mlsanityData to 7 Oct 2026

Augmented Equivariant Mesh Networks for Anatomical Segmentation

arXiv 2605.08172 ↗First posted 4 May 2026cs.CV cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Anatomical mesh segmentation requires models that operate directly on irregular surface geometry while remaining robust to changes in coordinate pose across meshes of varying resolution. Existing task-specific mesh and point-cloud methods are not equivariant, and can degrade sharply under test-time perturbation, for example dropping by 28-30 IoU points on 3D-IOSSeg at 40^circ rotation even with matched training augmentation. We present EAMS, an Equivariant Anatomical Mesh Segmentor built on Equivariant Mesh Neural Networks (EMNN), and evaluate it in four dataset settings across three clinical application areas, spanning edge-, vertex-, and face-level supervision. We combine intrinsic mesh descriptors with anatomy-aware priors, including PCA-derived frames for dental arches and liver surfaces, and augment message passing to provide lightweight global context. Across intracranial aneurysm and intraoral segmentation, EAMS variants are competitive with specialized baselines on unperturbed inputs while remaining stable under geometric perturbations, and on liver surfaces they expose a favorable trade-off between canonical-pose accuracy and rotation robustness. These results show that a lightweight (<2M parameters) equivariant framework can deliver robust anatomical mesh segmentation across diverse label types.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 19 of its 27 references that are in the map. First posted on arXiv 4 May 2026.

Builds on

19 of 27 references.

References · 19

  1. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,384
  2. 2018Dynamic Graph CNN for Learning on Point Cloudscited by 2,220
  3. 2021E(n) Equivariant Graph Neural Networkscited by 951
  4. 2018PointConv: Deep Convolutional Networks on 3D Point Cloudscited by 656
  5. 2023Point Transformer V3: Simpler, Faster, Strongercited by 579
  6. 2018MeshNet: Mesh Neural Network for 3D Shape Representationcited by 95
  7. 2020Gauge Equivariant Mesh CNNs: Anisotropic convolutions on geometric graphscited by 90
  8. 2019SpiralNet++: A Fast and Highly Efficient Mesh Convolution Operatorcited by 89
  9. 2022Contrastive Boundary Learning for Point Cloud Segmentationcited by 56
  10. 2020IntrA: 3D Intracranial Aneurysm Dataset for Deep Learningcited by 41
  11. 2018A Simple Approach to Intrinsic Correspondence Learning on Unstructured 3D Meshescited by 39
  12. 2023E(n) Equivariant Message Passing Simplicial Networkscited by 23
7 more
  1. 20233DTeethSeg'22: 3D Teeth Scan Segmentation and Labeling Challengecited by 14
  2. 2023Boundary Difference Over Union Loss For Medical Image Segmentationcited by 13
  3. 2022Equivariant Mesh Attention Networkscited by 12
  4. 2023Using Multiple Vector Channels Improves E(n)-Equivariant Graph Neural Networkscited by 10
  5. 2024E(3)-Equivariant Mesh Neural Networkscited by 9
  6. 20253D Dental Model Segmentation with Geometrical Boundary Preservingcited by 7
  7. 2025Fast and Distributed Equivariant Graph Neural Networks by Virtual Node Learningcited by 6

Built on by

Not cited yet.