ABSTRACT
Traditional ROC analysis focuses on assessment of the ability of a diagnostic system to reproduce the results of a “binary gold standard.” In the simplest scenario, a binary gold standard indicates the presence or absence of a specic condition (e.g., abnormality) in a specic subject. In other scenarios, it may indicate the locations of multiple abnormalities within an subject, hence associating any specic location with a binary indication of the presence or absence of a certain abnormality (see Chapter 8). In this chapter, we discuss a simple case in which the gold standard indicates a binary truth (normal or abnormal) for each subject (see Chapter 1). Note that the principles and concepts discussed here extend to more general scenarios.
