Research map / Detection and segmentation / Open-set and few-shot detection
Sim4Seg: Boosting Multimodal Multi-disease Medical Diagnosis Segmentation with Region-Aware Vision-Language Similarity Masks
Lingran Song, Yucheng Zhou, Jianbing Shen
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 RVLS2M similarity-mask prompt to SAM decoder (ablation) | LISA fine-tuned with diagnosis CoT · M3DS test (avg) | gIoU: 35.16 -> 40.34 | Improved |
| add RVLS2M similarity-mask prompt (ablation) | LISA fine-tuned with diagnosis CoT · M3DS test (avg) | diagnosis accuracy: 64.64 -> 69.90 | Improved |
| add diagnosis chain-of-thought to fine-tuning targets (vs diagnosis only) (ablation) | Sim4Seg · M3DS test (overall) | diagnosis accuracy: 54.33 -> 69.04 | Improved |
| add diagnosis CoT to fine-tuning targets (ablation) | Sim4Seg · M3DS test (overall) | gIoU: 51.00 -> 51.86 | No clear effect |
| add test-time scaling (multi-CoT, multi-mask selection) (ablation) | Sim4Seg (ft-CoT) · M3DS test (overall) | diagnosis accuracy: 69.04 -> 82.63 | Improved |
| add test-time scaling (ablation) | Sim4Seg (ft-CoT) · M3DS test (overall) | gIoU: 51.86 -> 53.11 | Improved |
| add RVLS2M without training (ablation) | LISA · M3DS test (zero-shot) | avg of gIoU and cIoU: 32.13 -> 35.85 (gIoU 32.43 -> 31.82, cIoU 31.83 -> 39.88) | Mixed |
| fine-tune on M3DS with diagnosis | LISA (without RVLS2M) · M3DS test (avg) | gIoU: 26.42 -> 35.21 | Improved |
| RVLS2M grid resolution 16x16 vs coarser/finer (ablation) | Sim4Seg · M3DS test | avg performance: 33.04 / 52.88 / 50.27 / 47.55 for 8/16/32/64 | Mixed |
| Sim4Seg vs LISA on held-out modalities | Sim4Seg · M3DS cross-modality | gIoU/cIoU: Endoscopy cIoU 35.86 -> 31.12; X-Ray gIoU 6.34 -> 10.45 | Mixed |
Closest papers by meaning
- Interactive Medical Image Segmentation: A Benchmark Dataset and Baseline (CVPR 2025 · similarity 0.87)
- Text4Seg: Reimagining Image Segmentation as Text Generation (ICLR 2025 · similarity 0.86)
- EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation (CVPR 2024 · similarity 0.85)
- ConDSeg: A General Medical Image Segmentation Framework via Contrast-Driven Feature Enhancement (AAAI 2025 · similarity 0.85)
- Generative Medical Segmentation (AAAI 2025 · similarity 0.85)
- Tyche: Stochastic In-Context Learning for Medical Image Segmentation (CVPR 2024 · similarity 0.85)
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.
