Table 1.

Summary of related techniques in the literature.

Reference	Dataset	Pretrained	Architecture	Input size	Stain normalization	Image-wise accuracy
Araújo et al. [5]	Bioimaging 2015	No	Custom CNN	512 × 512	Macenko	4-class: 77.8%
						2-class: 80.6%
Vo et al. [17]	Bioimaging 2015	No	3 × Inception-ResNet-v2	600 × 600	Macenko	4-class: 96.4%
				450 × 450		2-class: 99.5%
				300 × 300
Nawaz et al. [3]	ICIAR2018	Yes	AlexNet	512 × 512	Macenko	81.25%
Ferreria et al. [20]	ICIAR2018	Yes	Inception-ResNet-v2	244 × 244	Nonnormalized	90%
Kassani et al. [21]	ICIAR2018	Yes	VGG16	512 × 512	Macenko	83%
					Reinhard	87%
Kassani et al. [21]	ICIAR2018	Yes	VGG19	512 × 512	Macenko	80%
					Reinhard	84%
Kassani et al. [21]	ICIAR2018	Yes	Inception-ResNet-v2	512 × 512	Macenko	90%
					Reinhard	88%
Kassani et al. [21]	ICIAR2018	Yes	Xception	512 × 512	Macenko	91%
					Reinhard	94%
Kassani et al. [21]	ICIAR2018	Yes	Inception-v3	512 × 512	Macenko	90%
					Reinhard	90%