Skip to main content

Table 2 Performances of the prediction models in the test set for 5-year mortality and 5-year metachronous liver metastasis

From: Liver imaging features by convolutional neural network to predict the metachronous liver metastasis in stage I-III colorectal cancer patients based on preoperative abdominal CT scan

Predictors

Logistic regression classification

AUC (mean +/− standard deviation)

Random forest classification

AUC (mean, standard deviation)

Prediction model for 5-year metachronous liver metastasis

 Clinical*

0.709 +/− 0.038

0.692 +/− 0.038

 PC1

0.606 +/− 0.044

0.557 +/− 0.043

 PC1-PC2

0.600 +/− 0.042

0.536 +/− 0.042

 PC1-PC3

0.588 +/− 0.040

0.503 +/− 0.046

 PC1-PC4

0.580 +/− 0.040

0.520 +/− 0.042

Clinical + PC1

0.747 +/− 0.036

0.697 +/− 0.038

 Clinical + PC1-PC2

0.744 +/− 0.036

0.676 +/− 0.043

 Clinical + PC1-PC3

0.740 +/− 0.038

0.668 +/− 0.042

 Clinical + PC1-PC4

0.736 +/− 0.038

0.691 +/− 0.042

Prediction model for 5-year mortality

 Clinical*

0.704 +/− 0.028

0.679 +/− 0.030

 PC1

0.482 +/− 0.031

0.511 +/− 0.030

 Clinical + PC1

0.695 +/− 0.031

0.647 +/− 0.033

  1. *Clinical: Age, Sex, T stage, N stage