Research map / Detection and segmentation / Open-set and few-shot detection
COME: Dual Structure-Semantic Learning with Collaborative MoE for Universal Lesion Detection Across Heterogeneous Ultrasound Datasets
Lingyu Chen, Yawen Zeng, Yue Wang, Peng Wan, Guochen Ning, Hongen Liao, Daoqiang Zhang, Fang Chen
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 |
|---|---|---|---|
| remove both shared experts (STE and SEE) (ablation) | COME · inter-organ integration | mean AP: 0.5819 -> 0.6587 when added (-7.68% mean AP) | Improved |
| add clustering before routing (ablation) | COME · inter-organ integration | mean AP: 0.6264 -> 0.6587 | Improved |
| add source-specific traceability loss (ablation) | COME · inter-organ integration | mean AP: 0.6320 -> 0.6587 | Improved |
| train on integrated inter-organ data instead of single dataset | DINO · 8 US datasets | mean AP: 0.6125 (single) -> 0.5792 (inter-organ) | Hurt |
| use COME instead of DINO under joint training | US lesion detection · inter-organ integration | mean AP: 0.5792 -> 0.6587 | Improved |
| set the number of experts to 8 (equal to number of sources) (ablation) | COME · inter-organ integration | mean AP: 4 experts 0.6416, 10 experts 0.6459 -> 8 experts 0.6565 | Improved |
| use Top-K of 3 instead of 4 (ablation) | COME router · inter-organ integration | mean AP: 0.6333 -> 0.6587 | Improved |
| train on BUSI and test on other breast datasets | COME vs DINO · BUSC zero-shot | AP: 0.043 -> 0.297 | Improved |
| combine both clustering strategies (Mixing) (ablation) | COME · inter-organ integration | mean AP and inference time: 0.6587 -> 0.6337; 132 -> 845 us | Hurt |
| use Fine2Coarse instead of Multi-Step clustering (ablation) | COME · inter-organ integration | inference time: 615 -> 132 us with comparable mean AP | 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.
- DAMEX — DAMEX: Dataset-aware Mixture-of-Experts for visual understanding of mixture-of-datasets (NeurIPS 2023) · 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
- Leveraging Anatomical Consistency for Multi-Object Detection in Ultrasound Images via Source-free Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.87)
- U2-BENCH: Benchmarking Large Vision-Language Models on Ultrasound Understanding (ICLR 2026 · similarity 0.86)
- CSC-PA: Cross-image Semantic Correlation via Prototype Attentions for Single-network Semi-supervised Breast Tumor Segmentation (CVPR 2025 · similarity 0.86)
- Anatomical Knowledge Mining and Matching for Semi-supervised Medical Multi-structure Detection (AAAI 2025 · similarity 0.85)
- Unsupervised Domain Adaptation for Anatomical Structure Detection in Ultrasound Images (ICML 2024 · similarity 0.83)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · 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.
