Classification of Hematoxylin and Eosin-Stained Breast Cancer Histology Microscopy Images Using Transfer Learning with EfficientNets

Table 7

Comparison with previous approaches using pretrained architectures for classification of the ICIAR2018 dataset.

Reference	Architecture	Stain normalization	Accuracy
Nawaz et al. [3]	AlexNet	Macenko	81.25%
Ferreria et al. [20]	Inception-ResNet-v2	None	90%
Kassani et al. [21]	VGG16	Macenko	83%
		Reinhard	87%
Kassani et al. [21]	VGG19	Macenko	80%
		Reinhard	84%
Kassani et al. [21]	Inception-ResNet-v2	Macenko	90%
		Reinhard	88%
Kassani et al. [21]	Xception	Macenko	91%
		Reinhard	94%
Kassani et al. [21]	Inception-v3	Macenko	90%
		Reinhard	90%
Golatkar et al. [50]	Inception-v3	Vahadane	85%
Vesal et al. [51]	Inception-v3	Reinhard	97.08%
Vesal et al. [51]	ResNet-50	Reinhard	96.66%
Our approach	EfficientNet-B2	Reinhard	98.33%
Our approach	EfficientNet-B2	Macenko	96.67%