Research map / Reinforcement learning / Visual and representation RL
Glance and Focus Reinforcement for Pan-cancer Screening
Linshan Wu, Jia-Xin Zhuang, Hao Chen
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 Glance-Focus RL (GF-Screen) to SwinUNETR | SwinUNETR · FLARE23 | DSC (%): 41.5 -> 56.7 | Improved |
| add GF-Screen to nnUNet (ablation) | nnUNet · FLARE23 | DSC (%): 43.7 -> 45.6 | Improved |
| add GF-Screen to TransUNet (ablation) | TransUNet · FLARE23 | DSC (%): 27.2 -> 34.8 | Improved |
| add GF-Screen to UNETR (ablation) | UNETR · FLARE23 | DSC (%): 16.2 -> 27.3 | Improved |
| replace cross-entropy Glance training with group relative learning (ablation) | Glance model · FLARE23 | DSC (%): 37.6 -> 56.7 | Improved |
| replace PPO actor-critic with group relative learning (ablation) | Glance model · FLARE23 | DSC (%): 24.5 -> 56.7 | Improved |
| use binary detection reward instead of DSC reward (ablation) | GRL · FLARE23 | DSC (%): 43.2 -> 53.1 | Improved |
| add classification loss alpha CE to GRL objective (ablation) | GRL · FLARE23 | DSC (%): 53.1 -> 56.7 | Improved |
| add discard of healthy sub-volumes (ablation) | GF-Screen · FLARE23 | GFLOP: 13726 -> 2164 | Improved |
| increase group size N from 4 to 16 (ablation) | GRL · FLARE23 | DSC (%): 45.9 -> 56.5 | Improved |
| use 3D ResNet-34 instead of ResNet-18 as Glance model (ablation) | Glance model · FLARE23 | DSC (%): 56.7 -> 56.5 | No clear effect |
| swap Glance backbone to 3D UNet (ablation) | Glance model · FLARE23 | DSC (%): 56.7 -> 56.0 | No clear effect |
Closest papers by meaning
- Deep Reinforcement Learning for Early Diagnosis of Lung Cancer (AAAI 2024 · similarity 0.84)
- ScreenShot: A Foundation Model for Few-Shot Combination Drug Screening (NeurIPS 2026 · similarity 0.83)
- Sequential Attention-based Sampling for Histopathological Analysis (NeurIPS 2025 · similarity 0.83)
- CLIMB: Data Foundations for Large Scale Multimodal Clinical Foundation Models (ICML 2025 · similarity 0.83)
- Cross-Modal Interactive Perception Network with Mamba for Lung Tumor Segmentation in PET-CT Images (CVPR 2025 · similarity 0.82)
- MedEyes: Learning Dynamic Visual Focus for Medical Progressive Diagnosis (AAAI 2026 · similarity 0.82)
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.
