Table 3.

Summary of deep learning for medical image processing in COVID-19.

Ref.	Dataset	Methods used	Evaluation metrics	Research challenges
Ozturk et al. (2020)	Chest X-ray	CNN, DarkNet, DarkCovidNet	98.08% accuracy	Use of a limited number of COVID- 19 X-ray images
Chae et al. (2018)	OLS, ARIMA	DNN, LSTM	Average performance by 24% and 19%, respectively.	Prediction of infectious disease.
Zhou et al. (2020)	CT scan Images	logistic regression model	89.47% sensitivity, 67.42% specificity	The laboratory testing methods are not uniform among different hospitals.
Zhang, Yang, et al. (2020)	Electronic medical records	Univariate and multivariate Cox regression analysis	Hazard ratio and confidence interval was used to detect the adverse outcome	Predicting the adverse outcome at the early stage of COVID-19.
Chen, Wu, et al. (2020)	CT scan images	DL-based model	100% sensitivity, 93.55% specificity, 95.24% accuracy.	Achieving consistent results between the expert and model.
Li et al. (2020)	chest CT exams	COVNet	90% sensitivity, 96% specificity	Unable to categorize the disease into different severity levels.
Ai et al. (2020)	chest CT	DL	97% sensitivity	The clinical and laboratory data were limited when regional hospitals were overloaded.
Butt et al. (2020)	CT chest images	CNN models	0.996 AUC, 98.2% sensitivity, 92.2% specificity	Achieving fast and reliable detection of COVID-19 from chest CT datasets.
Gozes et al. (2020)	CT scan	2D Slice Analysis, 3D Volume Analysis	98.2% sensitivity, 0.996 AUC, 92.2% specificity	Challenging to achieve high accuracy in detection of Coronavirus as well as quantification and tracking of disease burden.
Wang and Wong (2020)	Chest X-ray	COVID-Net	93.3% test accuracy	AI systems leveraging the more readily available and accessible CXR imaging modality.