MRI feature-based radiomics models to predict treatment outcome after stereotactic body radiotherapy for spinal metastases

Insights into Imaging

Table 3 Discrimination performance of all the models

Model	Optimal ML classifier	AUC	Sensitivity	Specificity	Accuracy	Precision
Clinical
T1WI	QDA	0.779	0.436	0.936	0.799	0.738
T1WI + Clinical	SVM	0.816	0.478	0.895	0.778	0.652
T2WI	GP	0.823	0.460	0.944	0.814	0.751
T2WI + Clinical	GP	0.808	0.457	0.965	0.814	0.798
FS-T2WI	QDA	0.745	0.195	0.972	0.763	0.733
FS-T2WI + Clinical	LR	0.828	0.598	0.880	0.804	0.642
ALL	GP	0.825	0.473	0.957	0.825	0.830
ALL+ Clinical*	GP	0.828	0.511	0.950	0.830	0.813

ML Machine learning, AUC Area under the curve, ALL T1WI + T2WI + FS-T2WI sequences, QDA Quadratic discriminant analysis, SVM Support vector machine; GP Gaussian processes, LR Logistic regression; *best model