Augmented Equivariant Mesh Networks for Anatomical Segmentation

Daniel Saragih · 2026

Read Augmented Equivariant Mesh Networks for Anatomical Segmentation by Daniel Saragih (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-08172.txt.

Description

Daniel Saragih segments anatomical meshes with a model that respects geometry. Labeling points on organ surfaces must survive changes in pose. It must also survive changes in mesh resolution. Standard mesh and point-cloud models are not equivariant. They can lose thirty IoU points when a scan is rotated, even with matched training augmentation.

His EAMS segmentor builds on equivariant mesh neural networks. It mixes intrinsic surface descriptors with anatomy-aware priors. Those priors include PCA frames for dental arches and liver surfaces. Augmented message passing adds light global context. The system is tested in four dataset settings across three clinical areas. Supervision comes at edges, vertices, or faces.

Catalog record

CREATORDaniel Saragih
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE66 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.08172

Browse the full library · About the Tower of Babel