Medical Interaction Segmentation on MR-MedSeg (val)
77.7DiceMedSAM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MedSAMinput=bbox input2025.11 | 77.7 | 67.1 | 81 | |
| IMIS-Netinput=bbox input2025.11 | 77 | 67.5 | 87.2 | |
| MediRound + JCMIndependent Model (T)=true2025.11 | 58.4 | 49 | 58.9 | |
| MediRoundIndependent Model (T)=true2025.11 | 55.8 | 46.5 | 55.8 | |
| MediRound + READIndependent Model (T)=true2025.11 | 53.9 | 45.1 | 54.6 | |
| Human-Thinking + MediSeeIndependent Model (T)=false2025.11 | 43 | 33.5 | 35.4 | |
| Qwen3-VL + MediSeeIndependent Model (T)=false2025.11 | 42.7 | 34 | 34.5 | |
| SegLLM-13BIndependent Model (T)=true2025.11 | 38.9 | 31.2 | 39.1 | |
| GPT-4o + MediSeeIndependent Model (T)=false2025.11 | 36.8 | 28.7 | 29.4 | |
| SegLLM-7BIndependent Model (T)=true2025.11 | 36.3 | 27.5 | 36.8 | |
| Human-Thinking + MedPLIBIndependent Model (T)=false2025.11 | 33.7 | 25.3 | 24.5 | |
| Gemini-2.5-Pro + MediSeeIndependent Model (T)=false2025.11 | 33.5 | 26.1 | 27.6 | |
| Human-Thinking + IMIS-NetIndependent Model (T)=false2025.11 | 30.5 | 25.9 | 32.2 | |
| Gemini-2.5-Pro + IMIS-NetIndependent Model (T)=false2025.11 | 26.9 | 22.6 | 28 | |
| Qwen3-VL + IMIS-NetIndependent Model (T)=false2025.11 | 26.6 | 22.8 | 26 | |
| Human-Thinking + BiomedParseIndependent Model (T)=false2025.11 | 21.1 | 17 | 12.1 | |
| GPT-4o + IMIS-NetIndependent Model (T)=false2025.11 | 19.3 | 15.8 | 19.6 |