Table 2.
ML model prediction performance for external (EXT) and internal (INT) lymphedema for the two patient cohorts.
| EXT lymphedema (entire cohort) | Mean |
1 standard deviation |
||||
|---|---|---|---|---|---|---|
| Accuracy | F1-score | AUC | Accuracy | F1-score | AUC | |
| Logit | 67.2 | 71.7 | 77.1 | 3.5 | 3.8 | 12.3 |
| SVM | 64.5 | 68.4 | 77.6 | 2.8 | 5.8 | 13 |
| XGB | 77.7 | 85.1 | 78.2 | 2.9 | 2.7 | 12.7 |
| RF | 74.9 | 83.3 | 79.1 | 5.3 | 3.8 | 11.2 |
| INT lymphedema (entire cohort) | ||||||
| Logit | 69.9 | 59 | 74.5 | 8 | 17.4 | 9 |
| SVM | 63.3 | 55.8 | 70.7 | 7.4 | 18.1 | 9.2 |
| XGB | 76.3 | 66.9 | 79.5 | 6.7 | 13 | 7.8 |
| RF | 67.3 | 60.1 | 76 | 10 | 10.9 | 8.1 |
| EXT lymphedema (oropharengeal cohort) | ||||||
| Logit | 58.7 | 68.1 | 70.9 | 10.9 | 10 | 20.2 |
| SVM | 76.2 | 86 | 68.2 | 3.1 | 3 | 22.5 |
| XGB | 84.9 | 91.3 | 67.6 | 4.9 | 2.6 | 28.3 |
| RF | 78.4 | 87.6 | 74.3 | 6 | 3.5 | 22.4 |
| INT lymphedema (oropharengeal cohort) | ||||||
| Logit | 73.8 | 65.3 | 79.9 | 16.8 | 20.7 | 9.5 |
| SVM | 71.6 | 63.9 | 77.8 | 18.2 | 21.4 | 10.9 |
| XGB | 78.4 | 64.1 | 84.3 | 9.3 | 20 | 15.5 |
| RF | 80.7 | 70.4 | 83.7 | 12.3 | 20.4 | 15.3 |
Logit = Logistic regression, SVM = Support Vector Machine, XGB = Extreme Gradient Boosting,RF = Random Forest.