Coronary Artery Calcification Removal on Semi-synthetic CCTA (test)
32.45PSNRDense-MAE
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Dense-MAEBackbone=Dense-Unet, Training Protocol=Self-supervised pre-training + Fine-tuning, Pre-training Dataset=20,000 healthy coronary blocks2025.12 | 32.45 | 0.912 | 0.905 | 0.54 | |
| Dense-UnetTraining Protocol=From scratch (random initialization), Pre-training=None2025.12 | 29.88 | 0.854 | 0.865 | 0.78 | |
| Standard 3D-MAEBackbone=Vision Transformer (ViT), Inductive Bias=Standard ViT without dense convolutional bias2025.12 | 28.12 | 0.815 | 0.842 | 0.92 | |
| 3D InterpolationMethod Category=Non-learning traditional method, Algorithm=Cubic spline interpolation2025.12 | 22.45 | 0.682 | 0.714 | 1.85 |