Research map / Label-efficient and robust learning / Recognition and long-tail
Contrastive Cross-Bag Augmentation for Multiple Instance Learning-based Whole Slide Image Classification
Bo Zhang, Xinan Xu, Shuo Yan, Yu Bai, Zheng Zhang, Wufan Wang, Hui Gao, Wendong Wang
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 Cross-Bag Augmentation (all three components) (ablation) | C2Aug · CAMELYON-16 (ResNet50, TransMIL) | AUC: 90.9 -> 94.7 | Improved |
| add Multi-View Fusion (ablation) | C2Aug · CAMELYON-16 (ResNet50, TransMIL) | AUC: 91.5 -> 94.7 | Improved |
| add Instance Compression (ablation) | C2Aug · CAMELYON-16 (ResNet50, TransMIL) | AUC: 91.9 -> 94.7 | Improved |
| add Instance Expansion (ablation) | C2Aug · CAMELYON-16 (ResNet50, TransMIL) | AUC: 92.5 -> 94.7 | Improved |
| keep the input bag in Multi-View Fusion instead of replacing it with sampled instances (ablation) | C2Aug · CAMELYON-16 normal slides (ResNet50, TransMIL) | AUC: 92.1 -> 95.3 | Improved |
| keep the input bag in Instance Expansion (ablation) | C2Aug · CAMELYON-16 normal slides (ResNet50, TransMIL) | AUC: 92.6 -> 95.3 | Improved |
| sample instances from all bags with the same label instead of 4 bags (ablation) | C2Aug · CAMELYON-16 (ResNet50, TransMIL) | AUC: 92.3 -> 94.7 | Improved |
| use random instance counts per bag instead of equal counts when sampling from all bags (ablation) | C2Aug · CAMELYON-16 (ResNet50, TransMIL) | AUC: 93.9 -> 94.7 | Improved |
| add bag-level and group-level contrastive losses (ablation) | C2Aug · CAMELYON-16 test, bags with <1% tumor | ACC: 80.6 -> 82.4 | Improved |
| add bag-level and group-level contrastive losses (ablation) | C2Aug · CAMELYON-16 test, all bags | ACC: 89.4 -> 91.9 | Improved |
| use row-wise random masking instead of fixed views (no mask) (ablation) | Multi-View Fusion · CAMELYON-16 (ResNet50, TransMIL) | AUC: 92.2 -> 94.7 | Improved |
| use row-wise masking instead of element-wise masking (ablation) | Multi-View Fusion · CAMELYON-16 (ResNet50, TransMIL) | AUC: 93.0 -> 94.7 | Improved |
Closest papers by meaning
- OODML: Whole Slide Image Classification Meets Online Pseudo-Supervision and Dynamic Mutual Learning (AAAI 2025 · similarity 0.86)
- Boosting Multiple Instance Learning Models for Whole Slide Image Classification: A Model-Agnostic Framework Based on Counterfactual Inference (AAAI 2024 · similarity 0.86)
- Queryable Prototype Multiple Instance Learning with Vision-Language Models for Incremental Whole Slide Image Classification (AAAI 2025 · similarity 0.85)
- Dynamic Policy-Driven Adaptive Multi-Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.85)
- CaMIL: Causal Multiple Instance Learning for Whole Slide Image Classification (AAAI 2024 · similarity 0.84)
- ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification (CVPR 2024 · 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.
