Table 1.

Related work on classification of COVID-19 severity or from other lung infections.

Method	Year	Data	Class Groups	Results
Statistical approach [9]	2021	Plasma and single-cell proteomic	Mild, moderate, severe	AUCtraining = 0.799 AUCvalidation = 0.773
Deep Learning [10]	2022	CXR	COVID-19 and normal	Mean absolute error of 0.30
Deep Learning [11]	2021	CXR	Severity level (scores from 0 to 8)	confusion matrix: Sensitivity = 0.94 Specificity = 0.98
Deep learning [12] ResNet	2022	CXR	COVID-19, virus, bacteria, ARDS, SARS, Streptococcus, normal	Accuracy = 98%
Deep learning [13]	2023	CXR	COVID-19, virus, bacteria, and normal	Accuracy = 97.65%
Deep learning [14] CORONA-NET	2023	CXR	COVID-19, viral pneumonia, normal	Accuracy = 99.57%
Deep learning [15] ResNet	2023	CXR	COVID-19, non-COVID-19, and normal	Accuracy = 96%
Deep learning [16] DAM-Net	2023	CXR	COVID-19, pneumonia, and normal	Accuracy = 97.22% Sensitivity = 96.87% Specificity = 99.12%
Deep learning [17]	2021	CXR	Severity descriptors vs. radiologist’s severity ratings, normal vs. abnormal	Correlation = 0.68 (p < 0.0001) AUC = 0.78
Cytokine profiles with statistical approaches [20]	2022	Cytokine concentration	Severe, moderate, and mild	AUC = 0.83
Supervised machine learning models [21]	2023	Feature-dataset consisting of the routine blood values and demographic data that affect the prognosis of COVID-19	Severely and mildly	AUC range: 0.75 to 0.95 Accuracy range: 94.05% to 97.86%
Vision Transformers [22]	2022	CXR and CT	COVID-19, pneumonia, and normal	Accuracy = 94.62%
COVID_SDNet [23]	2020	CXR	Severe, moderate, and mild	Accuracy = 81.0% ± 2.9% Sensitivity = 76.8% ± 6.3% Specificity = 85.2% ± 5.4%
Multi-task vision [18]	2022	CXR	Severity degree from 0 to 6	Accuracy range: 78.7% to 97.7%
Multi-stage framework [19]	2023	CXR	Mild, moderate, severe, and critical	Accuracy = 97.63%