Skip to main content

Table 3 Model performance in predicting early recurrence of pancreatic cancer in the development and test cohorts

From: Preoperative prediction of early recurrence in resectable pancreatic cancer integrating clinical, radiologic, and CT radiomics features

 

AUC

P-value

Brier score

Accuracy

Sensitivity

Specificity

Development cohort

Radiomics model

0.73

(0.57–0.84)

0.048*

0.20

(0.16–0.25)

0.64

(0.50–0.77)

0.75

(0.38–0.96)

0.60

(0.46–0.84)

CR model

0.70

(0.60–0.77)

0.26*

0.20

(0.18–0.24)

0.70

(0.61–0.76)

0.50

(0.26–0.68)

0.81

(0.66–0.93)

CRR model

0.77

(0.63–0.86)

–

0.19

(0.14–0.24)

0.66

(0.51–0.78)

0.77

(0.49–0.96)

0.61

(0.47–0.84)

Test cohort

Radiomics model

0.69

(0.51–0.85)

0.038†

0.24

(0.22–0.25)

0.45

(0.43–0.50)

1.00

(1.00–1.00)

0.04

(0.00–0.13)

CR model

0.76

(0.56–0.83)

0.17†

0.21

(0.20–0.26)

0.70

(0.53–0.78)

0.41

(0.12–0.53)

0.91

(0.78–1.00)

CCR model

0.83

(0.65–0.94)

–

0.20

(0.18–0.24)

0.78

(0.60–0.85)

0.65

(0.29–0.77)

0.87

(0.70–1.00)

  1. Data are presented as mean (95% CI)
  2. *Pairwise comparison with CCR model using bootstrap
  3. †Pairwise comparison with CCR model using Delong’s test
  4. AUC, area under the receiver operating characteristic curve; CI, confidence interval; CR, clinical-radiologic; CRR, clinical-radiologic-radiomics