Table 4 Quantitative comparison of different architectures in segmenting COVID-19 lesions. The presented results are the average of the obtained results through a 10-fold cross-validation process.
From: COVID-rate: an automated framework for segmentation of COVID-19 lesions from chest CT images
Architecture | DSC | SPC | SEN | MAE |
|---|---|---|---|---|
Standard U-Net | 0.7793 | 0.9963 | 0.7622 | 0.0059 |
U-Net++ | 0.7891 | 0.9964 | 0.8118 | 0.0061 |
Attention U-Net | 0.7911 | 0.9974 | 0.7842 | 0.0056 |
Residual U-Net with CPB | 0.7921 | 0.9968 | 0.8239 | 0.0055 |
\(\text {COVID-Rate}\) without CPB | 0.7991 | 0.9968 | 0.8296 | 0.0054 |
\(\text {COVID-Rate}\) | 0.8036 | 0.9968 | 0.8340 | 0.0053 |
Enhanced \(\text {COVID-Rate}\) | 0.8069 | 0.9969 | 0.8354 | 0.0053 |
