Table 1.
Partial DL and AI research based on COVID-19 radiology images.
| First Author | Modality | Technique | Subject | Numbers of Cases | Nation | Sensitivity (%) |
Specificity (%) |
Accuracy (%)/AUC |
|---|---|---|---|---|---|---|---|---|
| Zhang [9] | CT | uAI | Huoshenshan Hospital | 2460 patients | China | N/A | N/A | N/N |
| Li [10] | CT | COVNet | 6 medical centers | 4352 scans/3322 patients | China | 90 | 96 | N/0.96 |
| Bai [28] | CT | EfficientNet B4 DNN | 10 medical centers | 1186 patients/132,583 slices) | multinational | 95 | 96 | 96/0.95 |
| Harmoon [13] | CT | AH-Net/Densnet-121 | multinational datasets | 2724 scans/2617 patients | multinational | N/A | N/A | N/0.949 |
| Wang [17] | CXR | DeepLabv3 | CC-CXRI/CC-CXRI-P | SYSU 145202 CC-CXRI-P 16196 |
China | 92.94 | 8704 | N/0.968 |
| Wehbe [18] | CXR | DeepCOVID-XR | multicenter | 2214 images/1192 COVID-19 | US | 75 | 93 | 83/0.90 |
| Jiao [44] | CXR | U-Net | 2 medical centers | 1834 patients | US | 73.8 | 85.3 | N/0.846 |
| Yu [34] | CT | CNN | multicenter | 421 patients | China | N/A | N/A | N/N |
| Mei [41] | CT | CNN | 18 medical centers | 905 patients | China | N/A | N/A | N/0.92 |
| Hwang [27] | CXR | Lunit INSIGHT CXR 2 | 4 medical centers | 172 CXRs/172 patients | Korea | 71.3 | 52.2 | 0.714 |
| Fung [25] | CT | 3D ResNet /LSTM | 2 medical centers | 1040 mild type/2543 patients | China | N/A | N/A | 0.92 |
| Wang [23] | CT | Full-uAI-Discover-NCP | single medical center | 31 patients | China | N/A | N/A | N/N |
| Shan [29] | CT | VB-Net | single medical center | 549 patients | China | N/A | N/A | N/N |
| Murphy [26] | CXR | CAD4COVID-XRay | 3 medical centers | 24,678 patients | the Netherlands | 85 | 61 | N/0.81 |
| Li [10] | CXR | Siamese neural network | CheXpert/1 medical center | 468 patients | US | N/A | N/A | N/0.80 |
| Lessman [31] | CT | CORADS-AI | 1 academic center/1 hospital | 843 patients | the Netherlands | 85.7 | 89.8 | N/0.95 |
| Wang [21] | CT | DenseNet121-FPN/COVID-19Net | multicenter | 5372 patients | China | 78.93 | 89.93 | N/0.90 |
| Schiaffino [43] | CT | SVM vs. MLP | 6 medical centers | 897 patients | Italy | N/A | N/A | N/0.747 vs. 0.844 |
| Bartolucci [33] | CT | 3DSlicer/RadAR | 2 medical centers | 115 patients | Italy | N/A | N/A | N/0.82 |
| Purkayastha [42] |
CT | CNN | multicenter | 981 patients | multinational | N/A | N/A | N/0.868 |
| Ma [24] | CT | Advanced decision tree based machine | single medical center | 244 patients | China | 82 | 84 | N/0.84 |
| Cai [20] | CT | RF | single medical center | 99 patients | China | N/A | N/A | N/0.917 –0.940 |
| Yue [22] | CT | logistic regression /random forest | 31 patients/ 72scans |
China | 100/75 | 100/89 | N/ 0.97/0.92 |
|
| Shiri [45] | CT | ResNet | 9 medical centers | 1141 patients | Switzerland | N/A | N/A | N/N |
| Sengupta [46] | CT | QNN | 5 open-source datasets | 9500 + patients | India | 97.7 | N/A | 96.92/ N |
| Wu [47] | CT | COVID-AL | CC-CCII | 962 patients | China | N/A | N/A | 86.6/0.968 |
| Fouladi [51] | CT | ResNet-50/VGG-16/CNN/CAENN | COVID-19 dataset | 2482 patients | Iran | N/A | N/A | 94/N |
N or N/A: not applicable.