Table 1.

Comparison of different approaches studied in the literature review.

S.No.	Method name	Empirical results	Strengths	Weaknesses
1	Patch-based DL approach14–17	Improved classification and segmentation accuracy	High accuracy and precision across datasets	Computational complexity, need for extensive labeled data
2	COVID-19 detection models21,33	High accuracy in detecting and segmenting COVID-19 infections	Advanced techniques like STM blocks and FME	Limited labeled data, high computational complexity
3	Graph-based and transfer learning models19,20	Effective COVID-19 detection and prediction	Utilizes GIN and transfer learning models	Dependence on large datasets for training
4	Capsule networks8,22	Superior performance with small datasets	Better handling of small datasets	Complex architecture
5	Regional feature-based prediction23	Overall accuracy of 91.66% on test data	Effective use of regional features	Limited generalizability
6	Deep feature learning with SMOTE28	Accurate COVID-19 prediction using CXR images	Improved accuracy with ResNet152 architecture	Potential overfitting
7	Ensemble and hybrid learning models30–32,34	High performance with web-based interface for rapid detection	Combines strengths of multiple models	Substantial computational resources required
8	Severity assessment models35–38	Efficient and reliable assessment of COVID-19 severity	Accurate severity computation	Complex preprocessing and segmentation steps