Research map / Label-efficient and robust learning / Recognition and long-tail
FACT: Fuzzy Alignment with Comorbidity Topology for Reliable Multi-Label Medical Image Diagnosis
Yingyu Chen, Yongqiang Huang, Yang Qin, Ziyuan Yang, Lang Yuan, Maosong Ran, Yi Zhang
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 Atomic Visual Space (vector-quantized shared visual atoms) (ablation) | FACT · ODIR-5K | mAP: 60.73 -> 70.31 | Improved |
| add Comorbidity Topology semantic anchors (ablation) | FACT · ODIR-5K | mAP: 60.73 -> 69.12 | Improved |
| add Comorbidity Topology on top of Atomic Visual Space (ablation) | FACT · ODIR-5K | mAP: 70.31 -> 72.61 | Improved |
| add set-level Jaccard loss L_set (ablation) | FACT · ODIR-5K | F1: 89.00 -> 90.70 | Improved |
| add VQ commitment loss L_commit (ablation) | FACT · ODIR-5K | mAP: 68.91 -> 72.11 | Improved |
| use fuzzy alignment with L_fuzzy instead of rigid classifier with BCE (ablation) | FACT · ODIR-5K | CER: 27.51 -> 16.22 | Improved |
| train Gaussian fuzzy membership with BCE instead of L_fuzzy (ablation) | FACT · ODIR-5K | CER: 16.22 -> 31.08 | Hurt |
| use clinical comorbidity prior instead of identity adjacency (ablation) | FACT · RSNA-IHD | mAP: 87.11 -> 89.72 | Improved |
| use clinical comorbidity prior instead of shuffled adjacency (ablation) | FACT · ODIR-5K | Avg: 84.02 -> 85.50 | Improved |
| use Gaussian kernel instead of Laplacian kernel (ablation) | FACT · ODIR-5K | mAP: 30.98 -> 72.61 | Improved |
| use Gaussian kernel instead of linear kernel (ablation) | FACT · NIH-Chest | mAP: 51.94 -> 58.95 | Improved |
| add fuzzy objective and topology on CXR-LT Task 2 (ablation) | FACT · CXR-LT Task 2 Tail | mean score: 67.16 -> 74.30 | Improved |
Closest papers by meaning
- Multi-Granularity Cross-modal Alignment for Generalized Medical Visual Representation Learning (NeurIPS 2022 · similarity 0.87)
- LLM-Guided Diagnostic Evidence Alignment for Medical Vision–Language Pretraining under Limited Pairing (ICML 2026 · similarity 0.86)
- Medical thinking with multiple images (ICLR 2026 · similarity 0.84)
- M3CoTBench: Benchmark Chain-of-Thought of MLLMs in Medical Image Understanding (ICLR 2026 · similarity 0.84)
- RAD: Towards Trustworthy Retrieval-Augmented Multi-modal Clinical Diagnosis (NeurIPS 2025 · similarity 0.83)
- Coherent Hierarchical Multi-Label Learning to Defer for Medical Imaging (NeurIPS 2026 · 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.
