Research map / Detection and segmentation / Open-set and few-shot detection
CancerUniT: Towards a Single Unified Model for Effective Detection, Segmentation, and Diagnosis of Eight Major Cancers Using a Large Collection of CT Scans
Jieneng Chen, Yingda Xia, Jiawen Yao, Ke Yan, Jianpeng Zhang, Le Lu, Fakai Wang, Bo Zhou, Mingyan Qiu, Qihang Yu, Mingze Yuan, Wei Fang and 13 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 query hierarchy (detection-to-diagnosis projection) vs plain queries (ablation) | CancerUniT · 8-cancer CT test set | average detection sensitivity (%): 89.5 -> 93.3 | Improved |
| link detection and diagnosis queries hierarchically instead of parallel groups (ablation) | CancerUniT · 8-cancer CT test set | tumor Dice: 0.608 -> 0.628 | Improved |
| link detection and diagnosis queries hierarchically instead of parallel groups (ablation) | CancerUniT · 8-cancer CT test set | normal-control specificity (%): 78.9 -> 81.7 | Improved |
| train one unified model for all organs instead of 8 separate nnUNets | nnUNet · 8-cancer CT test set | specificity on normal controls (%): 59.2 -> 78.9 | Improved |
| train one unified nnUNet instead of separate per-organ models | nnUNet · 8-cancer CT test set | patient-level diagnosis sensitivity (%): 68.1 -> 64.2 | Hurt |
| unified multi-organ training for breast tumor detection | nnUNet · 8-cancer CT test set | breast detection sensitivity (%): 96.0 -> 90.0 | Hurt |
| add Transformer query decoder with hierarchy to unified nnUNet | unified nnUNet · 8-cancer CT test set | diagnosis sensitivity (%): 64.2 -> 70.9 | Improved |
| use semantic segmentation instead of object detection (LENS) for tumor detection | tumor detection · 8-cancer CT test set | average detection sensitivity (%): 72.3 -> 88.0 | Improved |
| train on the 8-cancer dataset rather than DeepLesion | lesion detection · normal-control specificity | average specificity (%): 4.2 -> 64.8 | Improved |
| keep CancerUniT instead of TransUNet for class-agnostic lesion recall | CancerUniT · 8-cancer CT test set | lesion recall (%): 77.9 -> 75.0 | Hurt |
| unified CancerUniT instead of 8 single-organ models: speed and parameters | CancerUniT · Inference efficiency | inference time: 187s -> 42s | Improved |
Closest papers by meaning
- ZePT: Zero-Shot Pan-Tumor Segmentation via Query-Disentangling and Self-Prompting (CVPR 2024 · similarity 0.86)
- TumorChain: Interleaved Multimodal Chain-of-Thought Reasoning for Traceable Clinical Tumor Analysis (ICLR 2026 · similarity 0.86)
- "Why Is There a Tumor?": Tell Me the Reason, Show Me the Evidence (ICML 2025 · similarity 0.86)
- Towards Generalizable Tumor Synthesis (CVPR 2024 · similarity 0.86)
- CAT: Coordinating Anatomical-Textual Prompts for Multi-Organ and Tumor Segmentation (NeurIPS 2024 · similarity 0.86)
- LungNoduleAgent: A Collaborative Multi-Agent System for Precision Diagnosis of Lung Nodules (AAAI 2026 · 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.
