Skip to main content

Advertisement

Springer Nature is making SARS-CoV-2 and COVID-19 research free. View research | View latest news | Sign up for updates

TableĀ 3 Performance of the proposed CNNs with different parameters

From: Segmentation of lung parenchyma in CT images using CNN trained with the clustering algorithm generated dataset

Kernel size Kernel number Channels of normalization Output of FC Dropout probability Pooling type Batch size Epochs Learning rate Validation Favg Elapsed time
5 6 3 120 0.5 Max 128 50 0.01 0.9758 846.53
10 6 3 120 0.5 Max 128 50 0.01 0.9688 1077.55
5 3 3 120 0.5 Max 128 50 0.01 0.9609 806.078
5 6 1 120 0.5 Max 128 50 0.01 0.9657 997.13
5 6 3 240 0.5 Max 128 50 0.01 0.9765 1282.90
5 6 3 120 0.2 Max 128 50 0.01 0.9688 1593.91
5 6 3 120 0.1 Max 128 50 0.01 0.9541 1589.37
5 6 3 120 0.5 Avg 128 50 0.01 0.9682 1606.79
5 6 3 120 0.5 Max 256 50 0.01 0.9659 905.27
5 6 3 120 0.5 Max 128 80 0.01 0.9722 2660.79
5 6 3 120 0.5 Max 128 50 0.00001 0.9855 1566.26
5 6 3 120 0.5 Max 128 50 0.0001 0.9917 1609.34
  1. The kernel size and the kernel number are only related to the convolutional layer. FC represents the first fully connected layer in our proposed CNNs. The elapsed time indicates the time for training the CNN in different epochs
  2. The Italic in the first row indicates the default setting of parameters. The Italic in final row indicates the final optimized setting of parameters. The Italic in other rows indicates the modifed parameter compared with the default setting