Toward a strategy for consensus development on a quantitative approach to medical imaging.

This Guest Editorial is prompted by the article “A Quadratic Model for Combining Quantitative Diagnostic Assessments From Radiologist and Computer in Computeraided Diagnosis (CAD)” by Yulei Jiang and Charles Metz in the this issue of Academic Radiology (1). The report calls to mind a curiosity of research in contemporary computer-aided diagnosis and raises fundamental issues for further advances in medical image science. (Readers seeking more formal definitions of concepts and literature citations beyond those given here may refer to [2].) Here is the curious point. Sometimes it is found that radiologists using a computer aid perform better than unaided radiologists, yet computer-aided radiologists are outperformed by the computer in its stand-alone mode. A common reaction to this situation is to raise the challenge for radiologists to learn to perform with the computer as well as the computer performs alone. However, the investigators of the current report (1) remind us that as long as there is some degree of independence between the information that radiologists extract from the image and the information that the computer extracts from the image, the ultimate challenge is rather for the combined radiologist–computer system to perform better than either the unaided radiologist or the stand-alone computer. The investigators provide a quantitative formulation of a method