Research map / Image generation / Autoregressive and latent generation
Controllable Latent Space Augmentation for Digital Pathology
Sofiène Boutaj, Marin Scalbert, Pierre Marza, Florent Couzinie-Devy, Maria Vakalopoulou, Stergios Christodoulidis
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 WSI-wise HistAug augmentation | ABMIL-family MIL on UNI · UCEC (10% train) | C-index: 59.3 -> 63.2 | Improved |
| add WSI-wise HistAug augmentation | MIL on CONCH · UCEC (10% train) | C-index: 58.6 -> 64.9 | Improved |
| add WSI-wise HistAug augmentation | MIL on UNI · BRCA (10% train) | AUC: 86.1 -> 88.3 | Improved |
| add WSI-wise HistAug augmentation | MIL on CONCH · NSCLC (10% train) | AUC: 92.8 -> 94.6 | Improved |
| add instance-wise HistAug augmentation | MIL on UNI · BLCA (100% train) | C-index: 54.5 -> 60.3 | Improved |
| replace Gaussian noise perturbation with learned HistAug (WSI) augmentation (ablation) | MIL on CONCH · KIRC (10% train) | C-index: +6.4 | Improved |
| replace Gaussian noise perturbation with learned HistAug (WSI) augmentation (ablation) | MIL on UNI · BRCA (100% train) | AUC: -0.5 | No clear effect |
| add noise perturbation to features (ablation) | MIL · 10% and 100% train | performance | No clear effect |
| apply 10x-trained generator at 20x magnification | MIL on CONCH · UCEC (20x) | C-index: 60.5 -> 67.1 | Improved |
| add chunking of input embedding into tokens (ablation) | HistAug generator · generator training | training MSE loss | Improved |
| add AugDiff augmentation | DSMIL on UNI · BRCA/NSCLC classification (10% train) | AUC: 83.2 -> 76.2 | Hurt |
| add patch augmentation (PAug) | MIL on UNI · classification (100% train) | AUC: 95.4 -> 93.7 | Hurt |
Closest papers by meaning
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization (NeurIPS 2025 · similarity 0.91)
- Promptable Representation Distribution Learning and Data Augmentation for Gigapixel Histopathology WSI Analysis (AAAI 2025 · similarity 0.89)
- Rethinking Transformer for Long Contextual Histopathology Whole Slide Image Analysis (NeurIPS 2024 · similarity 0.88)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.88)
- Morphological Prototyping for Unsupervised Slide Representation Learning in Computational Pathology (CVPR 2024 · similarity 0.88)
- CPath-Omni: A Unified Multimodal Foundation Model for Patch and Whole Slide Image Analysis in Computational Pathology (CVPR 2025 · 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.
