Research map / Probabilistic and scientific modeling / Differential equations and PDEs
CNM-UNet: Continuous Ordinary Differential Equations for Medical Image Segmentation
Tianqi Xu, Yashi Zhu, Quansong He, Yue Cao, Kaishen Wang, Zhang Yi, Tao He
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 |
|---|---|---|---|
| replace all decoder layers with single CNM-Block (vs Continuous UNet) | CNM-UNet · Polyp-A | GFLOPs: 394.15 -> 22.53 | Improved |
| single continuous decoder vs full Continuous UNet | CNM-UNet · Polyp-C | DSC: 0.801 -> 0.814 | Improved |
| single continuous decoder vs Continuous UNet | CNM-UNet · ISIC2018 | DSC: 0.885 -> 0.879 | Hurt |
| continuous CNM-Block vs vanilla UNet (training time) | CNM-UNet · BKAI-IGH-NeoPolyp | training time (h): 1.38 -> 1.80 | Hurt |
| continuous CNM-Block vs discrete DiODE-UNet (DSC) | CNM-UNet · Polyp-C | DSC: 0.807 -> 0.814 | Improved |
| add DUSE test-time adaptation | CNM-UNet · OD/OC test set A | DSC: 0.335 -> 0.709 | Improved |
| add DUSE test-time adaptation | CNM-UNet · OD/OC test set D | DSC: 0.498 -> 0.712 | Improved |
| add DUSE test-time adaptation | CNM-UNet · OD/OC test set C | DSC: 0.732 -> 0.709 | Hurt |
| add DUSE test-time adaptation | CNM-UNet · Polyp test set C | DSC: 0.656 -> 0.626 | Hurt |
| add DUSE test-time adaptation | CNM-UNet · Polyp test set B | DSC: 0.237 -> 0.321 | Improved |
| add DUSE vs prompt-only VPTTA | CNM-UNet · Polyp test set A | DSC: 0.554 -> 0.559 | Improved |
| use parameter-updating TTA (AdaBN, Tent, CoTTA) instead of source only | CNM-UNet · OD/OC test set B | DSC: 0.789 -> 0.150 | Hurt |
Closest papers by meaning
- FuseUNet: A Multi-Scale Feature Fusion Method for U-like Networks (ICML 2025 · similarity 0.85)
- Improving Neural ODE Training with Temporal Adaptive Batch Normalization (NeurIPS 2024 · similarity 0.85)
- U-KAN Makes Strong Backbone for Medical Image Segmentation and Generation (AAAI 2025 · similarity 0.84)
- MedSegDiff-V2: Diffusion-Based Medical Image Segmentation with Transformer (AAAI 2024 · similarity 0.84)
- E2ENet: Dynamic Sparse Feature Fusion for Accurate and Efficient 3D Medical Image Segmentation (NeurIPS 2024 · similarity 0.83)
- MedFlowSeg: Flow Matching for Medical Image Segmentation with Frequency-Aware Attention (NeurIPS 2026 · similarity 0.83)
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.
