Table 4.
The best feature combinations that yield the highest predictive accuracy of statistical models for OAR overall toxicities‐based classification. Features that are consistently selected by all the models are bold
Nearest neighbors | Linear SVM | RBF SVM | Decision Tree | Random Forest | Ada‐boost | Naive Bayes | Linear DA | Quadratic DA | |
---|---|---|---|---|---|---|---|---|---|
Selected features | VolH | VolB | VolH | DisL‐B | VolB | VolH | VolB | VolH | VolB |
VolipsL | DisL‐B | VolH | |||||||
DisL‐B | VolipsL | ||||||||
DisH‐B | |||||||||
DisL‐B | |||||||||
Laterality | |||||||||
Thickness | |||||||||
VolH/VolL | |||||||||
Accuracy | 0.87 | 0.80 | 0.80 | 0.80 | 0.87 | 0.87 | 0.93 | 0.80 | 0.93 |
VolB: Breast Volume; VolH: Heart Volume; VolipsL: IpsiLung Volume; DisH‐B: Distance between Breast and Heart; DisL‐B: Distance between Breast and ipsilateral Lung; VolH‐in‐field: Volume of heart in the treatment field; Thickness: Deep breath motion thickness variation; VolH/VolL: Rate of heart volume to lung volume.