Table 1 Performance of radiation classifier with varying feature selection methods, feature set size and train/test split.
From: Radiomics approach for identifying radiation-induced normal tissue toxicity in the lung
Feature selection method | Size of feature set | Predictor variable | Classification model | Train/test split | AUC | Accuracy (%) | Confidence intervals | P-value (* if < 0.05) | Sensitivity | Specificity |
|---|---|---|---|---|---|---|---|---|---|---|
Recursive feature elimination | 16 | Radiation | Random forest | 50/50 | 0.909 | 94.3 | 80.8–99.4 | * 2.62 × 10−4 | 0.818 | 1.00 |
Feature importance | 16 | Radiation | Random forest | 50/50 | 0.909 | 94.3 | 80.8–99.4 | * 2.62 × 10−4 | 0.818 | 1.00 |
Principal components | 16 | Radiation | Random forest | 50/50 | 0.534 | 57.1 | 39.4–72.7 | 9.46 × 10−1 | 0.182 | 0.750 |
Recursive feature elimination | 16 | Radiation | Random forest | 60/40 | 0.913 | 92.9 | 76.5–99.1 | * 5.89 × 10−3 | 0.875 | 0.950 |
Recursive feature elimination | 16 | Radiation | Random forest | 70/30 | 0.917 | 95.2 | 76.2–99.9 | * 8.03 × 10−3 | 0.833 | 1.00 |
Recursive feature elimination | 16 | Radiation | Random forest | 80/20 | 0.955 | 92.9 | 66.1–99.8 | 1.65 × 10−1 | 1.00 | 0.909 |
Recursive feature elimination | 13 | Radiation | Random forest | 50/50 | 0.770 | 82.9 | 66.35–93.44 | * 2.09 × 10−2 | 0.583 | 0.957 |
Recursive feature elimination | 23 | Radiation | Random forest | 50/50 | 0.750 | 80.0 | 63.06–91.56 | * 1.02 × 10−2 | 0.500 | 1.00 |
