ROC-Curve from Ordinal Probit Model

I did a baesian multilevel ordinal regression to analyse a recogniton memory experiment (Confidence Rating 1-6). Is it possible to use the point estimates from my model to set up a ROC-Curve?

I thought I could use point estimators to get the response probabilities of the corresponding categories, i.e. the probability to rate a 1, 2, 3, 4, 5 or 6. And calculate the false-Alarm and Hit-Rate for the average person.