An MRI-based fusion model for preoperative prediction of perineural invasion status in patients with intrahepatic cholangiocarcinoma

Table 2 The performance comparison of different machine learning models

Model		AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
LR	Training set	0.727[0.644–0.810]	0.714	0.466	0.876	0.711	0.716
	Test set	0.673[0.512–0.834]	0.511	0.929	0.323	0.382	0.909
NaiveBayes	Training set	0.696[0.609–0.784]	0.667	0.586	0.719	0.576	0.727
	Test set	0.673[0.486–0.859]	0.689	0.643	0.710	0.500	0.815
KNN	Training set	0.755[0.680–0.830]	0.701	0.431	0.876	0.694	0.703
	Test set	0.634[0.467-0.800]	0.667	0.357	0.806	0.455	0.735
ExtraTrees	Training set	0.796[0.720–0.873]	0.762	0.672	0.820	0.709	0.793
	Test set	0.713[0.547–0.879]	0.689	0.643	0.710	0.500	0.815
MLP	Training set	0.760[0.679–0.841]	0.701	0.793	0.640	0.590	0.826
	Test set	0.705[0.543–0.867]	0.600	0.857	0.484	0.429	0.882

AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; LR, logistic regression; KNN, k-nearest neighbors; ExtraTrees, extremely randomized trees; MLP, multilayer perceptron

ISSN: 1477-7819