Multi-structure Cardiac Function Assessment on MIMIC-EchoQA
84.15Pericardium AccuracyEchoAgent
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| EchoAgent2026.04 | 84.15 | 82.58 | 81.61 | 75.26 | 80.82 | 77.42 | 70.37 | |
| Deepseek-VL22026.04 | 74.39 | 51.52 | 52.87 | 32.63 | 52.05 | 41.94 | 40.74 | |
| Qwen3-VL-Plus*Tool-augmentation=specialized Echo hands2026.04 | 73.17 | 56.82 | 56.32 | 57.37 | 69.86 | 48.39 | 51.85 | |
| GPT-5*Tool-augmentation=specialized Echo hands2026.04 | 69.51 | 60.61 | 59.77 | 47.89 | 63.01 | 41.94 | 66.67 | |
| Qwen3-VL-Plus2026.04 | 65.85 | 43.18 | 43.68 | 36.32 | 53.42 | 35.48 | 11.11 | |
| GPT-52026.04 | 60.98 | 40.91 | 36.78 | 26.32 | 36.99 | 38.71 | 44.44 | |
| Qwen2.5-7B-VL2026.04 | 50 | 18.18 | 29.89 | 36.84 | 27.4 | 25.81 | 29.63 | |
| DeepSeek-R1-Distill-Llama-8B2026.04 | 41.46 | 30.3 | 27.59 | 38.42 | 41.1 | 45.16 | 40.74 | |
| Janus-Pro-7B2026.04 | 39.02 | 26.52 | 27.59 | 40.53 | 30.14 | 51.61 | 37.04 | |
| LLaVA-Med2026.04 | 32.93 | 20.45 | 20.69 | 30.53 | 35.62 | 41.94 | 48.15 |