Classification of benign and malignant subtypes of breast cancer histopathology imaging using hybrid CNN-LSTM based transfer learning

Table 5 Comparison of proposed hybrid CNN-LSTM with existing state-of-the-art models

Study	Classifier	Transfer learning	Accuracy (%)
Study	Classifier	Transfer learning	40×	100×	200×	400×
Binary classification
Gupta et al. [20]	DR(DenseNet-169,XGB)	ImageNet	94.71 ± 0.88	95.9 ± 4.2	96.76 ± 1.09	89.11 ± 0.12
Nahid et al. [21]	NDCNN	None	94.4	95.93	97.19	96
Wei et al. [22]	NDCNN(GoogleNet)	ImageNet	97.89	97.64	97.56	97.97
Das et al. [23]	GoogleNet	ImageNet	94.82	94.38	94.67	93.49
Han et al. [24]	NDCNN(GoogleNet)	ImageNet	95.8 ± 3.1	96.9 ± 1.9	96.7 ± 2.0	94.9 ± 2.8
Gandomkar et al. [25]	ResNet-152	ImageNet	98.6	97.9	98.3	97
Gandomkar et al. [25]	Emdt(ResNet-152)	ImageNet	98.77 (overall)
Bardou et al. [26]	Eiter(NDCNN)	None	98.33	97.12	97.85	96.15
Proposed model	CNN-RNN hybrid	ImageNet	99.03	99.75	99.64	98.07
Multi-class classification
Han et al. [24]	NDCNN(GoogleNet)	ImageNet	92.8 ± 2.1	93.9 ± 1.9	93.7 ± 2.2	92.9 ± 1.8
Gandomkar et al. [25]	ResNet-152	Im-Break	95.6	94.8	95.6	94.6
Bardou et al. [26]	Eiter(NDCNN)	None	88.23	84.64	83.31	83.98
Nawaz et al. [27]	ResNet	ImageNet	95 (overall)
Proposed model	CNN-RNN hybrid	ImageNet	96.5	92.6	88.94	92.51

ISSN: 1471-2342