iENE component segmentation on iENE Grade 2 n=397 (5-fold cross-validation)
83.4DSCAMO-ENE
Evaluation results
| Method | Links | |
|---|---|---|
| AMO-ENEArchitecture Type=Vit + CNN, Pretraining=Yes, Parameters=79.4 M, Evaluation Protocol=Fine-tuned2026.04 | 83.4 | |
| SwinUNETRV2Architecture Type=Vit + CNN, Pretraining=No, Parameters=72.8 M, Evaluation Protocol=Fine-tuned2026.04 | 81.1 | |
| nnUnetArchitecture Type=CNN, Pretraining=No, Parameters=101.9 M, Evaluation Protocol=Fine-tuned2026.04 | 80.3 | |
| SwinUNETRArchitecture Type=Vit, Pretraining=Yes, Parameters=62.2 M, Evaluation Protocol=Fine-tuned2026.04 | 80 | |
| SamMed3D (1pt)Architecture Type=Vit + Prompt, Pretraining=Yes, Parameters=100.0 M, Evaluation Protocol=Inference-only (not trained), Prompts=1pt2026.04 | 55.1 | |
| SamMed3D (10pts)Architecture Type=Vit + Prompt, Pretraining=Yes, Parameters=100.0 M, Evaluation Protocol=Inference-only (not trained), Prompts=10pts2026.04 | 47.8 |