Table 3.
Subgroup analysis of diagnostic performance for EGFR status in lung cancer brain metastases cases.
| Subgroup | Classification | Best feature selection method | Optimal feature number | AUC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|---|---|---|
| Small BMs | |||||||
| RF | RF | 24 | 87.12 | 86.60 | 100 | 86.92 | |
| SVM | RF | 34 | 89.08 | 89.28 | 100 | 89.06 | |
| AdaBoost | mRMR | 35 | 87.37 | 88.21 | 100 | 86.92 | |
| LASSO-LR | RF | 26 | 64.16 | 65.17 | 71.42 | 63.51 | |
| Large BMs | |||||||
| RF | Laplacian | 18 | 76.04 | 62.96 | 89.13 | 79.45 | |
| SVM | RF | 4 | 78.22 | 62.96 | 93.47 | 82.19 | |
| AdaBoost | Relief | 42 | 76.48 | 70.37 | 82.60 | 78.08 | |
| LASSO-LR | L0 | 5 | 57.85 | 22.22 | 93.47 | 67.12 | |
Brain metastases (BM), epidermal growth factor receptor (EGFR); area under the curve (AUC); random forest (RF); support vector machine (SVM).