Table 1.

Baseline characteristics of datasets.

	Training	Internal validation	External validation
Patients	818	167	125
Images	1000	200	200
Patients with malignant masses	500	100	88
Images with malignant masses	500	100	100
Patients with benign masses	346	68	39
Images with benign masses	500	100	100
Age groups*
< 30 years	16 (2%)	3 (2%)	12 (6%)
30–50 years	431 (53%)	93 (56%)	91 (46%)
≥ 50 years	371 (45%)	71 (43%)	97 (49%)
Tissue composition**
Homogeneous, fat	20 (2%)	5 (3%)	14 (11%)
Homogeneous, fibroglandular	956 (96%)	191 (96%)	69 (55%)
Heterogeneous	24 (2%)	4 (2%)	42 (34%)
Lesion size**
< 2 cm	754 (75%)	149 (75%)	163 (82%)
2–5 cm	239 (24%)	50 (25%)	37 (19%)
≥ 5 cm	7 (1%)	1 (1%)	0
US machine**
Phillips	1000 (100%)	200 (100%)	16 (8%)
GE	0	0	169 (85%)
Siemens	0	0	15 (8%)

*Patient-level, **image-level.