Table 3 Performance in predicting the invasiveness of lung adenocarcinoma using machine learning-based pathomics models.
From: Identifying invasiveness to aid lung adenocarcinoma diagnosis using deep learning and pathomics
Model name | ACC | AUC | 95% CI | SE | SP | PPV | NPV | Cohort |
|---|---|---|---|---|---|---|---|---|
RandomForest | 0.817 | 0.897 | 0.8572–0.9373 | 0.795 | 0.865 | 0.925 | 0.667 | Training |
RandomForest | 0.814 | 0.807 | 0.6892–0.9239 | 0.900 | 0.632 | 0.837 | 0.750 | Test |
ExtraTrees | 0.848 | 0.930 | 0.8998–0.9611 | 0.801 | 0.946 | 0.969 | 0.693 | Training |
ExtraTrees | 0.814 | 0.797 | 0.6781–0.9166 | 0.900 | 0.632 | 0.837 | 0.750 | Test |
XGBoost | 0.974 | 0.997 | 0.9922–1.0000 | 0.968 | 0.986 | 0.993 | 0.936 | Training |
XGBoost | 0.695 | 0.736 | 0.5947–0.8764 | 0.625 | 0.889 | 0.893 | 0.516 | Test |
LightGBM | 0.796 | 0.880 | 0.8367–0.9242 | 0.795 | 0.797 | 0.892 | 0.648 | Training |
LightGBM | 0.780 | 0.803 | 0.6874–0.9179 | 0.850 | 0.632 | 0.829 | 0.667 | Test |
