Table 5.

Proposed DFL augmented CSDB CNN performance on the validation set for the four target classes (values are averaged across 5 folds). Also, the macro-averaged scores for each metric (computed from all classes per validation fold) are aggregated over all 5 folds.

Classes	Precision	Recall	F1-score	Accuracy	Specificity	AUC
COVID-19	98.36	96.07	97.20	99.80	99.94	98.01
Normal	99.54	97.89	98.71	98.25	99.03	98.46
Bacterial Pneumonia	95.43	98.74	97.05	98.91	98.94	98.84
Viral Pneumonia	92.03	97.46	94.66	98.92	99.08	98.27
Macro scores	96.34	97.54	96.90	97.94	99.25	98.39