Research map / Detection and segmentation / Open-set and few-shot detection
Alleviating Textual Reliance in Medical Language-guided Segmentation via Prototype-driven Semantic Approximation
Shuchang Ye, Usman Naseem, Mingyuan Meng, Jinman Kim
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 |
|---|---|---|---|
| use ProLearn prototype approximation instead of SGSeg | language-guided segmentation · MosMedData+ 1% text | Dice: 0.3452 -> 0.7218 | Improved |
| use ProLearn instead of LViT | language-guided segmentation · MosMedData+ 1% text | Dice: 0.1677 -> 0.7218 | Improved |
| use ProLearn instead of GuideSeg | language-guided segmentation · Kvasir-SEG 1% text | Dice: 0.5615 -> 0.8718 | Improved |
| use ProLearn instead of SGSeg | language-guided segmentation · QaTa-COV19 50% text | Dice: 0.8641 -> 0.8667 | Improved |
| reduce paired text from 50% to 1% for ProLearn | ProLearn · QaTa-COV19 | Dice: 0.8667 -> 0.8566 | No clear effect |
| reduce paired text from 50% to 1% for SGSeg | SGSeg · MosMedData+ | Dice: 0.7455 -> 0.3452 | Hurt |
| reduce paired text from 50% to 1% for ProLearn | ProLearn · MosMedData+ | Dice: 0.7539 -> 0.7218 | Hurt |
| use ProLearn with 1% reports instead of Swin U-Net image-only | image-only segmentation · Kvasir-SEG | Dice: 0.398 -> 0.872 | Improved |
| use ProLearn with 1% reports instead of BiomedParse | image-only segmentation · Kvasir-SEG | Dice: 0.828 -> 0.872 | Improved |
| use prototype retrieval instead of LLM text generation (GPT-2) | SGSeg-style text compensation · inference time | time: 136ms -> 4ms | Improved |
| use prototype retrieval instead of Llama3 | SGSeg-style text compensation · inference time | time: 1.2s -> 4ms | Improved |
| increase candidates k beyond a threshold (ablation) | ProLearn · Dice sensitivity | Dice | Hurt |
Closest papers by meaning
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.87)
- Bringing CLIP to the Clinic: Dynamic Soft Labels and Negation-Aware Learning for Medical Analysis (CVPR 2025 · similarity 0.86)
- SGTC: Semantic-Guided Triplet Co-training for Sparsely Annotated Semi-Supervised Medical Image Segmentation (AAAI 2025 · similarity 0.86)
- Enhancing SAM with Efficient Prompting and Preference Optimization for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.86)
- RADAR: Text-Guided Medical Image Segmentation via Residual Aggregation and Dense Alignment Representations (NeurIPS 2026 · similarity 0.86)
- Training Like a Medical Resident: Context-Prior Learning Toward Universal Medical Image Segmentation (CVPR 2024 · similarity 0.86)
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.
