Table 3 Validation and test accuracy of machine learning models and the human predictive factor (HPF). The accuracy values reflect the models’ ability to predict fibrosis risk based on CT-derived parameters and lung volumes.

From: Machine learning-driven imaging data for early prediction of lung toxicity in breast cancer radiotherapy

Model

Accuracy valid %

Accuracy test %

Tree

54.1

83.1

Kernel

55.4

81.9

kNN

62

100

HPF

62.81

72