CT radiomics to differentiate between Wilms tumor and clear cell sarcoma of the kidney in children

BMC Medical Imaging

Table 4 A comparison of radiomics model performance when the stepwise elimination approach was applied on the employed feature selection methods

Statistical method eliminated	Training set					Test set
Statistical method eliminated	AUC	95% CI	Accuracy	Sensitivity	Specificity	AUC	95% CI	Accuracy	Sensitivity	Specificity
ICC	0.889	0.811–0.967	0.864	0.739	0.914	0.792	0.616–0.968	0.857	0.600	0.960
PCC	0.889	0.811–0.967	0.864	0.739	0.914	0.792	0.616–0.968	0.857	0.600	0.960
LASSO	0.905	0.831–0.978	0.901	0.739	0.966	0.764	0.592–0.936	0.714	0.800	0.680
Multivariate stepwise logistic regression	0.901	0.826–0.976	0.889	0.826	0.914	0.784	0.604–0.964	0.829	0.600	0.920
Combined	0.889	0.811–0.967	0.864	0.739	0.914	0.792	0.616–0.968	0.857	0.600	0.960

AUC Area under the curve, CI Confidence interval, ICC Intraclass correlation coefficient, PCC Pearson correlation coefficient, LASSO Least Absolute Shrinkage and Selection Operator

ISSN: 1471-2342