|
| Author | Year | Data sets/samples | Sample collection location/source | Network type/technique used | Objective of study | Type of study/outcome of study |
|
| Mahmud et al. [22] | 2020 | Total of 5856 pictures | Guangzhou Medical Center, China and Sylhet Medical College, Bangladesh | CNN | To utilize COVID-19 chest X-rays for efficiently extracting diversified features from varying dilation rates | Detection accuracy: |
| 1583 normal X-rays, 1493 non-COVID, 2780 bacterial pneumonia | 97.4% for COVID-19/normal pneumonia |
| 96.9% for COVID-19/viral pneumonia |
| 94.7% for COVID-19/bacterial pneumonia |
|
| Gu et al. [23] | 2018 | JSRT, 241 images and Montgomery County, (MC, 138 images) | Guangzhou Women and Children’s Medical Center, China | FCN and DCNN | Deep learning in chest radiography for diagnosis of bacterial and viral childhood pneumonia | Experiments revealed that DCNN with transfer learning extracted features with greater accuracy (0.8048 ± 0.0202) and sensitivity (0.7755 ± 0.0296) |
| Li et al. [24] | 2020 | At six medical centers, 4,536 volumetric chest CT examinations (3D) were obtained from 3,506 individuals | 6 different hospitals in China | COVID-19 detection neural network (COVNet) | Distinguishing of COVID-19 from community-acquired pneumonia on chest CT using AI | For community-acquired pneumonia, the area under the receiver operating characteristic curve was 0.96 and 0.95, respectively |
| (From August 2016 and February 2020) |
|
| Rajpurkar et al. [25] | 2018 | 420 images from 14 different pathologies | Bethesda, Maryland, United States | CheXNeXt algorithm | To diagnosis chest radiograph using deep learning method | The 420 radiographs were labeled by radiologists in an average of 240 minutes, and the algorithm labeled them in 1.5 minutes |
|
| Chowdhury et al. [26] | 2020 | Total 423 images database (1485 pictures of viral pneumonia and 1579 pictures of normal chest X-rays) | Italian Society of Medical and Interventional Radiology, Italy | CNN | Screening of COVID-19 and pneumonia detection using AI | The networks were trained to distinguish between two types of pneumonia. For both methods, the classification accuracy, precision, sensitivity, and specificity were 99.7%, 99.7%, 99.7%, and 99.55% and 97.9%, 97.95%, 97.9%, and 98.8%, respectively |
|
| Liang and Zheng [27] | 2020 | Total 5856 chest X-ray images (training: 5232, testing: 624) | Guangzhou Women and Children’s Medical Center, China | CNN | Pediatric pneumonia diagnosis using transfer learning technique with a deep residual network | On a children’s pneumonia classification test, the method recall rate is 96.7%, and the F1 score is 92.7% |
|
| Ho and Gwak [28] | 2019 | Total of 112,120 X-ray images (70% training, | ILSVRC2014 dataset | CNN/DenseNet-121 model | CNN-based classification of thoracic disease in chest radiography | In compared to current reference baselines, techniques efficiently used interdependencies among target annotations to produce state-of-the-art classification results of 14 diseases |
| 10% validation, and 20% testing) |
|
| Roy et al. [29] | 2020 | There are 58,924 frames in 277 lung ultrasound recordings from 35 individuals | Italian COVID-19 lung ultrasound | CNN | Diagnosis of lung diseases in COVID-19 pandemic using deep learning | A novel deep network based on spatial transformer networks that predicts the illness severity with weakly supervised artefact localization |
| Database (ICLUS-DB), Italy |
|
