Reproduce “Sparse cubical complexes for efficient topology-preservation in image data” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Alexander H. Berger, Marco Fontana, Daniel Rueckert, Johannes C. Paetzold, Laurin Lux, Ulrich Bauer

SparseBM makes persistent-homology losses more practical for large-patch 3D segmentation by retaining cells relevant to prediction-label comparison and representing omitted background implicitly. On ATM’26 airway segmentation with 128³ patches, it reduced Betti-Matching error from 102 to 18.8 while Dice remained 0.9446. Whole-training time was 1.15× the Dice+CE baseline. This supports training-time topology-aware learning, not correct topology at inference.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.