Table 2.

The segmentation pipeline was applied four times with in-/excluded preprocessing and data augmentation in order to evaluate their performance influence on the model. Achieved results showing the median Dice similarity coefficient (DSC) on Lung and COVID-19 infection segmentation for each CV fold of the 5-fold cross-validation and the global average (AVG) based on the Ma et al. dataset.

	Data Augmentation:Excluded Preprocessing: Excluded		Data Augmentation: Included Preprocessing: Excluded		Data Augmentation: Excluded Preprocessing: Included		Data Augmentation: Included Preprocessing: Included
Fold	Lungs	COVID-19	Lungs	COVID-19	Lungs	COVID-19	Lungs	COVID-19
1	0.711	0.031	0.397	0.166	0.867	0.530	0.907	0.556
2	0.046	0.186	0.275	0.050	0.979	0.819	0.977	0.801
3	0.190	0.241	0.168	0.057	0.951	0.814	0.952	0.829
4	0.080	0.005	0.175	0.114	0.979	0.819	0.979	0.853
5	0.520	0.194	0.360	0.201	0.964	0.798	0.967	0.765
AVG	0.309	0.131	0.275	0.118	0.948	0.756	0.956	0.761