Table 5 Univariate analysis according to independent radiologist assessment

From: Diagnostic value of radiomics and machine learning with dynamic contrast-enhanced magnetic resonance imaging for patients with atypical ductal hyperplasia in predicting malignant upgrade

Imaging feature Reader 1 p-value Reader 2 p-value
Overall No-upgrade Upgrade Overall No-upgrade Upgrade
BPE     0.11     0.9
 Minimal 33 (24) 27 (24) 6 (21)   24 (17) 20 (18) 4 (14)  
 Mild 46 (33) 39 (35) 7 (25)   41 (29) 34 (31) 7 (25)  
 Moderate 50 (36) 35 (32) 15 (54)   55 (40) 42 (38) 13 (46)  
 Marked 10 (7.2) 10 (9) 0 (0)   19 (14) 15 (14) 4 (14)  
FGT     0.6     0.8
 Almost entirely fat 2 (1.4) 1 (0.9) 1 (3.6)   1 1 (0.9) 0 (0)  
 Scattered FGT 35 (25) 28 (25) 7 (25)   37 (27) 29 (26) 8 (29)  
 Heterogeneous FGT 91 (65) 72 (65) 19 (68)   73 (53) 57 (51) 16 (57)  
 Extreme FGT 11 (7.9) 10 (9) 1 (3.6)   28 (20) 24 (22) 4 (14)  
Depth      > 0.9     0.9
 Anterior 35 (25) 28 (25) 7 (25)   36 (26) 28 (25) 8 (29)  
 Middle 74 (53) 59 (53) 15 (54)   70 (50) 57 (51) 13 (46)  
 Posterior 30 (22) 24 (22) 6 (21)   33 (24) 26 (23) 7 (25)  
T2 signal intensity     0.9     0.8
 Hypointense 2 (1.4) 2 (1.8) 0 (0)   17 (12) 13 (12) 4 (14)  
 Isointense 99 (71) 78 (70) 21 (75)   85 (61) 67 (60) 18 (64)  
 Hyperintense 38 (27) 31 (28) 7 (25)   37 (27) 31 (28) 6 (21)  
DWI signal (75 lesions)     0.7     0.5
 Homogeneous 29 (39) 24 (41) 5 (29)   28 (37) 24 (41) 4 (24)  
 Heterogeneous 6 (8) 4 (6.9) 2 (12)   16 (21) 12 (21) 4 (24)  
 Rim 1 (1.3) 1 (1.7) 0 (0)   1 (1.3) 1 (1.7) 0 (0)  
 No correlate 39 (52) 29 (50) 10 (59)   30 (40) 21 (36) 9 (53)  
ADC signal (61 lesions)     0.4      > 0.9
 Hyperintense 4 (6.6) 3 (6.4) 1 (7.1)   4 (6.6) 3 (6.4) 1 (7.1)  
 Hypointense 7 (11) 4 (8.5) 3 (21)   10 (16) 8 (17) 2 (14)  
 No correlate 50 (82) 40 (85) 10 (71)   47 (77) 36 (77) 11 (79)  
BI-RADS     0.7      > 0.9
 3 13 (9.4) 10 (9) 3 (11)   2 (1.4) 2 (1.8) 0 (0)  
 4 126 (91) 101 (91) 25 (89)   137 (99) 109 (98) 28 (100)  
Enhancement type     0.6     0.6
 Mass like 70 (50) 58 (52) 12 (43)   66 (47) 55 (50) 11 (39)  
 Non-mass like 67 (48) 51 (46) 16 (57)   71 (51) 54 (49) 17 (61)  
 Mixed 2 (1.4) 2 (1.8) 0 (0)   2 (1.4) 2 (1.8) 0 (0)  
Shape (mass)     0.3     0.2
 Oval 19 (26) 18 (30) 1 (8.3)   13 (18) 13 (21) 0 (0)  
 Round 16 (22) 12 (20) 4 (33)   19 (26) 15 (24) 4 (33)  
 Irregular 38 (52) 31 (51) 7 (58)   42 (57) 34 (55) 8 (67)  
Margins (mass)     0.8     0.5
 Circumscribed 30 (41) 25 (41) 5 (42)   36 (49) 32 (52) 4 (3)  
 Irregular 34 (47) 29 (48) 5 (42)   32 (43) 25 (40) 7 (58)  
 Spiculated 9 (12) 7 (11) 2 (17)   6 (8.1) 5 (8.1) 1 (8.3)  
Enhancement (mass)     0.3     0.4
 Homogeneous 28 (38) 25 (40) 3 (25)   37 (49) 33 (52) 4 (33)  
 Heterogeneous 35 (47) 29 (47) 6 (50)   38 (51) 30 (48) 8 (67)  
 Rim enhancement 8 (11) 5 (8.1) 3 (35)   0 (0) 0 (0) 0 (0)  
 Dark internal septations 3 (4.1) 3 (4.8) 0 (0)   0 (0) 0 (0) 0 (0)  
Distribution (non-mass)     0.2     0.3
 Focal 23 (33) 20 (37) 3 (19)   26 (35) 22 (39) 4 (24)  
 Linear 27 (39) 22 (41) 5 (31)   26 (35) 21 (37) 5 (29)  
 Segmental 18 (26) 11 (20) 7 (44)   19 (26) 12 (21) 7 (41)  
 Regional 4 (2.9) 1 (1.9) 1 (6.2)   3 (4.1) 2 (3.5) 1 (5.9)  
 Multiple regions 0 (0) 0 (0) 0 (0)   0 (0) 0 (0) 0 (0)  
 Diffuse 0 (0) 0 (0) 0 (0)   0 (0) 0 (0) 0 (0)  
Enhancement (non-mass)     0.2     0.5
 Homogeneous 15 (21) 9 (17) 6 (38)   16 (22) 12 (21) 4 (24)  
 Heterogeneous 35 (50) 29 (54) 6 (38)   41 (55) 30 (53) 11 (65)  
 Clumped 20 (29) 16 (30) 4 (25)   17 (23) 15 (26) 2 (12)  
 Clustered rings 0 (0) 0 (0) 0 (0)   0 (0) 0 (0) 0 (0)  
  1. ADC apparent diffusion coefficient, BI-RADS breast imaging-reporting & data system, BPE background parenchymal enhancement, DWI diffusion-weighted imaging, FGT fibroglandular tissue