Table 2 Summary of significant Mann–Whitney U tests for percentage change in radiomics features between pre- and post-contrast images

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

Radiomic feature No-upgrade Upgrade p-value
Median (IQR) Median (IQR)
Minimum (FO) 122
(80 to 228)
184
(111 to 333)
0.041
Sum average (GLCM)  − 16.7
(− 23.9 to − 1.1)
 − 21.8
(− 35.4 to − 8.3)
0.047
Auto correlation (GLCM)  − 26.4
(− 38.7 to 0.5)
 − 32.3
(− 54.3 to − 12.8)
0.044
lrhgle (RLM)  − 35.7
(− 47.9 to − 13.6)
 − 44.8
(− 63.9 to − 22.6)
0.026
rlv (RLM)  − 30.0
(− 48.1 to − 5.8)
 − 39.2
(− 58.7 to − 24.0)
0.045
lzhgle (SZM)  − 46.7
(− 60.6 to − 20.0)
 − 63.8
(− 73.0 to − 39.7)
0.008
hde (NGLDM)  − 21.1
(− 36.8 to − 3.8)
 − 30.4
(− 44.4 to − 14.4)
0.047
hgce (NGLDM)  − 31.4
(− 41.6 to − 4.7)
 − 38.0
(− 56.8 to − 14.6)
0.047
hdhge (NGLDM)  − 42.4
(− 60.6 to − 20.0)
 − 61.8
(− 72.9 to − 30.5)
0.006
Busyness (NGTDM) 108
(47 to 176)
160
(63 to 206)
0.047
  1. Median values with associated interquartile range (IQR) are reported as non-parametric tests were performed. Features selected for advancement to model development are italicized
  2. FO first-order, GLCM gray-level co-occurrence matrix, RLM run-length matrix, SZM size zone matrix, NGLDM neighborhood gray-level dependence matrix, NGTDM neighborhood gray-tone difference matrix, lrhgle long-run high gray-level emphasis, rlv run-length variance, lzhgle large zone high gray-level emphasis, hde high dependence emphasis, hgce high gray-level count emphasis, hdhge high dependence high gray-level emphasis