Table 2.

Comparative analysis of binary classification accuracy (%) with other methods on the BreaKHis dataset.

Methods	40×	100×	200×	400×
ResHist model [12]	86.38	87.28	91.35	86.29
IRRCNN w/o augmentation [13]	97.16	96.84	96.61	95.78
IRRCNN with augmentation [13]	97.95	97.57	97.32	97.36
Alex Net [22]	85.6	83.5	82.7	80.7
class structure-based deep CNN [21]	92.8	93.9	93.4	92.9
Multi task CNN [40]	81.87	83.39	82.56	80.69
CNN & Fusion Rules [41]	90.0	88.4	84.6	86.1
VLAD encoding [42]	91.8	92.2	91.6	90.5
Structured Deep Learning [43]	95.8	96.9	96.7	94.9
IRV2+1-NN_Aug [16]	98.04	97.50	97.85	97.48
RBM [15]	88.7	85.3	88.6	88.4
DenseNet CNN [24]	93.64	97.42	95.87	94.67
PFTS Features + 1-NN [14]	80.9	80.7	81.5	79.4
PFTS Features + SVM [14]	81.6	79.9	85.1	82.3
VGGNET16-RF [26]	92.22	93.40	95.23	92.80
VGGNET16-SVM(POLY) [26]	94.11	95.12	97.01	93.40
Xception model [27]	95.26	93.37	93.09	91.65
Proposed method	97.58	97.44	97.28	97.02