Research map / Label-efficient and robust learning / OOD and anomaly detection
Exploiting Structural Consistency of Chest Anatomy for Unsupervised Anomaly Detection in Radiography Images
Tiange Xiang, Yixiao Zhang, Yongyi Lu, Alan Yuille, Chaoyi Zhang, Weidong Cai, Zongwei Zhou
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 space-aware memory (ablation) | SimSID · ZhangLab ablation | AUC (%): 77.6 -> 91.1 | Improved |
| add skip connections (ablation) | SimSID · ZhangLab ablation | AUC (%): 79.5 -> 91.1 | Improved |
| add feature-level in-painting block (ablation) | SimSID · ZhangLab ablation | AUC (%): 80.9 -> 91.1 | Improved |
| add hierarchical memory (three matrices) (ablation) | SimSID · ZhangLab ablation | AUC (%): 82.9 -> 91.1 | Improved |
| add teacher-student knowledge distillation (ablation) | SimSID · ZhangLab ablation | AUC (%): 85.4 -> 91.1 | Improved |
| add stop-gradient between teacher generator and encoder (ablation) | SimSID · ZhangLab ablation | AUC (%): 85.0 -> 91.1 | Improved |
| add Gumbel shrinkage (ablation) | SimSID · ZhangLab ablation | AUC (%): 86.2 -> 91.1 | Improved |
| replace memory queue with space-aware memory matrix (ablation) | SimSID · ZhangLab ablation | AUC (%): 86.7 -> 91.1 | Improved |
| use transformer instead of convolution for in-painting (ablation) | SimSID · ZhangLab ablation | AUC (%): 86.3 -> 91.1 | Improved |
| in-paint features instead of pixels (ablation) | SimSID · ZhangLab ablation | AUC (%): 79.1 -> 91.1 | Improved |
| activate top-5 patterns instead of top-1 in Gumbel shrinkage (ablation) | SimSID · ZhangLab ablation | AUC (%): -18.5% AUC for top-1 | Improved |
| use SimSID instead of SQUID | anomaly detector · ZhangLab | AUC (%): 87.6 -> 91.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PANDA — PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation (CVPR 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization (CVPR 2025 · similarity 0.85)
- Leveraging Anatomical Consistency for Multi-Object Detection in Ultrasound Images via Source-free Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.83)
- AFiRe: Anatomy-Driven Self-Supervised Learning for Fine-Grained Representation in Radiographic Images (AAAI 2025 · similarity 0.83)
- Adapting Visual-Language Models for Generalizable Anomaly Detection in Medical Images (CVPR 2024 · similarity 0.83)
- Novel Class Discovery in Chest X-rays via Paired Images and Text (AAAI 2024 · similarity 0.83)
- Unsupervised Domain Adaptation for Anatomical Structure Detection in Ultrasound Images (ICML 2024 · similarity 0.83)
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.
