Research map / Label-efficient and robust learning / Self-supervised and contrastive
Benchmarking Self-Supervised Learning on Diverse Pathology Datasets
Read the paperCheck 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 domain-aligned TCGA SSL pre-training instead of ImageNet-supervised pre-training | Pathology classification and nuclei segmentation · BACH, CRC, PCam, MHIST, CoNSeP | Top-1 accuracy or mPQ | Helped |
| Add pathology data from TULIP to TCGA pre-training | MoCo · BACH, CRC, PCam, MHIST, CoNSeP linear evaluation | Top-1 accuracy or mPQ: 80.83 -> 85.00; 90.93 -> 93.94; 80.79 -> 86.53; 76.25 -> 82.29; 33.49 -> 41.40 | Mixed |
| Use both 20x and 40x magnifications rather than one magnification (ablation) | Pathology SSL pre-training · BACH, CRC, PCam, MHIST, CoNSeP fine-tuning | Top-1 accuracy or mPQ | Mixed |
| Add random vertical flips and remove solarization (ablation) | BT pre-training · CoNSeP | mPQ: 50.71 -> 51.07 | Helped |
| Use RandStainNA or RandStainNA GMM instead of HED-light (ablation) | BT pre-training · CoNSeP | mPQ: 50.48; 50.86; 50.71 | Helped |
| Add random grayscale to RandStainNA GMM (ablation) | BT pre-training · CoNSeP | mPQ: 50.99 -> 51.13 | Helped |
| Use strong rather than weak color jitter with stain augmentation (ablation) | BT pre-training · CoNSeP | mPQ: 51.13 -> 50.27; 51.61 -> 50.51 | Hurt |
| Use RandStainNA GMM, grayscale, and weak color jitter (ablation) | BT pre-training · CoNSeP | mPQ: 50.99 -> 51.61 | Helped |
| Increase pre-training from 200 to 800 epochs (ablation) | MoCo, SwAV, and BT · CoNSeP | mPQ: 51.16 -> 52.39 | Mixed |
| Increase ViT input size from 224 to 272 (ablation) | DINO · CoNSeP | mPQ: 32.54 -> 35.81; 38.43 -> 41.13; 42.71 -> 40.08; 46.70 -> 44.24 | Mixed |
Closest papers by meaning
- Revisiting End-to-End Learning with Slide-level Supervision in Computational Pathology (NeurIPS 2025 · similarity 0.87)
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization (NeurIPS 2025 · similarity 0.87)
- Unsupervised Foundation Model-Agnostic Slide-Level Representation Learning (CVPR 2025 · similarity 0.87)
- Do Multiple Instance Learning Models Transfer? (ICML 2025 · similarity 0.87)
- PathView-Bench: Can Multimodal Large Language Models Achieve Fine-grained Multiscale Understanding of Pathology Images? (NeurIPS 2026 · similarity 0.87)
- Fast and Accurate Gigapixel Pathological Image Classification with Hierarchical Distillation Multi-Instance Learning (CVPR 2025 · similarity 0.86)
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.