Table 6 Evaluation of the proposed data augmentation method’s efficacy through a two-step 10-fold cross-validation approach.
From: COVID-rate: an automated framework for segmentation of COVID-19 lesions from chest CT images
Training set | Number of synthetic images | DSC | SPC | SEN | MAE |
|---|---|---|---|---|---|
Ave ± std | Ave ± std | Ave ± std | Ave ± std | ||
Existing training set | – | \(0.7624 \pm 0.02\) | \(0.9944 \pm 0.002\) | \(0.8342 \pm 0.06\) | \(0.0079 \pm 0.001\) |
Augmented training set | \(0.5 \times N\) | \(0.7696 \pm 0.02\) | \(0.9944 \pm 0.001\) | \(0.8573 \pm 0.04\) | \(0.0077 \pm 0.001\) |
\(1 \times N\) | \(0.782 \pm 0.01\) | \(0.9948 \pm 0.001\) | \(0.8614 \pm 0.02\) | \(0.0073 \pm 0.001\) | |
\(1.5 \times N\) | \(0.7604 \pm 0.02\) | \(0.9944 \pm 0.001\) | \(0.8393 \pm 0.05\) | \(0.0079 \pm 0.001\) | |
\(2 \times N\) | \(0.752 \pm 0.05\) | \(0.9953 \pm 0.001\) | \(0.79 \pm 0.1\) | \(0.0077 \pm 0.001\) |
