Table 6.

Testing performance for a patch-based approach for multi-class BC classification: center patches (CRP), augmented CRP, and random patches (RP)

CNN model	Criterions	Year	Sensitivity	Specificity	Accuracy	AUC
CNN [17]			–	–	0.667	–
CNN + SVM[17]			0.810	–	0.650	–
IRRCNN	CRP	2018	0.868	0.8733	0.8812	0.9169
IRRCNN	CRP + Aug	2018	0.9371	0.9809	0.9711	0.9905
IRRCNN	RP	2018	0.9290	0.9752	0.9634	0.9824