Table 6.

Comparison with state-of-the-art.

Study	Dataset Size	Diseases/ Classes	Techniques Used	Average Accuracy
Bhandari et al. [28]	Public dataset with 7132 chest X-ray images	COVID-19 Pneumonia Tuberculosis No-Finding	Deep Learning and XAI	Testing: 94.31 ± 1.01% Validation: 94.54 ± 1.33%
Venkataramana et al. [29]	Public dataset with 14,693 chest X-ray images	COVID-19 Pneumonia Tuberculosis No-Finding	SMOTE and Deep Learning	95.7% without Balancing 96.6% with Balancing
Bashar et al. [32]	Public dataset with 21,165 chest X-ray images	Normal, COVID-19, Pneumonia Lung Opacity	Deep learning models	Validation: 95.63%
Nasiri et al. [34]	ChestX-ray dataset with 1125 X-ray images	Binary (COVID-19/Healthy) Ternary (COVID-19, Healthy/Pneumonia)	Deep learning models DenseNet169 MobileNet	98.54% for binary and 91.11% for ternary
Liu et al. [35]	Public dataset with chest X-ray images	Binary (no finding/pneumonia) Multivariate (COVID-19/No findings/Pneumonia)	Deep Learning, Transfer Learning models	Binary (91.5%) Multivariate (91.11%)
Proposed Technique	Public dataset with 21,581 chest X-ray images	COVID-19 Pneumonia Tuberculosis No-Finding	Deep learning models (CNN)	Validation: 98.72%