial inter-subject variability in cortical folding while preserving stable anatomical landmarks.
• Move beyond isolated landmark matching by modeling higher-order folding communities (3HGs).
• Develop a subject-specific pipeline that yields cross-subject, anatomically grounded correspondences at the community level.

Proposed method

• Represent each three-hinge gyrus (3HG) with a dual-profile feature: topological context and structural (trace-map) connectivity.
• Cluster 3HGs within each subject’s hemisphere using a two-layer Graph Neural Network to obtain initial communities.
• Apply connectivity-constrained topological refinement to ensure spatial continuity of clusters.
• Establish cross-subject correspondence via Joint Morphological–Geometric Matching (JMGM) solved by a Hungarian assignment on multi-modal cluster features (morphology, geometry, and context).
• Train per-subject models (one for each hemisphere) to accommodate individual folding variability.

Experimental results

Research questions

• RQ1 Can community-level (rather than landmark-level) representations improve cross-subject alignment of gyral folding patterns?
• RQ2 Do 3HG folding communities show reduced morphometric variance and stronger modular organization across a large cohort?
• RQ3 Does joint morphological–geometric matching provide robust cross-subject correspondences that outperform atlas-based or isolated landmark approaches?

Key findings

• Folding communities exhibit substantially reduced morphometric variance across subjects.
• Community-level alignment yields stronger modular organization and improved cross-subject alignment compared with atlas-based and existing landmark- or embedding-based baselines.
• Hemispheric consistency and subject-specific folding representations are improved using the proposed pipeline.
• Cross-subject correspondences are established at the community level, enabling more reliable and fine-grained individualized brain mappings.
• The approach demonstrated robust performance across 1,064 HCP subjects, validating scalability and robustness.