DART: Scaling a Shape Representation Across Anatomical Classes and Imaging Modalities

Key points

• Current state-of-the-art segmentation models still produce anatomically implausible shapes, such as broken vessels and fragmented glands, on the structures that are hardest to segment.1
• Learned shape representations capture what structures look like and can help correct these errors, but most cover a fixed set of anatomical classes.1
• Using existing segmentation predictions, we scale a single shape representation from 50 to 377 anatomical classes without new voxel-level annotations for correction training.1
• These results suggest that sharing shape information within and across classes can reduce the need for new voxel-level annotations when extending segmentation to more anatomical classes and imaging modalities.1

Sources (1)

• [1]DART: Scaling a Shape Representation Across Anatomical Classes and Imaging Modalities
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 04:12 AM

  We propose Diffusion for Anatomical Reshaping with Text (DART), which learns a shared shape representation from segmentation probabilities.

  Current state-of-the-art segmentation models still produce anatomically implausible shapes, such as broken vessels and fragmented glands, on the structures that are hardest to segment.

Extractive summary: sentences quoted from the sources.