Methods for Evaluating the Effectiveness of Screening Mammography Are Not Necessarily Valid for Evaluating the Effectiveness of Computer-Aided Detection in Screening Mammography

摘要

It is important that the clinical effectiveness of computer-aided detection (CADe) systems be evaluated. Since CADe is used to improve the effectiveness of screening mammography, it might be expected that the same methods and endpoints that are used to evaluate screening mammography can be used to evaluated screening mammography when CADe is used. Unfortunately, this is not always true and when the assumption fails, erroneous conclusions are often made. In clinical studies the choice of endpoints, estimation of sensitivity, and the significance of DCIS are potential problem areas. Further, the use of ROC versus FROC can affect the measured performance.