Table 5.

Comparison of the performance of the models.

Model	Training set (n = 198)							Test set (n = 86)
	AUC (95% CI)	p-value (vs. CSRN)	ACC	SEN	SPE	NPV	PPV	AUC (95% CI)	p-value vs. CSRN	ACC	SEN	SPE	NPV	PPV
CSRN	0.905 (0.863–0.9472)	–	0.854	0.884 (0.813–0.935)	0.81 (0.699–0.887)	0.816 (0.717–0.893)	0.877 (0.800–0.931)	0.895 (0.828–0.962)	–	0.826	0.788 (0.653–0.889)	0.882 (0.725–0.967)	0.732 (0.579–0.914)	0.911 (0.783–0.957)
TRM	0.762 (0.695–0.829)	0.0004**	0.689	0.636 (0.544–0.722)	0.77 (0.656–0.855)	0.573 (0.478–0.707)	0.811 (0.713–0.864)	0.701 (0.588–0.814)	0.004**	0.686	0.635 (0.490–0.764)	0.765 (0.588–0.893)	0.578 (0.430–0.778)	0.805 (0.645–0.885)
TbRM	0.829 (0.771–0.888)	0.039*	0.773	0.818 (0.738–0.882)	0.701 (0.586–0.800)	0.711 (0.605–0.807)	0.812 (0.722–0.878)	0.769 (0.671–0.867)	0.039*	0.709	0.635 (0.490–0.764)	0.84 (0.655–0.932)	0.596 (0.449–0.813)	0.846 (0.691–0.911)
CSM	0.828 (0.769–0.887)	0.037*	0.808	0.909 (0.843–0.954)	0.649 (0.53–0.755)	0.820 (0.710–0.883)	0.803 (0.714–0.894)	0.761 (0.658–0.863)	0.033*	0.767	0.769 (0.632–0.875)	0.765 (0.588–0.893)	0.684 (0.527–0.847)	0.833 (0.687–0.913)
TCTbRM	0.860 (0.807–0.913)	0.072	0.808	0.785 (0.701–0.855)	0.844 (0.744–0.917)	0.714 (0.616–0.836)	0.888 (0.809–0.927)	0.817 (0.723–0.910)	0.046*	0.791	0.885 (0.766–0.956)	0.647 (0.46–0.803)	0.786 (0.610–0.890)	0.793 (0.645–0.917)

ACC, accuracy; SEN, sensitivity; SPE, specificity; NPV, negative predictive value; PPV, positive predictive value.

*Indicates significant difference after the DeLong test.

*p< 0.05, **p < 0.01.