Table 3 AUROC for LNM prediction

From: Predicting lymph node metastasis from primary tumor histology and clinicopathologic factors in colorectal cancer using deep learning

Model

Temporal validation

External validation 1a

External validation 1b

Clinical

0.667 [0.626, 0.708]

0.716 [0.674, 0.762]

0.719 [0.684, 0.752]

Clinical + ML

0.715 [0.674, 0.753]

0.740 [0.701, 0.780]

0.738 [0.705, 0.770]

Delta

0.048 [0.027, 0.069]

0.024 [−0.001, 0.047]

0.019 [0.000, 0.037]

  1. AUROCs for LNM predictions for logistic regressions with various feature sets. Clinical baseline clinicopathologic variables (age, sex, tumor grade, T-category, lymphatic invasion, venous invasion). Clinical + ML baseline clinicopathologic variables plus 5 machine-learned features. Delta the difference between Clinical + AI and Clinical. 95% confidence intervals computed via bootstrapping.
Back to article page