Table III.
Performance evaluation of the current literature and the proposed transfer learning model in terms of binary (COVID vs. pnemonia), ternary (normal, COVID, pnemonia), quaternary (normal, COVID, bacterial pnemonia, viral pnemonia) classification.a
| Type % | ACC | SEN | SPC | AUC |
|---|---|---|---|---|
| Binary | ||||
| Proposed | 100±1.0 | 99±2.0 | 100±0.0 | 100±0.0 |
| Zhang et al (14) | - | up to 96 | 70.6 | 95.1 |
| Narin et al (15) | 98.0 | 96.0 | 100.0 | - |
| Afshar et al (17) | 98.3 | 80.0 | 98.6 | - |
| Khalifa et al (19) | 98.7 | 98.7 | 98.7 | - |
| Apostolopoulos et al (22) | 96.7 | 98.6 | 96.46 | - |
| Chowdhury et al (37) | 98.3 | 96.7 | 100.0 | 99.8 |
| Ternary | ||||
| Proposed | 85±7.0 | 94±6 | 92.7±7.6 | 96±2.0 |
| Wang et al (16) | 92.6 | 91.3 | - | - |
| Abbas et al (18) | 95.1 | 97.9 | 91.8 | - |
| Ucar et al (21) | 98.2 | - | 99.1 | - |
| Apostolopoulos et al (22) | 94.7 | - | - | - |
| Chowdhury et al (37) | 98.3 | 96.7 | 99.0 | 99.0 |
| Quaternary | ||||
| Proposed | 76±8.0 | 93±9 | 91.8±7.6 | 93±3.0 |
aThe metrics are presented in mean ± standard deviation format, regarding the COVID-19 class for each case. The best performance is presented in bold. ACC, accuracy; SEN, sensitivity; SPC, specificity; AUC, area under the curve.