Table 6.

Patch-wise comparisons of the accuracy, sensitivity, and specificity metrics for the BreakHis dataset. The magnification factor is 200×, and other abbreviations (A, F, TA, PT, DC, LC, MC, and PC) are listed in Table 3. The best results are shown in bold.

Class	Model	Accuracy		Sensitivity
				Without Data								Data
				Augmentation								Augmentation
				Benign				Malignant				Benign				Malignant
Binary	DNet	0.896	0.909	0.75				0.97				0.75				0.99
	MSI-MFNet	0.92	0.98	0.76				0.99				0.94				0.99
				A	F	PT	TA	DC	LC	MC	PC	A	F	PT	TA	DC	LC	MC	PC
Multi	DNet	0.838	0.861	0.601	0.843	0.718	0.837	0.862	0.85	0.97	0.909	0.611	0.87	0.673	0.776	0.92	0.893	0.945	0.936
	MSI-MFNet	0.88	0.87	0.595	0.87	0.79	0.89	0.96	0.745	0.98	0.92	0.62	0.861	0.78	0.82	0.890	0.90	0.98	0.98
				Specificity
				Without Data								Data
				Augmentation								Augmentation
				Benign				Malignant				Benign				Malignant
Binary	DNet			0.97				0.74				0.95				0.93
	MSI-MFNet			0.98				0.76				0.99				0.94
				A	F	PT	TA	DC	LC	MC	PC	A	F	PT	TA	DC	LC	MC	PC
Multi	DNet			0.979	0.986	0.973	0.988	0.878	0.962	0.972	0.971	0.982	0.975	0.973	0.979	0.904	0.936	0.951	0.963
	MSI-MFNet			0.99	1.0	0.99	1.0	0.90	0.98	0.98	0.99	1.0	0.99	0.99	0.99	0.93	0.96	0.97	0.98