TABLE 4. Performance Metrics Achieved by the Top-3 Modality-Specific Knowledge Transfer Models on the Target Tasks.
| Task | Models | Acc. | AUC | Sens. | Prec. | F | MCC | Param. |
|---|---|---|---|---|---|---|---|---|
| Normal vs. COVID-19 | VGG-16 | 1 | 1 [1 1] | 1 | 1 | 1 | 1 | 19436354 |
| VGG-19 | 1 | 1 [1 1] | 1 | 1 | 1 | 1 | 24746050 | |
| Inception-V3 | 1 | 1[1 1] | 1 | 1 | 1 | 1 | 40645794 | |
| Normal vs. Bacterial vs. COVID-19 | VGG-16 | 0.9702 | 0.9977 [0.9935 1.0] | 0.9702 | 0.9709 | 0.9702 | 0.9465 | 19437379 |
| VGG-19 | 0.9563 | 0.9936 [0.9866 1.0] | 0.9563 | 0.9588 | 0.9563 | 0.9226 | 24747075 | |
| Inception-V3 | 0.9742 | 0.9969 [0.992 1.0] | 0.9742 | 0.9746 | 0.9742 | 0.9534 | 40646819 |
*Bold values stand for the model with a statistically significant better performance than the other models.