Machine learning-based multiparametric MRI radiomics for predicting poor responders after neoadjuvant chemoradiotherapy in rectal Cancer patients

Table 4 AUCs for the performance of the predictive models in the validation cohort

Feature selection methods	Machine Learning Classifiers
Feature selection methods	DT	RF	SVM	LR	Adaboost
No ComBat
REF	0.569 (0.441-0.698)	0.781 (0.679-0.884)	0.851 (0.759-0.943)	0.648 (0.534-0.762)	0.429 (0.316-0.542)
mRMR	0.735 (0.622-0.849)	0.841 (0.752-0.931)	0.864 (0.782-0.947)	0.745 (0.628-0.863)	0.664 (0.554-0.775)
LASSO	0.715 (0.597-0.833)	0.879 (0.8-0.957)	0.892 (0.816-0.967)	0.738 (0.627-0.849)	0.62 (0.509-0.731)
mRMR+LASSO	0.767 (0.653-0.882)	0.879 (0.804-0.953)	0.923 (0.862-0.985)	0.792 (0.686-0.897)	0.709 (0.604-0.814)
ComBat
REF	0.774 (0.669-0.879)	0.726 (0.607-0.844)	0.79 (0.689-0.891)	0.728 (0.608-0.848)	0.72 (0.622-0.818)
mRMR	0.754 (0.648-0.859)	0.821 (0.728-0.915)	0.852 (0.763-0.94)	0.702 (0.578-0.826)	0.722 (0.617-0.827)
LASSO	0.781 (0.675-0.887)	0.877 (0.797-0.958)	0.878 (0.801-0.954)	0.682 (0.572-0.792)	0.747 (0.654-0.841)
mRMR+LASSO	0.658 (0.53-0.786)	0.831 (0.739-0.923)	0.887 (0.806-0.969)	0.742 (0.625-0.86)	0.562 (0.446-0.677)

Values in parentheses are 95% confidence intervals
LASSO least absolute shrinkage and selection operator, RFE recursive feature elimination, mRMR minimum redundancy maximum relevance feature selection, DT decision tree, RF random forest, SVM support vector machine, LR logistic regression

ISSN: 1471-2407