Development and Validation of the Random Forest Model via Combining CT-PET Image Features and Demographic Data for Distant Metastases among Lung Cancer Patients
Table 2
The performance of the three random forest models.
Random forest models
Dataset
PPV (95% CI)
NPV (95% CI)
AUC (95% CI)
Accuracy (95% CI)
CT demographic data
Training set
0.556 (0.393–0.718)
0.949 (0.905–0.993)
0.880 (0.807–0.953)
0.843 (0.782–0.905)
CT demographic data
Testing set
0.429 (0.169–0.688)
0.833 (0.700–0.967)
0.716 (0.531–0.902)
0.705 (0.570–0.839)
PET demographic data
Training set
0.410 (0.286–0.533)
1.000 (1.000–1.000)
0.917 (0.865–0.969)
0.731 (0.656–0.806)
PET demographic data
Testing set
0.381 (0.173–0.589)
0.870 (0.732–1.000)
0.802 (0.633–0.970)
0.636 (0.494–0.779)
CT-PET
Training set
0.523 (0.375–0.670)
0.967 (0.930–1.000)
0.904 (0.850–0.959)
0.821 (0.756–0.886)
CT-PET
Testing set
0.364 (0.079–0.648)
0.818 (0.687–0.950)
0.797 (0.666–0.928)
0.705 (0.570–0.839)
CT- PET demographic data
Training set
0.714 (0.547–0.882)
0.953 (0.912–0.993)
0.923 (0.873–0.973)
0.903 (0.853–0.953)
CT-PET demographic data
Testing set
0.750 (0.450–1.000)
0.861 (0.748–0.974)
0.873 (0.757–0.990)
0.841 (0.733–0.949)
Note. CT: computed tomography; PET: positron emission tomography; CI: confidence interval; PPV: positive predictive value; NPV: negative predictive value; AUC: area under the curve.