Table 8.

Accuracy comparison with some state-of-the-art algorithms for the binary classification task on the BreakHis dataset.

Ref.	Mag. fac.	Acc. (%)	Prec. (%)	R (%)	F1 (%)
Nahid and Kong [50]	40X	94.40	94.00	96.00	95.00
	100X	95.93	98.00	96.36	97.00
	200X	97.19	98.00	98.20	98.00
	400X	96.00	95.00	97.79	96.00
Han et al. [24]	40X	95.8 ± 3.1	—	—	—
	100X	96.9 ± 1.9	—	—	—
	200X	96.7 ± 2.0	—	—	—
	400X	94.9 ± 2.8	—	—	—
Pratiher and Chattoraj [51]	40X	96.8	—	—	—
	100X	98.1	—	—	—
	200X	98.2	—	—	—
	400X	97.5	—	—	—
Bardou et al. [23]	40X	98.33	97.80	97.57	97.68
	100X	97.12	95.58	96.98	97.77
	200X	97.85	95.61	99.28	97.41
	400X	96.15	97.54	96.49	97.07
K c(top-10% pseudolabels)	40X	99.52 ± 0.33	99.50	99.23	99.38
K c(top-5% pseudolabels)	100X	99.44 ± 0.41	99.73	99.58	99.69
K c(top-10% pseudolabels)	200X	99.48 ± 0.30	99.84	99.80	99.49
K c(top-10% pseudolabels)	400X	99.47 ± 0.37	99.85	99.77	99.54

Acc. denotes the accuracy, Prec. is the precision, R is the recall, and F1 is the F1-score.