Prediction of lymph node status in patients with early-stage cervical cancer based on radiomic features of magnetic resonance imaging (MRI) images

Table 4 Comparison of the prediction performance of the Multinomial Naive Bayes (MNB) model using different features

Models	Cutoff	Sensitivity (95%CI)	Specificity (95%CI)	PPV (95%CI)	NPV (95%CI)	AUC (95%CI)	Accuracy (95%CI)
Radiomic features + clinical features model
Training set	0.508	0.371(0.211–0.532)	0.901(0.840–0.962)	0.591(0.385–0.796)	0.788(0.710–0.867)	0.611(0.607–0.615)	0.754(0.679–0.829)
Testing set	0.508	0.429(0.169–0.688)	0.900(0.807–0.993)	0.600(0.296–0.904)	0.818(0.704–0.932)	0.745(0.740–0.750)	0.778(0.667–0.889)
Clinical features model
Training set	0.116	0.514(0.349–0.680)	0.736(0.646–0.827)	0.429(0.279–0.578)	0.798(0.712–0.884)	0.641(0.637–0.644)	0.675(0.593–0.756)
Testing set	0.116	0.500(0.238–0.762)	0.800(0.676–0.924)	0.467(0.214–0.719)	0.821(0.700-0.941)	0.698(0.692–0.704)	0.722(0.603–0.842)
Radiomic features model
Training set	0.200	0.657(0.500-0.814)	0.440(0.338–0.542)	0.311(0.205–0.416)	0.769(0.655–0.884)	0.523(0.520–0.527)	0.500(0.413–0.587)
Testing set	0.200	0.929(0.794-1.000)	0.350(0.202–0.498)	0.333(0.185–0.481)	0.933(0.807-1.000)	0.632(0.627–0.637)	0.500(0.367–0.633)

Note: PPV, positive predictive value; NPV, negative predictive value; AUC, receiver operating characteristics curve area under the curve

ISSN: 1471-2342