Research map / Vision-language and multimodal / Cross-modal learning
CARE: A Molecular-Guided Foundation Model with Adaptive Region Modeling for Whole Slide Image Analysis
Di Zhang, Zhangpeng Gong, Xiaobo Pang, Jiashuai Liu, Junbo Lu, Hao Cui, Jiusong Ge, Zhi Zeng, Kai Yi, Yinghua Li, Si Liu, Tingsong Yu and 5 more
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 adaptive regions instead of fixed subregions (ablation) | CARE (random init) · BCNB-HER2 | ACC: 61.4 -> 64.8 | Improved |
| Use adaptive regions instead of fixed subregions (ablation) | CARE (random init) · EBRAINS-fine | ACC: 71.6 -> 72.5 | Improved |
| Add CLS-aggregated subregion feature to adaptive region generator (ablation) | CARE (random init) · BCNB-HER2 | ACC: 58.9 -> 64.8 | Improved |
| Add patch feature f_i to the region similarity (ablation) | CARE (random init) · EBRAINS-fine | ACC: 70.1 -> 72.5 | Improved |
| Add Region Structuring Loss target E*=0.5 instead of 0 or 2 (ablation) | CARE (random init) · BCNB-HER2 | ACC: 63.9 (E*=0), 59.1 (E*=2) -> 64.8 | Improved |
| Increase lambda_RSL from 0.1 to 0.3 (ablation) | CARE (random init) · BCNB-HER2 | ACC: 64.8 -> 58.6 | Hurt |
| Use window size k=8 instead of 6 or 10 (ablation) | CARE (random init) · BCNB-HER2 | ACC: 59.4 (k=6), 61.9 (k=10) -> 64.8 | Improved |
| Fuse coverage prior and semantic attention (lambda_SPF=0.5) instead of either alone (ablation) | CARE (random init) · BCNB-HER2 | ACC: 62.5 (0), 60.0 (1) -> 64.8 | Improved |
| Scale iBOT pretraining data from 24,000 WSIs to full (ablation) | CARE-iBOT · BCNB-HER2 (linear probing) | ACC: 56.9 -> 60.3 | Improved |
| Add RNA-guided cross-modal pretraining after iBOT (ablation) | CARE-iBOT · EBRAINS-fine (linear probing) | ACC: 72.2 -> 73.7 | Improved |
| Add RNA-guided cross-modal pretraining after iBOT (ablation) | CARE-iBOT · BCNB-HER2 (linear probing) | ACC: 60.3 -> 59.7 | Hurt |
| Add protein-guided pretraining after RNA stage (ablation) | CARE-RNA · EBRAINS-fine (linear probing) | ACC: 73.7 -> 74.0 | Improved |
Closest papers by meaning
- Semantic and Visual Crop-Guided Diffusion Models for Heterogeneous Tissue Synthesis in Histopathology (NeurIPS 2025 · similarity 0.89)
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization (NeurIPS 2025 · similarity 0.89)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.89)
- Multi-Resolution Pathology-Language Pre-training Model with Text-Guided Visual Representation (CVPR 2025 · similarity 0.89)
- Morphological Prototyping for Unsupervised Slide Representation Learning in Computational Pathology (CVPR 2024 · similarity 0.88)
- STimage-1K4M: A histopathology image-gene expression dataset for spatial transcriptomics (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.
