Receiver operating characteristic (ROC) methodology: the state of the art.

摘要: Receiver operating characteristic (ROC) methodology has been increasingly used in medical applications the last 10 years. The text by Swets and Pickett popularized technique journal Medical Decision Making (1981--) provides a forum for further methodologic issues. In this article, I will (1) describe nature of data generated ROC studies; (2) evaluate choices summary indices performance (accuracy); (3) outline data-analytic techniques used, how to incorporate from multiple observers "readings"; (4) review proposed alternatives commonly binormal model; (5) discuss issues, such as verification bias, challenges, multicenter comparative imaging studies difficulty obtaining "truth data", which need be addressed when adapting methods contexts.