Hierarchical Mesh Transformers with Topology-Guided Pretraining for Morphometric Analysis of Brain Structures

2026-04-06

AI Summary

Existing methods struggle to effectively integrate multimodal vertex-level morphometric features on heterogeneous brain meshes and are constrained by fixed topologies, limiting their generalizability. This work proposes a hierarchical Transformer framework that constructs spatially adaptive tree partitions based on arbitrary-order simplicial complexes, enabling unified processing of both volumetric and surface meshes. A feature projection module decouples geometric structure from clinical attributes, while a self-supervised pretraining strategy employing multi-channel masked reconstruction facilitates cross-task transfer. The approach is the first to support unified modeling of brain meshes with arbitrary topology and achieves state-of-the-art performance in Alzheimer’s disease classification and amyloid burden prediction on the ADNI dataset, as well as focal cortical dysplasia detection on the MELD dataset.

Abstract

Representation learning on large-scale unstructured volumetric and surface meshes poses significant challenges in neuroimaging, especially when models must incorporate diverse vertex-level morphometric descriptors, such as cortical thickness, curvature, sulcal depth, and myelin content, which carry subtle disease-related signals. Current approaches either ignore these clinically informative features or support only a single mesh topology, restricting their use across imaging pipelines. We introduce a hierarchical transformer framework designed for heterogeneous mesh analysis that operates on spatially adaptive tree partitions constructed from simplicial complexes of arbitrary order. This design accommodates both volumetric and surface discretizations within a single architecture, enabling efficient multi-scale attention without topology-specific modifications. A feature projection module maps variable-length per-vertex clinical descriptors into the spatial hierarchy, separating geometric structure from feature dimensionality and allowing seamless integration of different neuroimaging feature sets. Self-supervised pretraining via masked reconstruction of both coordinates and morphometric channels on large unlabeled cohorts yields a transferable encoder backbone applicable to diverse downstream tasks and mesh modalities. We validate our approach on Alzheimer's disease classification and amyloid burden prediction using volumetric brain meshes from ADNI, as well as focal cortical dysplasia detection on cortical surface meshes from the MELD dataset, achieving state-of-the-art results across all benchmarks.

Problem

• morphometric analysis
• mesh representation learning
• neuroimaging
• heterogeneous mesh topology
• vertex-level descriptors

Innovation

• Hierarchical Mesh Transformers
• Topology-Guided Pretraining
• Morphometric Analysis
• Simplicial Complexes
• Self-Supervised Learning