Research map / Image generation / Autoregressive and latent generation
Neural Image Compression for Gigapixel Histopathology Image Analysis
David Tellez, Geert J. S. Litjens, Jeroen A. W. M. van der Laak, Francesco Ciompi
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 |
|---|---|---|---|
| use a BiGAN encoder instead of a VAE encoder (ablation) | NIC · Camelyon16 WSI metastasis (All) | AUC: 0.661 -> 0.725 | Improved |
| use a BiGAN encoder instead of a contrastive encoder (ablation) | NIC · Camelyon16 WSI metastasis (All) | AUC: 0.608 -> 0.725 | Improved |
| use a BiGAN encoder instead of a VAE encoder for tissue classification (ablation) | frozen encoder + MLP · Rectum global tissue classification | accuracy: 0.639 -> 0.765 | Improved |
| use a VAE encoder instead of a contrastive encoder (ablation) | frozen encoder + MLP · Rectum global tissue classification | accuracy: 0.520 -> 0.639 | Improved |
| replace the mean-RGB color baseline with a BiGAN learned encoder (ablation) | NIC · Camelyon16 WSI metastasis (All) | AUC: 0.582 -> 0.725 | Improved |
| use a supervised tumor encoder instead of unsupervised BiGAN on macro metastases (ablation) | NIC · Camelyon16 WSI metastasis (Macro) | AUC: 0.720 -> 0.914 | Improved |
| use a BiGAN encoder instead of a supervised-tumor encoder for tumor proliferation (ablation) | NIC · TUPAC16 WSI proliferation speed (All) | Spearman correlation: 0.427 -> 0.522 | Improved |
| use a larger code size to compress the images (ablation) | NIC · Synthetic data | accuracy | Improved |
| use a smaller stride, giving embedded images with larger spatial resolution (ablation) | NIC with oracle encoder · Synthetic data | accuracy | Hurt |
| shrink the target lesion below 10% of the image size (ablation) | NIC · Synthetic data | accuracy | Hurt |
| use macro-metastasis-only positives instead of all positives for evaluating BiGAN | NIC with BiGAN · Camelyon16 WSI metastasis | AUC: 0.704 -> 0.720 | Improved |
Closest papers by meaning
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.87)
- Sequential Attention-based Sampling for Histopathological Analysis (NeurIPS 2025 · similarity 0.86)
- Promptable Representation Distribution Learning and Data Augmentation for Gigapixel Histopathology WSI Analysis (AAAI 2025 · similarity 0.86)
- Dynamic Entity-Masked Graph Diffusion Model for Histopathology Image Representation Learning (AAAI 2025 · similarity 0.86)
- STimage-1K4M: A histopathology image-gene expression dataset for spatial transcriptomics (NeurIPS 2024 · similarity 0.85)
- Beyond ViT Tokens: Masked-Diffusion Pretrained Convolutional Pathology Foundation Model for Cell-Level Dense Prediction (NeurIPS 2026 · similarity 0.85)
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.
