Research map / Label-efficient and robust learning / OOD and anomaly detection
Normal and Abnormal Pathology Knowledge-Augmented Vision-Language Model for Anomaly Detection in Pathology Images
Jinsol Song, Jiamu Wang, Anh Tien Nguyen, Keunho Byeon, Sangjeong Ahn, Sung Hak Lee, Jin Tae Kwak
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 Ano-NAViLa instead of the best competing anomaly detector | pathology anomaly detection · Camelyon16 (external, max score) | AUROC: gains of 0.1436--0.3778 in AUROC | Improved |
| use text-augmented embeddings with both normal and abnormal term pools instead of removing any of the three embeddings (ablation) | Ano-NAViLa · GastricLN and Camelyon16 | AUROC/AUPR | Improved |
| use text embeddings only (drop image embedding) (ablation) | Ano-NAViLa · GastricLN and Camelyon16 | AUROC/AUPR | Hurt |
| add normal text embedding to image embedding (ablation) | Ano-NAViLa · GastricLN vs Camelyon16 | AD performance | Mixed |
| add abnormal text embedding on top of image and normal text embeddings (ablation) | Ano-NAViLa · GastricLN and Camelyon16 | AD performance | Improved |
| vary anomaly scoring (max, L2-norm, AE, SVM, GMM instead of sum) (ablation) | Ano-NAViLa · GastricLN and Camelyon16 | AUROC/AUPR: GastricLN patch AUROC 0.9681 (sum) vs 0.9669 (max) | No clear effect |
| remove erosion smoothing of heatmap (ablation) | Ano-NAViLa · GastricLN and Camelyon16 | AD performance | No clear effect |
| use frozen VLM with trainable MLP instead of larger models | Ano-NAViLa · efficiency | trainable parameters (M): 0.69 M vs 2.78 M (STFPM next lowest) | Improved |
| use zero-shot CONCH with the same term pools (ablation) | CONCH · lymph node metastasis detection | AD performance | Hurt |
Closest papers by meaning
- Adapting Visual-Language Models for Generalizable Anomaly Detection in Medical Images (CVPR 2024 · similarity 0.86)
- Decomposing Disease Descriptions for Enhanced Pathology Detection: A Multi-Aspect Vision-Language Pre-training Framework (CVPR 2024 · similarity 0.85)
- Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization (CVPR 2025 · similarity 0.85)
- CPLIP: Zero-Shot Learning for Histopathology with Comprehensive Vision-Language Alignment (CVPR 2024 · similarity 0.84)
- PathAsst: A Generative Foundation AI Assistant towards Artificial General Intelligence of Pathology (AAAI 2024 · similarity 0.84)
- PathView-Bench: Can Multimodal Large Language Models Achieve Fine-grained Multiscale Understanding of Pathology Images? (NeurIPS 2026 · similarity 0.84)
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.
