Validation of Monte Carlo Estimates of Three-Class Ideal Observer Operating Points for Normal Data

The statistical estimation method should be key in the development of a pragmatic performance metric for evaluation of observers in classification tasks with three or more classes.