Research map / Vision-language and multimodal / CLIP and zero-shot
MedCLIPSeg: Probabilistic Vision-Language Adaptation for Data-Efficient and Generalizable Medical Image Segmentation
Taha Koleilat, Hojat Asgariandehkordi, Omid Nejatimanzari, Berardino Barile, Yiming Xiao, Hassan Rivaz
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 PVL adapters (ablation) | MedCLIPSeg · ID/OOD DSC | OOD DSC: 55.23 -> 79.02 | Improved |
| use probabilistic instead of deterministic attention (ablation) | MedCLIPSeg · OOD DSC | OOD DSC: 63.12 -> 79.02 | Improved |
| use probabilistic attention instead of deterministic (ablation) | MedCLIPSeg · ID DSC | ID DSC: 87.68 -> 89.11 | Improved |
| add probabilistic modeling for calibration (ablation) | MedCLIPSeg · ID/OOD Brier score | Brier score (%): 23.9 -> 11.1 (ID); 25.3 -> 11.8 (OOD) | Improved |
| add visual-side adaptation (ablation) | MedCLIPSeg · OOD DSC | OOD DSC: 64.40 -> 79.02 | Improved |
| add textual-side adaptation (ablation) | MedCLIPSeg · ID DSC | ID DSC: 88.83 -> 89.11 | No clear effect |
| add bidirectional interaction (ablation) | MedCLIPSeg · HM DSC | HM DSC: 82.85 -> 83.76 | Improved |
| add soft patch-level contrastive loss (ablation) | MedCLIPSeg · HM DSC | HM DSC: 81.84 -> 83.76 | Improved |
| use soft instead of hard contrastive targets (ablation) | MedCLIPSeg · HM DSC | HM DSC: 82.65 -> 83.76 | Improved |
| use concise informative prompts instead of underdescriptive ones (ablation) | MedCLIPSeg · HM DSC | HM DSC: 56.82 -> 83.76 | Improved |
| use UniMedCLIP backbone instead of CLIP (ablation) | MedCLIPSeg · OOD DSC | OOD DSC: 74.81 -> 79.02 | Improved |
| use beta=2.35 confidence weighting (ablation) | MedCLIPSeg · HM DSC | HM DSC | 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.
- CAT-Seg — CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024) · better on 9 of 9 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
- RADAR: Text-Guided Medical Image Segmentation via Residual Aggregation and Dense Alignment Representations (NeurIPS 2026 · similarity 0.87)
- UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation (NeurIPS 2026 · similarity 0.87)
- Bringing CLIP to the Clinic: Dynamic Soft Labels and Negation-Aware Learning for Medical Analysis (CVPR 2025 · similarity 0.87)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.87)
- BiomedCoOp: Learning to Prompt for Biomedical Vision-Language Models (CVPR 2025 · similarity 0.87)
- MedFlowSeg: Flow Matching for Medical Image Segmentation with Frequency-Aware Attention (NeurIPS 2026 · 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.
