Research map / Detection and segmentation / CNN features
HarmonySeg: Tubular Structure Segmentation with Deep-Shallow Feature Fusion and Growth-Suppression Balanced Loss
Yi Huang, Ke Zhang, Wei Liu, Yuanyuan Wang, Vishal M. Patel, Le Lu, Xu Han, Dakai Jin, Ke Yan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add D2SD deep-to-shallow decoding (ablation) | nnU-Net · HVS | Dice: 5.1% increase in Dice and a 2.6% reduction in HD | Improved |
| concatenate CT and vesselness maps as input (VSCat) (ablation) | nnU-Net · HVS | Dice: 60.15 -> 60.07 | No clear effect |
| add SADF with D2SD (ablation) | nnU-Net · HVS | clDice: 70.30 -> 73.08 | Improved |
| add GSB losses (ablation) | nnU-Net · HVS | Dice: 6.4% | Improved |
| add relaxed supervision loss Lr-sup alone (ablation) | nnU-Net · HVS | Dice: 2.94% | Improved |
| add connectivity loss Lcon alone (ablation) | nnU-Net · HVS | Dice: 1.11% | Improved |
| add spatial prior loss Lspatial alone (ablation) | nnU-Net · HVS | Dice: 1.67% | Improved |
| add mix-equivalence loss Lmix alone (ablation) | nnU-Net · HVS | Dice: 2.01% | Improved |
| combine all four GSB losses (ablation) | nnU-Net · HVS | Dice: 60.15% to 64.0% | Improved |
| raise noise suppression weight lambda from 0 to 1 (ablation) | HarmonySeg GSB · HVS | Dice: 61.82 -> 64.00 | Improved |
| raise lambda above 1 (ablation) | HarmonySeg GSB · HVS | Dice: 64.00 -> 63.64 | Hurt |
| HarmonySeg vs second best on HVS | HarmonySeg · HVS | Dice: 5.1% | Improved |
Closest papers by meaning
- Dynamic Snake Convolution Based on Topological Geometric Constraints for Tubular Structure Segmentation (ICCV 2023 · similarity 0.90)
- Deep Distance Transform for Tubular Structure Segmentation in CT Scans (CVPR 2020 · similarity 0.90)
- Skeleton Recall Loss for Connectivity Conserving and Resource Efficient Segmentation of Thin Tubular Structures (ECCV 2024 · similarity 0.89)
- GraphMorph: Tubular Structure Extraction by Morphing Predicted Graphs (NeurIPS 2024 · similarity 0.89)
- Upping the Game: How 2D U-Net Skip Connections Flip 3D Segmentation (NeurIPS 2024 · similarity 0.85)
- SegVol: Universal and Interactive Volumetric Medical Image Segmentation (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
