|
Abstract:
|
For the typical diagnostic radiology study, several readers assign confidence-of-disease ratings to each case (i.e., subject) based on one or more corresponding radiologic images. These studies are often used to compare different imaging modalities with respect to reader performance. Often measures of reader performance are functions of the estimated receiver-operating-characteristic (ROC) curve, such as the area under the ROC curve (AUC). A commonly used method for analyzing reader performance outcomes that allows conclusions to generalize to both the reader and case populations is the Obuchowski and Rockette (OR) method, first proposed in 1995. However, the OR model has been developed for only a few basic balanced ANOVA models. In this talk I show how the OR model can be reframed as a conventional linear mixed model (LMM) that can be fitted using conventional LMM software, provided that a fixed-reader covariance matrix is additionally computed to specify some of the variance components in the model and a work-around is used to estimate the degrees of freedom. The advantage of the proposed implementation is that more complex OR models can be easily fitted.
|
