Table 2.
Diagnosis ML and DL algorithms based on CT scans for COVID-19 patients.
| Ref. | Year | Model | Task | Dataset | Evaluation Metrics | ||
|---|---|---|---|---|---|---|---|
| ACC | P | SN | |||||
| [40] | March 2020 | 3D CNN model | Using CT chest images infiltrative biomarkers | 498 CT scans from 151 positive COVID_19 subjects and 497 CT scans from different subjects with various types of pneumonia | 70.02 | - | - |
| [22] | June 2020 | Desenet201 pre-trained model with CNN | Object detection, binary classification | 1260 COVID-19 images and 1232 CT from health patients | 96.21 | 96.20 | 96.20 |
| [28] | June 2020 | CNN Model | Binary classification | 413 of COVID-19 images and 439 of health images | 93.01 | 95.18 | 91.45 |
| [24] | May 2020 | 3D CNN model | Multiclass classification | 219 CT scans from COVID-19 patients, 220 from IAVP and 174 from healthy people | 83.90 | 81.30 | 86.70 |
| [29] | March 2020 | Segmentation models (V-Net, U-Net, FCN) and classification models (ResNet, inception) | Detection | 732 COVID chest CT scan (400 from normal cases and 332 from COVID_19 cases | 92.22 | - | 97.21 |
| [31] | May 2019 | CNN model | Multiclass classification | 10,000 CT images related to four classes, including COVID-19, non-viral pneumonia, influenzas, and non-pneumonia | - | 95.75 | 90.11 |
| [35] | March 2020 | ResNet-50 model | Multiclass classification | 60,457 CT chest scan images were collected from 100 COVID-19 cases, 102 non-COVID-19 viral pneumonia, and 200 normal lungs. | 98.81 | 98.20 | 94.52 |
| [36] | June 2020 | DenseNet121 model | COVID-19 prognostic tool | 4106 CT images (925 COVID-19, 342 pneumonia) | 78.33 | 76.61 | 80.39 |
| [37] | March 2020 | Hybrid classification technique (CNN and ML) | Predicting the recurrences in both SARS and COVID-19 cases | 51 SARS and COVID-19 CT chest scans from the Kaggle benchmark dataset. | 96.20 | 96.12 | 96.77 |
| [41] | March 2020 | Segmentation techniques (SegNet, DRUNET) and ResNet classification model | Multiclass classification | 3000 CT images of COVID_19 and pneumonia then testing on external data | - | 94.33 | 91.22 |
| [23] | June 2020 | 3D CNN model | Object detection and binary classification | 618 CT images (219 images from 110 COVID-19 patients with mean age 50, 224 from IVAP patients with mean age 61, and 175 CT images from healthy people. | 86.60 | 86.77 | 98.21 |
| [42] | May 2020 | U-net and ResNet32 models | Examine the effect of synthetic data on COVID-19 classification | 2143 chest CTs related to 327 COVID-19-positive subjects across seven countries | 90.06 | - | - |
| [39] | March 2020 | ML (RF and SVM) and CNN models | Utilizing CT images, patient symptoms for a binary classification task | 626, negative cases 279 patients | 83.77 | 81.8 | 84.2 |
| [43] | June 2020 |
Multi-objective CNN model | Multiclass classification | 312 CT scan images in addition to patient symptoms aggregated from COVID-19 patients in 9 days | 93.40 | 91.00 | 89.00 |
| [27] | August 2020 |
CNN based on ResNet 50 model | Binary classification | 622 CT chest images from 122 for COVID-19 positive cases and 500 for normal cases | 97.95 | 97.44 | 97.31 |
| [44] | May 2020 | DL model | Classification COVID-19 from pneumonia at early stages | 219 images from 110 patients with COVID-19 (with mean age 50 years), 224 images from 224 patients with IAVP (mean age 61 years), and 175 images from 175 healthy cases (mean age 39 years) | 86.72 | 86.5 | 86.5 |
| [45] | June 2020 | ImageNet and pre-trained model (ResNet50 and ResNet100) and CNN model | Binary classification | - | 89.22 | - | 89.61 |
| [46] | April 2020 | Fully connected DL model | Binary classification | CT images from 1186 patients (132,583 CT slices). Data was divided into training, validation, and test datasets with percentage 7:2:1 | 96.21 | 95.0 | 96.21 |
| [47] | May 2020 |
Using Generative Adversarial Networks and ResNet pretrained model to classify COVID-19 images | Binary classification | 1- pneumonia dataset that includes (5863 X-ray images categorized: normal and pneumonia. 2- 624 images selected from normal and COVID-19 cases to demonstrate the effectiveness of the model |
98.77 | 9.875 | 99.21 |