Research map / Label-efficient and robust learning / Self-supervised and contrastive
Dual-Stream Multiple Instance Learning Network for Whole Slide Image Classification With Self-Supervised Contrastive Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the dual-stream aggregator instead of ABMIL | DSMIL · Camelyon16 | accuracy: 0.8450 -> 0.8682 | Helped |
| Use the dual-stream aggregator instead of ABMIL | DSMIL · Camelyon16 | FROC: 0.4056 -> 0.4296 | Helped |
| Add multiscale attention to DSMIL (ablation) | DSMIL · Camelyon16 | accuracy: 0.8682 -> 0.8992 | Helped |
| Use self-supervised contrastive features instead of ImageNet features (ablation) | DSMIL with fixed aggregator · Camelyon16 | accuracy: 0.6202 -> 0.8682 | Helped |
| Use self-supervised contrastive features instead of max-pooling-trained features (ablation) | DSMIL with fixed aggregator · Camelyon16 | accuracy: 0.7099 -> 0.8682 | Helped |
| Use self-supervised contrastive features instead of patch-based features (ablation) | DSMIL with fixed aggregator · Camelyon16 | accuracy: 0.6977 -> 0.8682 | Helped |
| Use self-supervised contrastive features instead of max-pooling-trained features (ablation) | DSMIL with fixed aggregator · TCGA lung cancer | accuracy: 0.7714 -> 0.9190 | Helped |
| Replace pyramidal multiscale fusion with single-scale 20× features (ablation) | DSMIL · Camelyon16 | accuracy: 0.8682 -> 0.8992 | Hurt |
| Replace pyramidal fusion with concatenation (ablation) | DSMIL · Camelyon16 | accuracy: 0.8682 -> 0.8992 | Hurt |
| Replace pyramidal fusion with max pooling across scales (ablation) | DSMIL · Camelyon16 | accuracy: 0.8604 -> 0.8992 | Hurt |
| Mix instances from different magnifications in one bag (ablation) | DSMIL · Camelyon16 | accuracy: 0.8837 -> 0.8992 | Hurt |
| Add the 1.25× scale to the 5× + 20× fusion (ablation) | DSMIL · Camelyon16 | accuracy: 0.8992 -> 0.8760 | Hurt |
Closest papers by meaning
- Queryable Prototype Multiple Instance Learning with Vision-Language Models for Incremental Whole Slide Image Classification (AAAI 2025 · similarity 0.90)
- CaMIL: Causal Multiple Instance Learning for Whole Slide Image Classification (AAAI 2024 · similarity 0.89)
- Dynamic Policy-Driven Adaptive Multi-Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.89)
- Boosting Multiple Instance Learning Models for Whole Slide Image Classification: A Model-Agnostic Framework Based on Counterfactual Inference (AAAI 2024 · similarity 0.89)
- ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.88)
- FAST: A Dual-tier Few-Shot Learning Paradigm for Whole Slide Image Classification (NeurIPS 2024 · similarity 0.88)
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.