Skip to main content

Table 3 Texture features that showed significantly different mean values comparing malignant versus benign solid lesions and corresponding area under the curve (AUC)

From: Diagnostic performance of machine learning applied to texture analysis-derived features for breast lesion characterisation at automated breast ultrasound: a pilot study

Feature

Malignant lesions (mean ± standard deviation)

Benign lesions (mean ± standard deviation)

p value

AUC (95% confidence interval)

Entropy

5.28 ± 0.38

5.67 ± 0.16

< 0.00001 for all

0.86 (0.82–0.89)

Skewness

0.74 ± 0.33

0.54 ± 0.41

0.66 (0.61–0.70)

Kurtosis

0.53 ± 0.67

0.31 ± 0.66

0.61 (0.56–0.65)

Contrast

24.45 ± 10.61

28.40 ± 13.66

0.58 (0.53–0.63)

GLN (GLCM)

96.33 ± 87.47

60.58 ± 61.73

0.67 (0.62–0.71)

RLN (GLCM)

2,218.97 ± 1,834.77

1,523.25 ± 1,681.38

0.66 (0.62–0.71)

HGRE (GLCM)

1,163.54 ± 20.49

1,171.63 ± 19.35

0.62 (0.57–0.67)

SRHGE (GLCM)

1,067.79 ± 32.58

1,080.79 ± 29.69

0.63 (0.58–0.67)

GLN (GLRLM)

96.43 ± 87.45

60.70 ± 61.72

0.67 (0.62–0.71)

RLN (GLRLM)

2,218.89 ± 1,828.26

1,525.31 ± 1,677.17

0.67 (0.62–0.71)

HGRE (GLRLM)

1,165.00 ± 20.59

1,173.06 ± 19.57

0.62 (0.57–0.67)

SRHGE (GLRLM)

1,069.20 ± 33.13

1,082.28 ± 29.75

0.62 (0.58–0.67)

  1. GLN Grey-level non-uniformity, GLCM Grey-level co-occurrence matrix, RLN Run length non-uniformity, HGRE High grey-level run emphasis, SRHGE Short-run high grey-level emphasis, GLN Grey-level non-uniformity, GLRLM Grey-level run length matrix