Skip to main content

Table 3 CRC Performance by Validation Procedure

From: Machine learning enables detection of early-stage colorectal cancer by whole-genome sequencing of plasma cell-free DNA

Validation

Mean AUC (95% CI)

Mean Sensitivity at 85% Specificity (95% CI)

k-fold

0.92 (0.91–0.93)

85% (83–86%)

Binned-age

0.91 (0.89–0.94)

79% (73–87%)

k-batch

0.91 (0.88–0.94)

85% (80–89%)

Ordered k-batch

0.90 (0.83–0.94)

73% (53–88%)

Balanced k-batch

0.83 (0.79–0.86)

71% (63–76%)

  1. CRC performance by cross-validation procedure in 50–84 year-old patients. Confidence intervals are calculated from bootstrapped distributions of the metric across folds