Table 3.
Benchmarking of six AI models with the existing work on COVID-19 classification
| Row# | Authors | Dataset | Model | Accuracy | Performance |
|---|---|---|---|---|---|
| R1 | Polsinelli et al. [29] | 360 CT scans of COVID-19 subjects and 397 CT scans of other kinds of illnesses | SqueezeNet | 0.83 | 0.8333 of F1 Score |
| R2 | Hasan et al. [30] | 321 chest CT scans (118-COVID, 96, pneumonia, 107 healthy) | LSTM | 1.00 | X |
| R3 | Jaiswal et al. [24] | 1262 CT COVID-19-positive CT images, 1230 CT images of non-COVID patients | DenseNet201 | 0.962 | 0.97 AUC |
| R4 | Loey et al. [31] | 345 images—COVID, 397 images—non-COVID CT scans | ResNet50 | 0.829 | Sensitivity of 77.66% and specificity of 87.62% |
| R5 | Apostolopoulos et al. [32] | 224 images—COVID-19, 714—bacterial pneumonia, 504—normal patients X-ray | MobileNet v2 | 0.967 | Sensitivity of 98.66% and specificity of 96.46% |
| R6 | Proposed Study |
2788 CoP/990 NCoP CT scans |
iCNN | 1.00 | 0.993 AUC |