Table 3.

DenseNet161 model comparison between top-ranking 2-class and equivalent 3-class variant.

Model	Hyperparameters			Validation Performance ($\mu \pm 95CI$)
Binary vs. Multiclass	InputShape	Num. Epochs	Batch Size	AUC	Accuracy	F1 Score	Sensitivity/Recall	Specificity	Precision
2-Class	128 × 128	300	32	0.9128 ± 0.0052	0.8403 ± 0.0076	0.7630 ± 0.017	0.7739 ± 0.0199	0.8735 ± 0.0128	0.7580 ± 0.0231
3-Class	128 × 128	300	32	0.7546 ± 0.0151	0.7246 ± 0.0118	0.3820 ± 0.0436	0.4148 ± 0.0250	0.8605 ± 0.0153	0.3694 ± 0.0469