Skip to main content

Table 2 Performance index of different models in two validation cohorts

From: Attention mechanism based multi-sequence MRI fusion improves prediction of response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer

Cohort

Model

AUC

Accuracy

Sensitivity

Specificity

PPV

NPV

Internal Validation

CFDL

0.898

(0.834–0.964)

77.8%

(68.8-87.0%)

76.7%

(64.3-89.9%

78.9%

(65.8-91.9%

80.5%

(68.2-92.9%

75.0%

(61.3-89.1%

DL

0.835

(0.745–0.921)

75.3%

(65.4-84.8%)

76.7

(63.9-89.1%)

73.7%

(58.9-87.9%)

76.7%

(63.5-89.6%)

73.7%

(59.4-87.3%)

CFRS

0.812

(0.711–0.916)

76.5%

(67.3-85.9%)

76.7%

(64.2-89.7%)

76.3%

(62.7-89.8%)

78.6%

(65.7-91.2%)

74.4%

(60.8-88.5%)

RS

0.757

(0.648–0.869)

69.1%

(59.1-79.3%)

67.4%

(53.1-81.7%)

71.1%

(57.0-85.4%)

72.5%

(58.6-86.3%)

65.9%

(51.5-80.4%)

External Validation

CFDL

0.873

(0.806–0.938)

86.5%

(81.1-91.7%)

91.9%

(86.9-96.9%)

72.1%

(58.8-85.2%)

89.6%

(84.1-95.1%)

77.5%

(64.3-90.2%)

DL

0.837

(0.764–0.910)

80.0%

(73.7-86.5%)

83.0%

(76.3-90.2%)

72.1%

(58.4-85.8%)

88.6%

(82.2-94.9%)

62.0%

(48.5-76.2%)

CFRS

0.832

(76.6-89.9%)

79.4%

(72.9-85.9%)

89.3%

(83.5-94.9%)

53.5%

(39.1-69.9%)

83.3%

(76.6-90.4%)

65.7%

(49.8-81.5%)

RS

0.806

(0.731–0.883)

81.3%

(75.1-87.6%)

97.3%

(94.4-100%)

39.5%

(24.8-54.8%)

80.7%

(74.0-87.6%)

85.0%

(69.8-100%)

  1. Data are given as value (95% confidence interval)
  2. PPV, positive predictive value; NPV, negative predictive value