Voxel2Mesh: 3D Mesh Model Generation from Volumetric Data

TL;DR

This paper introduces Voxel2Mesh, a novel deep learning architecture that directly converts 3D volumetric data into surface meshes, eliminating post-processing artifacts and improving accuracy over existing segmentation methods.

Contribution

The paper presents a new end-to-end model that generates 3D surface meshes directly from volumetric data, bypassing traditional post-processing steps.

Findings

Outperforms current state-of-the-art segmentation methods.

Effective on Electron Microscopy, MRI, and CT datasets.

Produces high-quality 3D surface meshes without post-processing.

Abstract

CNN-based volumetric methods that label individual voxels now dominate the field of biomedical segmentation. However, 3D surface representations are often required for proper analysis. They can be obtained by post-processing the labeled volumes which typically introduces artifacts and prevents end-to-end training. In this paper, we therefore introduce a novel architecture that goes directly from 3D image volumes to 3D surfaces without post-processing and with better accuracy than current methods. We evaluate it on Electron Microscopy and MRI brain images as well as CT liver scans. We will show that it outperforms state-of-the-art segmentation methods.