Research map / Detection and segmentation / Open-set and few-shot detection
Scaling Tumor Segmentation: Best Lessons from Real and Synthetic Data
Qi Chen, Xinze Zhou, Chen Liu, Hao Chen, Wenxuan Li, Zekun Jiang, Ziyan Huang, Yuxuan Zhao, Dexin Yu, Junjun He, Yefeng Zheng, Ling Shao and 2 more
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 3x synthetic tumors to real training scans (~435 real) (ablation) | SegResNet · JHH pancreatic, out-of-distribution (PANORAMA) | DSC: 36.8 -> 52.8 | Improved |
| add 3x synthetic tumors at full real scale (3159 scans) (ablation) | SegResNet · JHH pancreatic, out-of-distribution (PANORAMA) | DSC: 41.2 -> 53.5 | Improved |
| add synthetic tumors to 278 real scans, in-distribution (ablation) | SegResNet · JHH pancreatic, in-distribution | DSC: 52.7 -> 58.1 | Improved |
| add synthetic tumors at 3159 real scans, in-distribution (ablation) | SegResNet · JHH pancreatic, in-distribution | DSC: 59.7 -> 59.3 | No clear effect |
| scale real scans from 1500 to 3159 | SegResNet · JHH pancreatic, in-distribution | DSC: 59.3 -> 59.7 | No clear effect |
| scale real scans from 1500 to 3159 | SegResNet · JHH pancreatic, out-of-distribution (PANORAMA) | DSC: 38.4 -> 41.2 | Improved |
| train ResEncM on Cancerverse instead of public datasets | ResEncM · Cancerverse val, pancreatic tumor | DSC: 57.0 -> 65.8 | Improved |
| train ResEncM on Cancerverse instead of public datasets | ResEncM · 3D-IRCADb liver tumor (external) | DSC: 67.1 -> 81.1 | Improved |
| train on Cancerverse, colon tumor NSD | ResEncM · Cancerverse val, colon tumor | NSD: 48.7 -> 47.6 | Hurt |
| train on Cancerverse, liver tumor sensitivity | ResEncM · Cancerverse val, liver tumor | sensitivity: 89.1 -> 83.7 | Hurt |
| train ResEncM on Cancerverse instead of public datasets | ResEncM · MSD Task07 Pancreas (test) | DSC: 62.3 -> 67.2 | Improved |
| train ResEncM on Cancerverse instead of public datasets | ResEncM · Kipa kidney tumor (external) | DSC: 76.4 -> 83.6 | Improved |
Closest papers by meaning
- Label-Free Liver Tumor Segmentation (CVPR 2023 · similarity 0.89)
- Towards Generalizable Tumor Synthesis (CVPR 2024 · similarity 0.88)
- Touchstone Benchmark: Are We on the Right Way for Evaluating AI Algorithms for Medical Segmentation? (NeurIPS 2024 · similarity 0.87)
- PanTS: The Pancreatic Tumor Segmentation Dataset (NeurIPS 2025 · similarity 0.87)
- "Why Is There a Tumor?": Tell Me the Reason, Show Me the Evidence (ICML 2025 · similarity 0.85)
- ZePT: Zero-Shot Pan-Tumor Segmentation via Query-Disentangling and Self-Prompting (CVPR 2024 · similarity 0.85)
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.
