Summary of the test set performance of the ICS-based score system and two classical score systems (M1 and M2) for the prediction of malignancy of adnexal masses.

The classical score systems are based on a logistic regression model using the variables selected with ICS (LR). In a second step, the variables are manually divided into intervals. M1 uses a high number of intervals for continuous variables, M2 uses fewer intervals. The ICS approach is able to obtain good performance using a small number of intervals. The classical score systems are able to obtain good performance provided that a large number of intervals is considered.