Skip to main content

Table 4 Performances for the automatic model

From: MRI-based radiomics to predict response in locally advanced rectal cancer: comparison of manual and automatic segmentation on external validation in a multicentre study

  Construction set: centre A + centre B
  AUC
(95% CI)
ACC %
(95% CI)
[rate]
SE %
(95% CI)
[rate]
SP %
(95% CI)
[rate]
NPV %
(95% CI)
[rate]
PPV %
(95% CI)
[rate]
AP+Bayes 0.75
(67–83)
72
(59–82)
[45/63]
74
(54–89)
[20/27]
69
(52–84)
[25/36]
77
(65–87)
[25/32]
65
(51–76)
[20/31]
MRMR+Bayes 0.78
(69–85)
74
(60–83)
[46/63]
81
(62–94)
[22/27]
67
(49–81)
[24/36]
83
(68–92)
[24/29]
65
(44–70)
[22/34]
Ranking+EL (Bag) 0.90
(87–95)
68
(55–79)
96
(81–99)
47
(30–65)
94
(71–99)
58
(50–65)
[43/63] [26/27] [17/36] [17/18] [26/45]
AP+SVM polynomial 0.70
(64–82)
67
(54–78)
81
(62–94)
56
(38–72)
80
(63–90)
58
(48–67)
  [42/63] [22/27] [20/36] [20/25] [22/38]
Ranking+SVM linear 0.83
(78–90)
74
(62–85)
70
(50–86)
78
(66–87)
78
(66–87)
70
(55–82)
  [47/63] [19/27] [28/36] [28/36] [19/27]
Ranking+SVM Gaussian 0.86
(78–94)
78
(66–87)
[49/63]
81
(62–94)
[22/27]
75
(58–88)
[27/36]
84
(71–92)
[27/32]
71
(57–82)
[22/31]
  Validation set: centre C
Ranking+SVM Gaussian 0.81
(60–89)
75
(53–88)
[21/28]
80
(50–95)
[12/15]
69
(37–90)
[9/13]
75
(54–86)
[9/12]
75
(55–83)
[12/16]
  1. AUC Area under the curve, ACC Accuracy, NPV Negative predictive value, PPV Positive predictive value, SE Sensitivity, SP Specificity