Table 4.

Results of the ROC analysis

	AUC (95% CI)	Specificity (%)	Sensitivity (%)	PPV (%)	NPV (%)
Neural network (training)	0.96 (0.92–0.99)	96.9	87.5	82.4	97.9
Neural network (validation)	0.84 (0.75–0.93)	80.3	84.2	32.0	97.9
Reader 1	0.89 (0.79–0.98)	89.0	84.2	45.7	98.1
Reader 2	0.89 (0.79–0.98)	82.7	84.2	34.8	97.9
Reader 3	0.79 (0.69–0.89)	72.8a	73.7a	23.0	96.2

AUC, area under the receiver operating characteristic curve; PPV, positive predictive value; NPV, negative predictive value.

^ap < 0.05 compared to both other readers and the NN.