Reducing the number of unnecessary biopsies for mammographic BI-RADS 4 lesions through a deep transfer learning method

Table 2 Downgraded lesions by the DTL model and histopathology results in the validation set

R	N1			N2				Histologically benign			Histologically malignant
R	4 A	4B	4 C	4 A	4B	4 C		4 A	4B	4 C	4 A	4B	4 C
R1	29	22	19	27	19	16	DTL benign	26	18	14	1	1	2
							DTL malignant	1	1	0	1	2	3
R2	32	20	22	29	16	17	DTL benign	28	15	15	1	1	2
							DTL malignant	1	0	1	2	4	4
R3	29	19	20	26	17	16	DTL benign	25	15	14	1	2	2
							DTL malignant	1	0	1	2	2	3
R4	31	24	21	27	21	17	DTL benign	26	19	14	1	2	3
							DTL malignant	1	0	0	3	3	4
R5	30	20	24	28	16	19	DTL benign	27	14	16	1	2	3
							DTL malignant	0	0	1	2	4	4
Total	151	105	106	137	89	85		136	82	76	15	23	30

Note: R, resident; N1, number of BI-RADS 4 lesions diagnosed by residents; N2, number of downgraded lesions according to the DTL model

ISSN: 1471-2342