Table 4.

Results for the EfficientNet architectures with each stain normalization method.

EfficientNet	Stain normalization technique	Precision	Recall	F1-score	Accuracy
B0	Reinhard	93.65	93.33	93.32%	93.33%
	Macenko	91.67	91.67	91.67%	91.67%
	Nonnormalized	92.37	91.67	91.37%	91.67%
					92.9%
B1	Reinhard	94.99	95.00	94.94%	95.00%
	Macenko	94.99	95.00	94.94%	95.00%
	Nonnormalized	93.32	93.33	93.27%	93.33%
					94.7%
B2	Reinhard	98.44	98.33	98.33%	98.33%
	Macenko	96.88	96.67	96.66%	96.67%
	Nonnormalized	94.03	93.33	93.29%	93.33%
					95.3%
B3	Reinhard	92.50	91.67%	91.62%	91.67%
	Macenko	93.54	93.33	93.27%	93.33%
	Nonnormalized	91.94	91.67	91.66%	91.67%
					92.2%
B4	Reinhard	92.38	91.67	91.69%	91.67%
	Macenko	91.94	91.67	91.66%	91.67%
	Nonnormalized	88.56	88.33	88.31%	88.33%
					90.6%
B5	Reinhard	88.44	88.33	88.14%	88.33%
	Macenko	91.80	91.67	91.53%	91.67%
	Nonnormalized	92.49	91.67	91.56%	91.67%
					90.6%
B6	Reinhard	92.65	91.67	91.67%	91.67%
	Macenko	92.46	91.67	91.68%	91.67%
	Nonnormalized	93.54	93.33	93.39%	93.33%
					92.2%