Simulation of unequal-variance binormal multireader ROC decision data: an extension of the Roe and Metz simulation model.

摘要: Rationale and Objectives Roe Metz (RM) proposed a model for simulating multireader multicase (MRMC) data collected from factorial study design in which readers read the same cases all modalities. However, major weakness of RM is that it generates according to an equal-variance binormal each reader. This article extends by allowing diseased nondiseased decision-variable distributions have unequal variances Materials Methods I show how modify so unequal-variance In doing so, preserve other important characteristics original simulation input values. The mean-to-sigma ratio, describes relationship between means distributions, constrained value representative many sets. last point illustrated with example comparing performances spin echo cine magnetic resonance imaging detecting thoracic aortic dissection. Results A performed assess performance MRMC methods Dorfman, Berbaum, Obuchowski Rockette using variance extension model. either excellent or acceptable when there are at least five 25 normal abnormal cases. Conclusions more similar radiological studies.