Medical Image Segmentation on COVID-19 lung region segmentation
86.58Dice CoefficientAttention U-Net
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Attention U-NetAugmented=true, Key Features=Attention U-Net, data augmentation, post-processing2025.05 | 86.58 | 83.16 | 0.3888 | 9.8995 | |
| Attention U-NetAugmented=false, Key Features=Attention U-Net2025.05 | 85.02 | 74.45 | 0.3907 | 8.4853 | |
| COVID-Rate (Nastaran Enshaei et al.)Key Features=Automated segmentation framework with CT dataset2025.05 | 85 | — | — | — | |
| Jun Ma et al. (2020)Key Features=Traditional U-Net architecture2025.05 | 83.45 | 72.04 | 0.5 | 10.34 | |
| Automated 3D U-Net Segmentation (Dominik Müller et al.)Key Features=On-the-fly patch generation, 3D U-Net2025.05 | 83.2 | 72.8 | — | — | |
| COVID-CT-Net (Jinyu Zhao et al.)Key Features=Ensemble learning and transfer learning2025.05 | 82 | 74 | 0.45 | 9.5 | |
| Comparative Study (Sofie Tilborghs et al.)Key Features=Multi-center evaluation with various algorithms2025.05 | 81 | — | — | — | |
| Inf-Net (Deng-Ping Fan et al.)Key Features=Incorporation of edge-attention mechanisms2025.05 | 78 | 68 | 0.52 | 11.2 |