作者: Adam Wunderlich , Craig K. Abbey
DOI: 10.1118/1.4823755
关键词:
摘要: Purpose: Studies of lesion detectability are often carried out to evaluate medical imaging technology. For such studies, several approaches have been proposed measure observer performance, as the receiver operating characteristic (ROC), localization ROC (LROC), free-response (FROC), alternative (AFROC), and exponentially transformed FROC (EFROC) paradigms. Therefore, an experimenter seeking carry a study is confronted with array choices. Traditionally, arguments for different made on basis practical considerations (statistical power, etc.) or gross level analysis (case-level lesion-level). This article contends that careful consideration utility should form rationale matching assessment paradigm clinical task interest. Methods: In theory, performance commonly evaluated total expected utility, which integrates various event utilities against probability each event. To formalize relationship between summary curve associated paradigm, concept “natural” structure proposed. A natural defined when variables axes sufficient computing assuming disease prevalence known. Results: Natural structures ROC, LROC, FROC, AFROC, EFROC curves introduced, clarifying how correct incorrect decisions aggregated by curves. Further, conditions given under general localization-based methodologies reduce case-based assessment. Conclusions: Overall, findings reveal correspond diagnostic tasks, suggesting motivating principle choosing paradigm.