# Deploying Machine and Deep Learning Models for Efficient Data-Augmented Detection of COVID-19 Infections

^{1}

^{2}

^{3}

^{4}

^{5}

^{6}

^{7}

^{8}

^{9}

^{10}

^{11}

^{12}

^{*}

## Abstract

**:**

## 1. Introduction

- Manual extraction and selection of features.
- Poor performance when dealing with imbalanced datasets.
- Over-fitting.
- Complexity and time consumption.

## 2. Data Augmentation and Efficient Learning by DLMs

#### 2.1. Data Augmentation Process Based on Image Transformations

#### 2.2. Convolutional GAN

#### 2.3. Technical Specifications of DLMs

#### 2.3.1. ConvLSTM

#### 2.3.2. The Convolutional Layer

#### 2.3.3. Pooling Layer

#### 2.4. Classification Network

#### 2.5. Network Training

## 3. Experimental Validation

#### 3.1. Evaluation Metrics

#### 3.1.1. Accuracy of Detection (ACCD)

#### 3.1.2. Logarithmic Loss

_{ab}indicates if the sample belongs to class a or b or not and ${p}_{ab}$ denotes the probability that sample (a) belongs to class (b). Log loss values closer to zero indicate a higher level of accuracy.

#### 3.1.3. Receiver Operating Characteristic (ROC) Curve

#### 3.1.4. Testing the Execution Time

#### 3.2. Simulation Results

#### 3.2.1. COVID-19 Detection Based on Traditional Machine Learning Techniques

#### 3.2.2. Scenario 1: DLMs without Data Augmentation

#### 3.2.3. Scenario 2: DLMs with Prior Data Augmentation Using Image Transformations (DADLM)

#### 3.2.4. Analysis of the Impact of the Proposed DADLM on the Second Dataset

_{2}DADLM). On the other hand, it oscillates between 60% and 99% for the validation, i.e., testing, curves for the ConvLSTM DADLM (i.e., ConvLSTM

_{2}DADLM). Similarly, marginal variations are observed for the training set. Therefore, it can be surmised that the tracking of the training and validation is reliable since stability is observed in both curves.

_{2}DADLM degrades from 0.5 to zero, while that for the ConvLSTM

_{2}DADLM (i.e., in Figure 21b) presents a more complicated outcome.

_{2}and ConLSTM

_{2}. Specifically, from Figure 23a,b, we see that both DADLMs have accuracy reaching 99%. These outcomes are better than those reported for Scenarios 1 and 2 using the previous dataset. Furthermore, unlike the results reported for the first dataset, here, the curves support the deduction that both the training and validation tests are more stable and reliable. The stability is traced to the distended dataset size arising from our proposed data augmentation. These findings validate the earlier claims regarding the impact of data augmentation on enhancing the accuracy of image-based detection of COVID-19. Additionally, Figure 24 and Figure 25 present the confusion matrices and ROC curves for the CNN

_{2}and ConvLSTM

_{2}DADLMs, respectively. Like the previous results, the baseline of 99% accuracy is reported throughout.

#### 3.2.5. DLMs with Prior Data Augmentation Based on CGAN (CGAN DADLM)

_{1}CGAN DADLM) are presented in the curves in Figure 24a and b, respectively. Similarly, Figure 25a and b present the curves for accuracy and logarithmic loss for both training and validation phases of the proposed ConvLSTM with GAN data augmentation (i.e., ConvLSTM

_{1}CGAN DADLM). In addition, Figure 26 presents the confusion matrix and ROC curve for the proposed CNN

_{1}CGAN DADLM, while Figure 27 presents similar confusion matrix and ROC curve for the proposed ConvLSTM

_{1}CGAN DADLM. From these results, we see that the proposed CNN

_{1}CGAN DLM records a testing accuracy of 99%, while the proposed ConvLSTM

_{1}CGAN DADLM achieved an accuracy of 96%.

_{2}CGAN DADLM, while Figure 29a,b presents similar curves for accuracy and logarithmic loss for both training and validation phases for the proposed ConvLSTM

_{2}CGAN DADLM. Additionally, Figure 30 presents confusion matrix and ROC curve for the proposed CNN

_{2}CGAN DADLM, while Figure 31 presents the confusion matrix and ROC curve for the proposed ConvLSTM

_{2}CGAN DADLM. From these results, we deduce that the proposed CNN

_{2}CGAN DADLM records a testing accuracy of 83%, while an accuracy of 81% is reported for the proposed ConvLSTM

_{2}CGAN DADLM.

## 4. Discussion

_{1}DADLM and CNN

_{1}CGAN DADLM presented similar results across all six metrics, the CNN

_{1}DADLM has a slight edge. Similarly, for the CT images in the second dataset, head to head evaluation shows that the CNN

_{2}DADLM presents the best performance in terms of specificity, PPV, and MCC. Hence, for both datasets, the image-based transformation focused data augmentation DLM edges out all the other models. Metrics such as those presented here provide the conclusions regarding image-based contributions to support doctors and other specialists in making efficient diagnosis of COVID-19.

_{1}DLM and ConvLSTM

_{1}DLM) fall short of the accuracy reported the image transformations-based versions of both DLMs, i.e., CNN

_{1}DADLM and ConvLSTM

_{1}DADLM, as well as their CGAN data-augmented versions, i.e., CNN

_{1}CGAN DADLM and ConvLSTM

_{1}CGAN DADLM.

_{1}DADLM and CNN

_{1}CGAN DADLM) report best performance with 99% detection accuracy for COVID-19 positive (i.e., infected) and negative (i.e., normal) classes.

_{2}CGAN DADLM and ConvLSTM

_{2}CGAN DADLM) is noteworthy. The difference of 16 and 18% between the CGAN and image transformations-based data augmentation DADLMs is remarkable. A plausible explanation would be, as with most DLMs, there are losses associated at each layer of the CGAN used in our CGAN DADLM. This causes distortions to subsequent layers which accumulate recursively. In comparison, the pixel-wise, image-based transformations used in our image transformation-based data-augmented DLM, i.e., DADLM, is immune to these types of losses.

_{1}DADLM, ConvLSTM

_{1}DADLM, CNN

_{1}CGAN DADLM, and ConvLSTM

_{1}CGAN DADLM.

_{1}DADLM and CNN

_{1}CGAN DADLM) outperform most of the methods reported.

## 5. Concluding Remarks

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Chen, H.; Guo, J.; Wang, C.; Luo, F.; Yu, X.; Zhang, W.; Li, J.; Zhao, D.; Xu, D.; Gong, Q.; et al. Clinical characteristics and intrauterine vertical transmission potential of COVID-19 infection in nine pregnant women: A retrospective review of medical records. Lancet
**2020**, 395, 809–815. [Google Scholar] [CrossRef] [Green Version] - Zhou, F.; Yu, T.; Du, R.; Fan, G.; Liu, Y.; Liu, Z.; Guan, L. Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: A retrospective cohort study. Lancet
**2020**, 395, 1054–1062. [Google Scholar] [CrossRef] - Ye, Q.; Wang, B.; Mao, J. Cytokine storm in COVID-19 and treatment. J. Infect.
**2020**, 80, 607–613. [Google Scholar] [CrossRef] [PubMed] - Berger, A.; Drosten, C.; Doerr, H.W.; Stürmer, M.; Preiser, W. Severe acute respiratory syndrome (SARS)—Paradigm of an emerging viral infection. J. Clin. Virol.
**2004**, 29, 13–22. [Google Scholar] [CrossRef] - Hemida, M.G.; Elmoslemany, A.; Al-Hizab, F.; Alnaeem, A.; Almathen, F.; Faye, B.; Peiris, M. Dromedary camels and the transmission of Middle East respiratory syndrome coronavirus (MERS-CoV). Transbound. Emerg. Dis.
**2017**, 64, 344–353. [Google Scholar] [CrossRef] [Green Version] - Huang, C.; Wang, Y.; Li, X.; Ren, L.; Zhao, J.; Hu, Y.; Zhang, L.; Fan, G.; Xu, J.; Gu, X.; et al. Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet
**2020**, 395, 497–506. [Google Scholar] [CrossRef] [Green Version] - Guo, Y.R.; Cao, Q.D.; Hong, Z.S.; Tan, Y.Y.; Chen, S.D.; Jin, H.J.; Tan, K.S.; Wang, D.-J.; Yan, Y. The origin, transmission and clinical therapies on coronavirus disease 2019 (COVID-19) outbreak–an update on the status. Mil. Med. Res.
**2020**, 7, 1–10. [Google Scholar] [CrossRef] [Green Version] - Ather, A.; Patel, B.; Ruparel, N.B.; Diogenes, A.; Hargreaves, K.M. Coronavirus disease 19 (COVID-19): Implications for clinical dental care. J. Endod.
**2020**, 46, 584–595. [Google Scholar] [CrossRef] - Gabutti, G.; d’Anchera, E.; Sandri, F.; Savio, M.; Stefanati, A. Coronavirus: Update related to the current outbreak of COVID-19. Infect. Dis. Ther.
**2020**, 9, 1–13. [Google Scholar] [CrossRef] - Bourouiba, L. Turbulent gas clouds and respiratory pathogen emissions: Potential implications for reducing transmission of COVID-19. Jama
**2020**, 323, 1837–1838. [Google Scholar] [CrossRef] - Manolis, A.S.; Manolis, T.A. Cardiovascular Complications of the Coronavirus (COVID-19) Infection: COVID-19 and the Heart. Rhythmos
**2020**, 15, 23–28. [Google Scholar] - McCloskey, B.; Dar, O.; Zumla, A.; Heymann, D.L. Emerging infectious diseases, and pandemic potential: Status quo and reducing risk of global spread. Lancet Infect. Dis.
**2014**, 14, 1001–1010. [Google Scholar] [CrossRef] - Hammad, M.; Maher, A.; Wang, K.; Jiang, F.; Amrani, M. Detection of abnormal heart conditions based on characteristics of ECG signals. Measurement
**2018**, 125, 634–644. [Google Scholar] [CrossRef] - Alghamdi, A.S.; Polat, K.; Alghoson, A.; Alshdadi, A.A.; Abd El-Latif, A.A. A novel blood pressure estimation method based on the classification of oscillometric waveforms using machine-learning methods. Appl. Acoust.
**2020**, 164, 107279. [Google Scholar] [CrossRef] - Khalil, H.; El-Hag, N.; Sedik, A.; El-Shafie, W.; Mohamed, A.E.N.; Khalaf, A.A.; El-Fishawy, A.S. Classification of Diabetic Retinopathy types based on Convolution Neural Network (CNN). Menoufia J. Electr. Eng. Res.
**2019**, 28, 126–153. [Google Scholar] [CrossRef] - Haggag, N.T.; Sedik, A.; Elbanby, G.M.; El-Fishawy, A.S.; Khalaf, A.A. Classification of Corneal Pattern Based on Convolutional LSTM Neural Network. Menoufia J. Electr. Eng. Res.
**2019**, 28, 158–162. [Google Scholar] [CrossRef] - Alghamdi, A.; Hammad, M.; Ugail, H.; Abdel-Raheem, A.; Muhammad, K.; Khalifa, H.S.; El-Latif, A.A.A. Detection of Myocardial Infarction Based on Novel Deep Transfer Learning Methods for Urban Healthcare in Smart Cities. Multimed Tools Appl.
**2020**, 1–22. [Google Scholar] [CrossRef] [Green Version] - Amrani, M.; Hammad, M.; Jiang, F.; Wang, K.; Amrani, A. Very deep feature extraction and fusion for arrhythmias detection. Neural Comput. Appl.
**2018**, 30, 2047–2057. [Google Scholar] [CrossRef] - Saraiva, A.A.; Ferreira, N.; Lopes de Sousa, L.; Costa, N.; Sousa, J.; Santos, D.; Valente, A.; Soares, S. Classification of images of childhood pneumonia using convolutional neural networks. In Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2019), Prague, Czech Republic, 22–24 February 2019; pp. 112–119. [Google Scholar]
- Sun, Y.; Lo, F.P.W.; Lo, B. EEG-based user identification system using 1D-convolutional long short-term memory neural networks. Expert Syst. Appl.
**2019**, 125, 259–267. [Google Scholar] [CrossRef] - Jin, Y.; Qin, C.; Huang, Y.; Zhao, W.; Liu, C. Multi-domain modeling of atrial fibrillation detection with twin attentional convolutional long short-term memory neural networks. Knowl. Based Syst.
**2020**, 193, 105460. [Google Scholar] [CrossRef] - Oktay, A.B.; Kocer, A. Differential diagnosis of Parkinson and essential tremor with convolutional LSTM networks. Biomed. Signal Process. Control
**2020**, 56, 101683. [Google Scholar] [CrossRef] - Langr, J.; Bok, V. GANs in Action; Manning Publishing: New York, NY, USA, 2019. [Google Scholar]
- Weng, S.F.; Vaz, L.; Qureshi, N.; Kai, J. Prediction of premature all-cause mortality: A prospective general population cohort study comparing machine-learning and standard epidemiological approaches. PLoS ONE
**2019**, 14, e0214365. [Google Scholar] [CrossRef] [PubMed] - Molaei, S.; Zare, H.; Veisi, H. Deep learning approach on information diffusion in heterogeneous networks. Knowl. Based Syst.
**2020**, 189, 105153. [Google Scholar] [CrossRef] [Green Version] - Goodfellow, I.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Bengio, Y. Generative Adversarial Nets. In Advances in Neural Information Processing Systems; Curran: Norwich, UK, 2014; pp. 2672–2680. [Google Scholar]
- Covid-Chestxray-Dataset. 2020. Available online: https://github.com/ieee8023/covid-chestxray-dataset (accessed on 15 April 2020).
- COVID-CT-Dataset: A CT Scan Dataset About COVID-19. 2020. Available online: https://github.com/UCSD-AI4H/COVID-CT (accessed on 15 April 2020).
- Waheed, A.; Goyal, M.; Gupta, D.; Khanna, A.; Al-Turjman, F.; Pinheiro, P.R. CovidGAN: Data augmentation using auxiliary classifier GAN for improved Covid-19 detection. IEEE Access
**2020**, 8, 91916–91923. [Google Scholar] [CrossRef] - Wu, H.; Xie, P.; Zhang, H.; Li, D.; Cheng, M. Predict pneumonia with chest X-ray images based on convolutional deep neural learning networks. J. Intell. Fuzzy Syst.
**2020**, 1–15. [Google Scholar] [CrossRef] - Hochreiter, S.; Schmidhuber, J. Long Short-Term Memory. Neural Comput.
**1997**, 9, 1735–1780. [Google Scholar] [CrossRef] [PubMed] - Sutskever, I.; Vinyals, O.; Le, Q.V. Sequence to Sequence Learning with Neural Networks. In Advances in Neural Information Processing Systems; Curran: Norwich, UK, 2014; pp. 3104–3112. [Google Scholar]
- Shi, X.; Chen, Z.; Wang, H.; Yeung, D.Y.; Wong, W.K.; Woo, W.C. Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting. In Advances in Neural Information Processing Systems; Curran: Norwich, UK, 2015; Volume 9199, pp. 802–810. [Google Scholar]
- Refregier, P.; Javidi, B. Optical image encryption based on input plane and Fourier plane random encoding. Opt. Lett.
**1995**, 20, 767–769. [Google Scholar] [CrossRef] - Tato, A.; Nkambou, R. Improving adam optimizer. In Proceedings of the 6th International Conference on Learning Representations (ICLR 2018 Workshop Track), Vancouver, BC, Canada, 30 April–3 May 2018. [Google Scholar]
- Courtade, T.A.; Weissman, T. Multiterminal source coding under logarithmic loss. IEEE Transact. Inf. Theory
**2013**, 60, 740–761. [Google Scholar] [CrossRef] [Green Version] - Hanley, J.A.; McNeil, B.J. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology
**1982**, 143, 29–36. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Kalayeh, M.M.; Idrees, H.; Shah, M. NMF-KNN: Image annotation using weighted multi-view non-negative matrix factorization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Columbus, OH, USA, 24–27 January 2014; pp. 184–191. [Google Scholar]
- Beygelzimer, A.; Kakade, S.; Langford, J. Cover trees for nearest neighbor. In Proceedings of the 23rd International Conference on Machine Learning, Pittsburgh, PA, USA, 25–29 June 2006; pp. 97–104. [Google Scholar]
- Chen, Y.; Zhou, L.; Bouguila, N.; Zhong, B.; Wu, F.; Lei, Z.; Li, H. Semi-convex hull tree: Fast nearest neighbor queries for large scale data on GPUs. In Proceedings of the IEEE International Conference on Data Mining (ICDM 2018), Singapore, Singapore, 7–20 November 2018; pp. 911–916. [Google Scholar]
- Kalbkhani, H.; Shayesteh, M.G.; Zali-Vargahan, B. Robust algorithm for brain magnetic resonance image (MRI) classification based on GARCH variances series. Biomed. Signal Process. Control
**2013**, 8, 909–919. [Google Scholar] [CrossRef] - Jain, S.; Salau, A.O. An image feature selection approach for dimensionality reduction based on kNN and SVM for AkT proteins. Cogent Eng.
**2019**, 6, 1599537. [Google Scholar] [CrossRef] - Sadhukhan, S.; Upadhyay, N.; Chakraborty, P. Breast Cancer Diagnosis Using Image Processing and Machine Learning. In Emerging Technology in Modelling and Graphics; Springer: Singapore, 2020; pp. 113–127. [Google Scholar]
- Boughorbi, S.B. Optimal classifier for imbalanced data using Matthew’s correlation coefficient metric. PLoS ONE
**2017**, 12, e0177678. [Google Scholar] - Kermany, D.S.; Goldbaum, M.; Cai, W.; Valentim, C.C.; Liang, H.; Baxter, S.L.; Dong, J. Identifying medical diagnoses and treatable diseases by image-based deep learning. Cell
**2018**, 172, 1122–1131. [Google Scholar] [CrossRef] [PubMed] - Stephen, O.; Sain, M.; Maduh, U.J.; Jeong, D.U. An efficient deep learning approach to pneumonia classification in healthcare. J. Healthc. Eng.
**2019**, 2019, 7. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Liang, G.; Zheng, L. A transfer learning method with deep residual network for pediatric pneumonia diagnosis. Comput. Methods Prog. Biomed.
**2019**, 187, 104964. [Google Scholar] [CrossRef] - Loey, M.; Smarandache, F.; Khalifa, M.N.E. Within the Lack of Chest COVID-19 X-ray Dataset: A Novel Detection Model Based on GAN and Deep Transfer Learning. Symmetry
**2020**, 12, 651. [Google Scholar] [CrossRef] [Green Version] - Khalifa, N.E.; Taha, M.H.; Hassanien, A.E.; Elghamrawy, S. Detection of Coronavirus (COVID-19) Associated Pneumonia based on Generative Adversarial Networks and a Fine-Tuned Deep Transfer Learning Model using Chest X-ray Dataset. arXiv
**2020**, arXiv:2004.01184. [Google Scholar]

**Figure 1.**Architecture showing data flow in the proposed image transformation deep learning models (DLMs).

**Figure 3.**Outline of data flow convolutional generative adversarial network (CGAN) data augmentation.

**Figure 7.**Technical specifications of the proposed deep learning models: (

**a**) CNN-based DLM and (

**b**) ConvLSTM-based DLM.

**Figure 8.**Visualisation of performance indicators for ROC curves. (

**a**) Low performance. (

**b**) High performance.

**Figure 10.**Outcomes for SVM machine learning technique showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 11.**Outcomes for k-NN machine learning technique showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 12.**Outcomes for CNN

_{1}DLM without data augmentation for the first dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 13.**Outcomes for ConvLSTM

_{1}DLM without data augmentation for the first dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 14.**Outcomes for CNN

_{1}DLM without data augmentation for the first dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 15.**Outcomes for ConvLSTM

_{1}DLM without data augmentation for the first dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 16.**Outcomes for CNN

_{1}DADLM with prior data augmentation for the first dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 17.**Outcomes for ConvLSTM

_{1}DADLM with prior data augmentation for the first dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 18.**Outcomes for CNN

_{1}DADLM with prior data augmentation for the first dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 19.**Outcomes for ConvLSTM

_{1}DADLM with prior data augmentation for the first dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 20.**Accuracy for DADLMs with prior data augmentation for the second dataset. (

**a**) CNN

_{2}DADLM and (

**b**) ConvLSTM

_{2}DADLM.

**Figure 21.**Logarithmic loss for DADLMs with prior data augmentation for the second dataset. (

**a**) CNN

_{2}DADLM and (

**b**) ConvLSTM

_{2}DADLM.

**Figure 22.**Outcomes for CNN

_{2}DADLM with prior data augmentation for the second dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 23.**Outcomes for ConvLSTM

_{2}DADLM with prior data augmentation for the second dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 24.**Outcomes for CNN

_{1}CGAN DADLM with prior GAN data augmentation for the first dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 25.**Outcomes for ConvLSTM

_{1}CGAN DADLM with prior GAN data augmentation for the first dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 26.**Outcomes for CNN

_{1}CGAN DADLM with prior data augmentation for the first dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 27.**Outcomes for ConvLSTM

_{1}CGAN DADLM with prior data augmentation for the first dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 28.**Outcomes for CNN

_{1}CGAN DADLM with prior GAN data augmentation for the second dataset showing (

**a**) Accuracy and (

**b**) Logarithmic loss.

**Figure 29.**Outcomes for ConvLSTM

_{1}CGAN DADLM with prior CGAN data augmentation for the second dataset showing (

**a**) Accuracy (

**b**) Logarithmic loss.

**Figure 30.**Outcomes for CNN

_{1}CGAN DADLM with prior CGAN data augmentation for the second dataset showing (

**a**) Confusion matrix and (

**b**) ROC curve.

**Figure 31.**Outcomes for ConvLSTM

_{1}CGAN DADLM with prior CGAN data augmentation for the second dataset showing (a) Confusion matrix and (

**b**) ROC curve.

Epochs | CNN | ConvLSTM | ||
---|---|---|---|---|

Accuracy | Loss | Accuracy | Loss | |

10 | 72 | 2.623 | 65.3 | 1.9612 |

20 | 79 | 1.664 | 67.8 | 2.463 |

30 | 79.33 | 1.5187 | 79.3 | 1.159 |

40 | 73.4 | 2.919 | 73.5 | 2.205 |

50 | 0.553 | 87.5 | 87.5 | 0.553 |

60 | 74.4 | 3.035 | 75.2 | 1.396 |

70 | 87.2 | 0.9277 | 78.5 | 0.997 |

80 | 69.7 | 3.559 | 72 | 2.807 |

90 | 72.3 | 2.779 | 72.3 | 2.479 |

100 | 82.5 | 1.639 | 82.5 | 1.639 |

110 | 73.9 | 3.366 | 72.8 | 3.126 |

120 | 76.5 | 2.563 | 76.5 | 2.563 |

130 | 77.5 | 2.289 | 87.5 | 1.047 |

140 | 81 | 1.958 | 89 | 0.992 |

150 | 91 | 0.559 | 91 | 0.541 |

**Table 2.**Outcomes of binary classification quality metrics for proposed DADLMs and traditional ML techniques (all results expressed as percentages).

Model | Sensitivity | Specificity | PPV | NPV | F1-Score | MCC |
---|---|---|---|---|---|---|

CNN_{1} DLM | 98.7 | 83.6 | 85.7 | 98.5 | 91.7 | 83.2 |

ConvLSTM_{1} DLM | 98.4 | 83.3 | 85.7 | 98.0 | 91.6 | 82.7 |

CNN1 DADLM | 99.7 | 98.7 | 98.7 | 99.7 | 99.0 | 98.4 |

ConvLSTM_{1} DADLM | 100 | 90.1 | 91.0 | 100 | 95.3 | 90.6 |

CNN_{2} DADLM | 99.7 | 98.7 | 98.7 | 99.7 | 99.0 | 98.4 |

ConvLSTM_{2} DADLM | 99.6 | 98.6 | 98.6 | 99.6 | 99.0 | 98.1 |

SVM | 95.5 | 80.7 | 83.1 | 94.7 | 88.8 | 76.9 |

k-NN | 95.5 | 74.3 | 79.5 | 94.0 | 86.7 | 71.6 |

CNN_{1} CGAN DADLM | 100 | 97.8 | 97.7 | 100 | 99.0 | 97.7 |

ConvLSTM_{1} CGAN DADLM | 100 | 92.4 | 91.6 | 100 | 95.7 | 92.0 |

CNN_{2} CGAN DADLM | 87.5 | 80.0 | 75.0 | 90.0 | 80.7 | 66.4 |

ConvLSTM_{2} CGAN DADLM | 87.1 | 74.1 | 79.4 | 83.3 | 83.1 | 61.9 |

**Table 3.**Comparison of average detection accuracy from proposed models alongside those from traditional and recent methods.

Model | Year | Implementation | Description | Accuracy (%) |
---|---|---|---|---|

Proposed | 2020 | X-ray and CT images for COVID-19 detection | CNN_{1} DADLM | 99.0 |

ConvLSTM_{1} DADLM | 95.0 | |||

CNN_{1} CGAN DADLM | 99.0 | |||

ConvLSTM_{1} CGAN DADLM | 96.0 | |||

[45] | 2018 | X-ray images (pneumonia) | CNN | 92.8 |

[46] | 2019 | X-ray images (pneumonia) | CNN + 2 Dense Layers + Augmentation | 93.7 |

[19] | 2019 | X-ray images (pneumonia) | CNN + 3 Dense Layers | 95.3 |

[47] | 2019 | X-ray images (pneumonia) | CNN + 2 Dense Layers | 96.7 |

[23] | 2020 | X-ray images (pneumonia) | CNN + Random Forest | 97.0 |

[29] | 2020 | X-ray images (COVID-19) | CNN + Actual data + Synthetic Augmentation | 95.0 |

[48] | 2020 | X-ray images (COVID-19) | Alexnet + GAN data augmentation | 80.6 |

Googlenet + GAN data augmentation | 85.2 | |||

Resnet18 + GAN data augmentation | 100 | |||

[49] | 2020 | X-ray images (pneumonia) | Alexnet + GAN data augmentation | 96.1 |

Squeeznet + GAN data augmentation | 97.8 | |||

Google + GAN data augmentation | 96.8 | |||

Resnet18 + GAN data augmentation | 99.0 |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Sedik, A.; Iliyasu, A.M.; Abd El-Rahiem, B.; Abdel Samea, M.E.; Abdel-Raheem, A.; Hammad, M.; Peng, J.; Abd El-Samie, F.E.; Abd El-Latif, A.A.
Deploying Machine and Deep Learning Models for Efficient Data-Augmented Detection of COVID-19 Infections. *Viruses* **2020**, *12*, 769.
https://doi.org/10.3390/v12070769

**AMA Style**

Sedik A, Iliyasu AM, Abd El-Rahiem B, Abdel Samea ME, Abdel-Raheem A, Hammad M, Peng J, Abd El-Samie FE, Abd El-Latif AA.
Deploying Machine and Deep Learning Models for Efficient Data-Augmented Detection of COVID-19 Infections. *Viruses*. 2020; 12(7):769.
https://doi.org/10.3390/v12070769

**Chicago/Turabian Style**

Sedik, Ahmed, Abdullah M Iliyasu, Basma Abd El-Rahiem, Mohammed E. Abdel Samea, Asmaa Abdel-Raheem, Mohamed Hammad, Jialiang Peng, Fathi E. Abd El-Samie, and Ahmed A. Abd El-Latif.
2020. "Deploying Machine and Deep Learning Models for Efficient Data-Augmented Detection of COVID-19 Infections" *Viruses* 12, no. 7: 769.
https://doi.org/10.3390/v12070769