Reproduce “TRiST: Topology-Reinforced Swin Transformer for Medical Image Analysis” 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
Pengfei Gu, Huimin Li, Guangyu Meng, Hao Zheng, Haoteng Tang, Bin Fu, Danny Z. Chen
TRiST adds multiscale persistent-homology descriptors to a Swin-based medical-image model and uses them to bias window attention. It computes connected-component and loop features at four stages, refining them by persistence band. In benchmark experiments, it improves over Swin-based baselines; on Synapse, its full model reaches 82.19 Dice versus 76.91 for downsampled topology, while stronger 3D methods score higher.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.