Research map / Label-efficient and robust learning / OOD and anomaly detection
PDD: Manifold-Prior Diverse Distillation for Medical Anomaly Detection
Xijun Lu, Hongying Liu, Fanhua Shang, Yanming Hui, Liang Wan
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 |
|---|---|---|---|
| dual-teacher InA+MMU instead of single-teacher RD (ablation) | PDD · ZhangLab Chest X-ray | AUROC: 81.5 -> 90.8 | Improved |
| add MPA module (prior injection) to dual-teacher single student (ablation) | PDD · ZhangLab Chest X-ray | AUROC: 90.8 -> 92.9 | Improved |
| add MPA to single-teacher RD (ablation) | RD (1t1s) · ZhangLab Chest X-ray | AUROC: 81.5 -> 82.0 | Improved |
| upgrade to dual-student design (ablation) | PDD · ZhangLab Chest X-ray | AUROC: 92.9 -> 94.0 | Improved |
| use teacher-student cosine for anomaly map instead of student-student cosine (with Ldiv) (ablation) | PDD · BrainMRI | AUROC: 32.50 -> 96.67 | Improved |
| add diversity loss Ldiv (with teacher-student scoring) (ablation) | PDD · BrainMRI | AUROC: 93.41 -> 96.67 | Improved |
| vary tau_low/tau_high thresholds (ablation) | PDD · HeadCT | AUROC: 96.15 -> 97.31 | No clear effect |
| PDD vs strongest baseline SimSID | PDD · CheXpert | AUROC: 79.7 -> 79.1 | No clear effect |
| PDD vs MambaAD on Uni-Medical | PDD · Uni-Medical (mean) | F1 max: 82.0 -> 85.4 | Improved |
| PDD vs MambaAD on Uni-Medical | PDD · Uni-Medical (mean) | AUROC: 83.7 -> 81.4 | Hurt |
| PDD vs Skip-TS on HeadCT | PDD · HeadCT | AUROC: 85.7 -> 97.5 | Improved |
Closest papers by meaning
- Adapting Visual-Language Models for Generalizable Anomaly Detection in Medical Images (CVPR 2024 · similarity 0.87)
- Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization (CVPR 2025 · similarity 0.86)
- UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025 · similarity 0.85)
- Scale-Aware Contrastive Reverse Distillation for Unsupervised Medical Anomaly Detection (ICLR 2025 · similarity 0.85)
- Manifold Induced Biases for Zero-shot and Few-shot Detection of Generated Images (ICLR 2025 · similarity 0.84)
- Detecting Generated Images by Fitting Natural Image Distributions (NeurIPS 2025 · 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.
