Skip to main content

Table 2 Univariate analysis of patient and lesion characteristics (DCE+DTI) for discriminating between malignant and benign lesions

From: Diffusion tensor imaging for characterizing tumor microstructure and improving diagnostic performance on breast MRI: a prospective observational study

 

Pathology status*

  

Univariate model

Malignant (N = 95)

Benign (N = 143)

AUC

(95% CI)

OR†

(95% CI)

p value

Clinical/DCE-MRI parameters

 Age, years

54.7 ± 11.3

49.3 ± 11.8

0.63

(0.56–0.71)

1.60

(1.19, 2.16)

0.002

 Post-menopausal

61 (64.2)

67 (46.9)

0.59

(0.52–0.66)

2.04

(1.14, 3.62)

0.016

 MRI indication: known cancer

77 (81.1)

84 (58.7)

0.61

(0.55–0.67)

3.00

(1.54, 5.85)

0.001

 Dense breasts

57 (60.0)

117 (81.8)

0.61

(0.55–0.67)

0.33

(0.18, 0.62)

0.001

 BPE category (1–4)

1.9 ± 1.0

2.3 ± 0.9

0.62

(0.54–0.70)

0.65

(0.47, 0.88)

0.006

 Lesion size, cm‡

26.9 ± 27.1

16.3 ± 17.7

0.64

(0.57–0.71)

1.66

(1.27, 2.16)

< 0.001

 Mass vs. NMLE/focus

54 (56.8)

81 (56.6)

0.50

(0.43–0.57)

1.01

(0.58, 1.76)

0.98

 Washout on delayed phase kinetics

87 (91.6)

111 (77.6)

0.57

(0.53–0.61)

3.14

(1.41, 7.00)

0.005

 BI-RADS 5 vs. 4

17 (17.9)

3 (2.1)

0.58

(0.53–0.62)

10.17

(2.11, 49.10)

0.004

DTI parameters

 Mean ADC, 10−3 mm2/s

1.26 ± 0.32

1.55 ± 0.30

0.75

(0.68–0.82)

0.37

(0.25, 0.54)

< 0.001

 Mean axial diffusivity, 10−3 mm2/s

1.62 ± 0.41

1.91 ± 0.36

0.73

(0.66–0.80)

0.42

(0.28, 0.64)

< 0.001

 Mean radial diffusivity, 10−3 mm2/s

1.08 ± 0.35

1.37 ± 0.33

0.74

(0.67–0.81)

0.40

(0.28, 0.59)

< 0.001

 Mean FA‡

0.28 ± 0.15

0.23 ± 0.13

0.61

(0.53–0.68)

1.45

(1.11, 1.91)

0.007

 Mean λ1 − λ3‡, 10−3 mm2/s

0.69 ± 0.46

0.69 ± 0.40

0.52

(0.44–0.60)

0.96

(0.74, 1.25)

0.77

  1. AUC area under the ROC curve, ROC receiver operating characteristic curve, OR odds ratio for malignancy, CI confidence interval
  2. *Values are no. (%) or mean ± SD
  3. †For continuous variables, ORs are scaled to show change per 1-SD increase in the corresponding variable
  4. ‡Variable was log-transformed prior to inclusion in the logistic regression model to reduce right-skewness
  5. NS = variable was included as a candidate predictor but was not selected by the LASSO; A blank cell indicates that the corresponding variable was not included as a candidate predictor in the model