Fig. 2.

Receiving Operator Characteristic (ROC) curve for the most complex decision tree model proposed in this work. The range of AUC values (95% C.I.) is indicated. The model achieves an AUC of 0.93. For each trade-off between sensitivity and specificity, the corresponding threshold on the model output is shown (right vertical axis). Dashed lines indicate the position of the optimal trade-off, alongside with the high-sensitivity and high-specificity trade-offs (see text for additional details).