|
Author | Purpose | Method | Advantages | Disadvantages | DL architecture | Results |
|
Luz et al. [144] | To provide an accurate and efficient method for COVID-19 screening with chest X-rays for memory and processing time | Using EfficientNet artificial neural networks | High accuracy | Large and heterogeneous database | The use of algorithms, or artificial neural networks, is inspired by the brain's structure and function | From the hierarchical classification, experiments were performed to evaluate the performance of the neural network in the COVID-19 data. It became possible to use data transfer techniques and reinforce data. The accuracy was 93.9 |
Liu et al. [12] | Determining areas of infection and examining the lungs with the help of a chest CT scan | Use of CT scans to evaluate COVID-19, evaluation of system performance based on DL. This experiment was performed on 249 patients | Patient availability, high accuracy | Lack of sufficient information | Classification based on DL of VB-Net neural network | A DL system was developed for the segmentation and measurement of infection areas in CT scans of patients with COVID-19. The quantitative evaluation showed high accuracy for the infected area based on POI criteria |
Li et al. [13] | Development of artificial intelligence CT imaging tools to diagnose coronavirus and isolate sick people away from healthy people | Uses powerful 2D and 3D DL models. Modifies and adapts existing AI models. This experiment was performed on 157 patients | High accuracy | Complexity | 2D and 3D DL models were used | The AI-based analysis is rapidly evolving in the diagnosis of coronavirus, and the detection is being made with great accuracy |
Wang and Wong [14] | Assist physicians in improving COVID-19 screening | Polymer reverse chain reaction screening from RT-PCR to diagnose COVID-19 | Better understanding and character analysis by physicians, accelerating the development of high-precision DL solutions | Complexity, being time-consuming | Use of artificial intelligence systems based on DL, hardening and evaluation of COVID-Net prototype using Keras DL library with TensorFlow background | Assist physicians in improving screening, use of CXR images to diagnose COVID-19 |
Ghoshal and Tucker [145] | Evaluation of prop weights-based elliptic irritable neural networks (BCNN) to improve performance for COVID-19 diagnosis | Using the transmission learning method in COVID-19 X-ray images | Improvements in the diagnosis of COVID-19 | Uncertainty in detection for radiologists | Use of DL for classified tasks, as well as chest radiographic diagnosis for COVID-19 | Estimated uncertainty with DL can warn radiologists of incorrect predictions, which increases the use of DL in diagnosing the disease |
Narin et al. [146] | Evaluation of the use of focal neural network-based methods to detect an infected patient using X-ray radiography of the chest | Use InceptionV3, ResNet50, and InceptionResNetV50 to diagnose infected patient | High-performance ResNet50 model (i) High accuracy (ii) Cost reduction | Ambiguity in matrices | Using COVID-19 X-ray images for DL models, using transition learning methods | Preliminary diagnosis of COVID-19 patients to prevent the spread of this disease in other people, using the ResNet50 model with 98% accuracy |
Hemdon et al. [20] | The implementation of a DL framework for automated COVID-19 diagnosis of X-ray images | It was performed on 50 chest X-ray photographs of 25 positive COVID-19 cases. Seven distinct architectures from deep concealer neural network models are used in COVIDX-Net | Automatic diagnosis of COVID-19 | Complexity | Use the COVID-19 classification to diagnose COVID-19 on X-ray images automatically using one of the DL frameworks | X-ray images based on the COVIDX-Net framework proposed |
Zhang et al. [147] | Detection and differentiation of viral patient from the nonviral patient | Experiment on COVIDX data including 106 COVID-19 cases | Strengthen and improve the model, more efficiency in treatment | Complex calculations | Analysis of medical images including staging detection and drawing of pathological abnormalities, X-ray image change | Detection of the abnormality by viral pneumonia screening works well on chest X-ray images The learning model is useful for predicting job failure We have the CAAD model and we have never seen such cases in COVID-19, the data had 83.61% AUC, and the sensitivity was 70.71% |
Maghdid et al. [21] | Provide artificial intelligence tools for fast and accurate detection of COVID-19, create a comprehensive set of X prototypes | Using X-rays and scanning CT images and using DL algorithms | Creating intelligent detection methods with higher efficiency (i) Increasing detection speed (ii) Increasing accuracy | Complexity | Build a DL-based detection system to detect COVID-19 pneumonia using DL algorithms | Accelerate the diagnosis of COVID-19 using the CNN model |
Zeraati et al. [148] | Automatic classification of lung diseases including COVID-19 with X-ray images | Use of advanced convolutional neural network called MobileNet, use of 3905 X-ray images of more than 6 patients | Automatic diagnosis of COVID-19 from medical images (i) Low cost (ii) High speed | Limitations | Use DL to extract large-sized features from medical images | Low-cost, fast, and automatic diagnosis of COVID-19. Different infections may be differentiated by computer and detection using features extracted by DL |
Li et al. [149] | Provide a fast and reliable way to diagnose COVID-19 | Use 1020 CT images of 108 patients with COVID-19, use of ten confidential neural networks to diagnose COVID-19 | Rapid diagnosis of COVID-19, being valid | High expenses | Use of CAD system based on DL to classify COVID-19 against other pneumonia | Using the textual CAD method on CT images to differentiate COVID-19 from other pneumococcal diseases. ResNet-101 can be used to diagnose COVID-19 |
Arora et al. [26] | Predict the number of new coronaviruses | Use LSTM-based RNN for forecasting | High accuracy of forecasting | High complexity and volume of data | Use the LSTM DL model | Two-way LSTM gives the best result and confidential LSTM gives the worst result. biLSTM gives very accurate results for short-term forecasts, such as 1 to 3 days, with less than 3% error |
Huang et al. [150] | Quantitative evaluation of changes in lung tolerance in patients with COVID-19 using CT scan with an automated DL method | CT images show the entire lung and are measured and compared by commercial DL software | Classification of different groups and better detection | Loss of initial findings | Chest CT image evaluation using DL | The lung failure rate in COVID-19 was measured using a DL instrument based on a chest CT image and there was a significant difference between different groups |
Oh et al. [151] | Use of the neural network to diagnose COVID-19 | Inspired by CXR radiographic imaging | The usefulness of this method for the diagnosis of COVID-19 and patient triage | Difficulty in training deep neural network, difficulty in collecting big data | Use of X-ray chest images to classify COVID-19 | Use artificial intelligence to improve CXR performance for detection |
Liang et al. [152] | Use of a DL model to predict disease | Inclusion of 1590 patients from 575 medical centers, Use of DL models | Early detection | Complexity of calculations | Use of an integrated Cox model called Survival Cox DL | At least 60% of the data were used for prediction. A DL model was used to predict which was efficient |
|