Identifying factors that may influence the classification performance of radiomics models using contrast-enhanced mammography (CEM) images

Table 1 Summary of the study cohort and image features

Characteristics	Benign lesions (n = 47)	Malignant lesions (n = 114)	P value
Age (year) ^a	46.0 ± 7.9	50.7 ± 9.2	0.002
Lesion size (mm) ^a	17.1 ± 10.3	28.8 ± 15.6	< 0.001
Breast density			0.004
a-b	4/47 (8.5)	33/114 (28.9)
c-d	43/47 (91.5)	81/114 (71.1)
Degree of BPE			< 0.001
Minimal or Mild	20/47 (42.6)	86/114 (75.4)
Moderate or Marked	27/47 (57.4)	28/114 (24.5)
Rim Artifacts			1.000
Absent	42/47 (89.3)	100/114 (87.7)
Present	5/47 (10.7)	14/114 (12.3)
Ripple Artifacts			0.005
Absent	36/47 (76.6)	60/114 (52.6)
Present	11/47 (23.4)	54/114 (47.4)
Vascular Artifacts			0.042
Absent	41/47 (87.2)	82/114 (71.9)
Present	6/47 (12.8)	32/114 (28.1)
Air Trapping Artifacts			0.104
Absent	43/47 (87.2)	92/114 (71.9)
Present	4/47 (12.8)	22/114 (28.1)
SNR^a	183.2 ± 134.0	359.7 ± 141.9	< 0.001
CNR^a	194.5 ± 153.0	431.7 ± 193.1	< 0.001
BCR^a	108.6 ± 84.4	230.2 ± 103.2	< 0.001

BPE background parenchymal enhancement, SNR signal-to-noise ratio, CNR contrast-to-noise ratio, BCR background contrast ratio
^a Data are shown as the mean values ± standard deviations. Other data are shown as proportions with percentages in parentheses

ISSN: 1470-7330