Skip to main content
. 2024 Jul 31;9(8):103595. doi: 10.1016/j.esmoop.2024.103595

Table 1.

Evaluating model performances in the three cohorts at targeted 90% specificity

Training cohort Actual
Lung cancer Benign nodule
Predict, n
 Lung cancer 175 18
 Benign nodule 18 26
Sensitivity (95% CI) 90.7% (85.7% to 94.4%)
Specificity (95% CI) 59.1% (43.2% to 73.7%)
PPV (95% CI) 92.5% (85.1% to 96.9%)
NPV (95% CI) 59.1% (43.2% to 73.7%)
Accuracy (95% CI)
84.8% (79.6% to 89.1%)
Validation cohort Actual
Lung cancer Benign nodule
Predict, n
 Lung cancer 86 7
 Benign nodule 10 15
Sensitivity (95% CI) 89.6% (81.7% to 94.9%)
Specificity (95% CI) 68.2% (45.1% to 86.1%)
PPV (95% CI) 92.5% (85.1% to 96.9%)
NPV (95% CI) 60% (38.7% to 78.9%)
Accuracy (95% CI)
85.6% (77.9% to 91.4%)
External test cohort Actual
Lung cancer Benign nodule
Predict, n
 Lung cancer 52 15
 Benign nodule 6 26
Sensitivity (95% CI) 89.7% (78.8% to 96.1%)
Specificity (95% CI) 63.4% (46.9% to 77.9%)
PPV (95% CI) 77.6% (65.8% to 86.9%)
NPV (95% CI) 81.2% (63.6% to 92.8%)
Accuracy (95% CI) 78.8% (69.4% to 86.4%)

CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value.