Background
Methods
Data acquisition
Prediction process
Image pre-processing and Radiomics features extraction
Features selection and machine learning methods
Methods | Software | Packages | Website Links |
---|---|---|---|
PCC | SML toolbox | corr | |
KCC | |||
SCC | |||
MI | MIToolbox | mi | |
CI | Hisc | rcorr.cens | |
Cox | survival | coxph | |
GB-Cox | mboost | mboost | |
GB-Cindex | mboost | mboost | |
CoxBoost | CoxBoost | CoxBoost | |
BST | ipred | bagging | |
RFS | randomForestSRC | rfsrc | |
SR | survival | survreg | |
SVCR | survivalsvm | survivalsvm |
Parameters tuning
Evaluation methods
Results
Methods | FS | Maximum CI | CFI of Maximum CI |
---|---|---|---|
GB-Cox | CI | 0.682 | [0.620, 0.744] |
CoxBoost | CI | 0.674 | [0.615, 0.731] |
Cox | MI | 0.646 | [0.578, 0.714] |
GB-Cindex | SCC | 0.357 | [0.290, 0.423] |
RFS | PCC | 0.627 | [0.558, 0.695] |
SR | MI | 0.380 | [0.310, 0.452] |
BST | SCC | 0.385 | [0.318, 0.450] |
SVCR | KCC | 0.405 | [0.341, 0.470] |
Methods | Parameters | Range of Parameters |
---|---|---|
Cox | ||
GB-Cox | Number of boosting steps | [1, 500] |
GB-Cindex | Number of boosting steps | [1, 500] |
Coxboost | Number of boosting steps | [1, 500] |
BST | Minsplit | [1, 10] |
Number of trees | [1, 500] | |
RFS | Average terminal node size of forest | [1, 10] |
Number of trees | [1, 500] | |
SR | Assumed distribution | Weibull, Gaussian, Exponential |
SVCR | Parameter of regularization | [0.01, 1] |