Table 4. MSB classification performance of baseline models.

Methods	Magnification	Image_accuracy(%)	Patient_accuracy(%)	Sensitivity(%)	Precision(%)	F1_score(%)
AlexNet	40×	86.38±3.46	87.21±3.85	87.48±6.60	93.85±2.22	90.41±3.16
	100×	88.17±3.52	88.66±3.35	87.23±4.25	96.55±1.93	91.62±2.82
	200×	89.48±3.38	88.67±3.45	90.79±5.80	94.93±1.28	92.72±2.79
	400×	89.13±3.18	88.50±3.11	91.10±5.21	93.87±2.99	92.37±2.65
VGG16	40×	90.26±2.75	90.50±2.96	91.35±4.47	95.50±2.48	93.30±2.02
	100×	90.06±4.83	89.40±4.54	91.25±4.96	95.11±2.75	93.11±3.64
	200×	93.87±2.00	93.15±2.88	93.88±2.51	97.84±1.08	95.80±1.50
	400×	90.82±2.62	89.83±1.99	93.74±2.31	93.75±3.43	93.71±1.92
ResNet50	40×	90.24±4.27	90.39±4.54	90.64±5.32	95.87±2.08	93.14±3.52
	100×	92.09±3.28	92.17±3.21	94.10±3.16	95.13±2.63	94.60±2.52
	200×	91.04±2.42	89.57±2.28	93.01±3.09	94.94±1.33	93.94±1.80
	400×	88.75±2.56	87.96±2.55	91.23±5.79	93.42±3.35	92.15±2.28
GoogLeNet	40×	91.61±3.16	91.29±3.30	91.66±3.83	96.87±2.14	94.16±2.47
	100×	91.73±3.42	91.97±3.76	91.82±5.03	96.82±0.90	94.20±2.79
	200×	92.00±2.31	90.72±2.47	92.90±3.96	96.32±2.33	94.52±1.88
	400×	90.39±2.28	89.61±2.45	90.74±3.32	96.04±4.07	93.22±1.82
SqueezeNet	40×	86.99±1.53	85.69±3.06	87.91±5.38	94.54±3.19	90.94±1.46
	100×	91.35±3.99	91.04±4.22	93.73±3.74	94.51±2.84	94.10±2.98
	200×	93.09±1.22	92.57±2.37	94.59±1.98	96.20±2.40	95.36±0.92
	400×	88.89±4.20	87.93±4.30	89.76±7.25	94.70±1.72	92.01±3.68
DenseNet201	40×	88.23±2.84	88.62±2.95	87.45±3.70	96.36±1.89	91.66±2.46
	100×	90.38±3.83	90.84±3.70	90.49±5.99	96.40±2.29	93.24±3.16
	200×	91.03±2.46	90.14±2.72	90.69±3.92	97.14±1.63	93.76±2.01
	400×	89.23±2.43	88.78±2.44	89.55±4.38	95.48±2.22	92.34±2.09
Inception-ResNet-V2	40×	88.14±2.44	88.29±4.15	93.05±3.09	91.43±4.23	92.14±1.67
	100×	91.27±0.90	91.16±0.76	95.21±2.55	93.33±3.18	94.20±0.76
	200×	90.66±0.66	89.74±1.38	95.11±3.46	92.86±3.34	93.88±0.34
	400×	87.93±1.90	87.25±2.12	91.80±2.43	91.91±4.38	91.76±1.34