Table 1 Review on methods and quantitative results for the classification of COVID-19 CT-Scan Images

T. D. Pham [26]

Augmentation techniques involving random reflection, translation, and scaling

Pre-trained Convolutional Neural Network (CNN) model with weights from the ImageNet dataset.

The highest performance obtained by DenseNet201. Accuracy: 96.20%, AUC = 0.98 ± 0.03.

V. Shah et al [27]

Unprocessed

Pre-trained CNN model with ImageNet weight on CT scan images

The highest performance obtained by VGG19. Accuracy: 94.50%.

Y. Pathak et al. [28]

Unprocessed

Pre-trained CNN model with ImageNet weight

The highest performance obtained by ResNet32. Accuracy: 93.01%

R. Tiwari. et al [29]

Data augmentation, resizing images and channels slicing and stacking.

Channel based overlapping CNN tower architecture

The obtained accuracy is 99.40%, AUC = 0.99.

A. Sinha et al. [30]

Manual assessment to compromise segmentation accuracy

ML prediction model based on clinical parameters and automated CT scan features

The highest performance obtained by Random Forest Model ALLR. AUC: 0.91.

M. Subramanian, et al [31]

Image augmentation techniques based on “in-place” and “on-the-fly” methods

Learning without forgetting by leveraging transfer learning using CNN

The highest performance obtained by Wide ResNet. Accuracy: 98.12%

E.H. Lee, et al [32]

Scaling, data augmentation, applying a clipping function that truncates all Hounsfield unit intensity values above a fixed pre-determined value.

CNN that uses the entire chest CT volume applied on data in different countries

Highest accuracy: 93.2%, Highest AUC: 0.994