Use of ROC Curves for Analysis of Diagnostic Radiologic Imaging Data: Review and Outlook

Department of Biomedical Engineering
Department of Psychological and Brain Sciences

Joint Colloquium

This talk will first review the definition and purpose of receiver operating characteristic (ROC) curves and then review the history of their use in the field of radiology to assess reader performance, starting in 1992. That year, the first practical method for generalizing results from ROC curve analyses to both the reader and patient populations was proposed by Dorfman, Berbaum, and Metz (DBM). The speaker will discuss this and subsequent developments, including his contributions. Although statistical concepts will be discussed, this will be primarily a conceptual talk. In addition, the speaker will briefly demonstrate new R software for performing ROC curve analyses that he and Brian Smith have developed.

Stephen L. Hillis is a Research Professor in the Departments of Radiology and Biostatistics at the University of Iowa. He earned a PhD in statistics in 1987 and an MFA in music 1978, both from the University of Iowa. He is the author of more than 100 peer-reviewed journal articles and four book chapters. Since 1998, when he began collaborating with Don Dorfman, his research has focused on methodology for multi-reader diagnostic radiologic imaging studies. Presently he is supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) of the National Institutes of Health under Award Number R01EB025174 as the Principal Investigator (PI) of the grant “Generalized Obuchowski-Rockette Methodology for Analysis of Radiologic Diagnostic Imaging Studies.”