Skip to main content

Table 4 Results of machine learning models for predicting local lung fibrosis

From: Radiomics for prediction of radiation-induced lung injury and oncologic outcome after robotic stereotactic body radiotherapy of lung cancer: results from two independent institutions

  Coxnet Gradient boost
Features number of features CCI train-set CCI cross-valid CCI test-set Number of features CCI train-set CCI cross-valid CCI test-set
Clinical/dosimetric 3§ 0.71
p < 0.005
0.68 ± 0.11 0.65
p = 0.04*
3§ 0.73
p < 0.005
0.64 ± 0.12 0.62
n.s
Radiomics 10 0.79
p < 0.005
0.64 ± 0.13 0.58
n.s
2 0.75
p < 0.005
0.72 ± 0.11 0.59
p = 0.02*
Combined 4 + 7 0.74
p < 0.005
0.67 ± 0.12 0.66
p = 0.03*
0 + 2 0.72
p < 0.005
0.72 ± 0.11 0.59
p = 0.02*
  1. CCI concordance index, means ± standard deviation are shown, p-values: significance level of the model risk score in univariate Cox regression analysis
  2. §Age/ GTVMeanDose/LungD1ml
  3. wavelet_HLH_glcm_MCC/wavelet_HLL_glcm_MCC (= GrayLevelCo-occurrence matrix maximal correlation coefficient)