Relationship between Roe and Metz simulation model for multireader diagnostic data and Obuchowski-Rockette model parameters.

摘要: For the typical diagnostic radiology study design, each case (ie, patient) undergoes several tests (or modalities) and resulting images are interpreted by readers. Often, reader is asked to assign a confidence-of-disease rating for test, compared with respect reader-performance outcomes that functions of receiver operating characteristic (ROC) curves, such as area under ROC curve. These frequently analyzed using Obuchowski Rockette method, which allows conclusions generalize both populations. The simulation model proposed Roe Metz (RM) in 1997 emulates data collected from studies has been an important tool empirically evaluating various analysis methods. However, because RM parameters expressed terms continuous decision variable rather than outcomes, it not possible evaluate realism model. I derive relationships between Obuchowski-Rockette empirical curve outcome. make parameter models assess performance estimates. An example illustrates application assessing upper one-sided confidence bound test-by-reader variance component, useful sample size estimation.