Computer-assisted mammography in clinical practice: Another set of problems to solve

To be adopted in radiological practice, computer-assisted diagnosis must address a domain of realistic complexity and have a high performance in terms of speed and reliability. Use of a microcomputer-based system of mammographic diagnoses employing discriminant function analysis resulted in significantly fewer false-positive diagnoses while producing a similar level of correct diagnoses of cancer as normal reporting. Although such a system is a valuable teaching aid, its clinical use is constrained by the problems of unambiguously codifying descriptors, data entry time, and the tendency of radiologists to override predicted diagnoses which conflict with their own