New developments in observer performance methodology in medical imaging.

摘要: A common task in medical imaging is assessing whether a new system, or variant of an existing one, improvement over technology. Imaging systems are generally quite complex, consisting several components—for example, image acquisition hardware, processing and display hardware software, interpretation by radiologists- each which can affect performance. Although it may appear odd to include the radiologist as "component" chain, because radiologist's decision determines subsequent patient care, effect human has be included. Physical measurements such modulation transfer function, signal-to-noise ratio, useful for characterizing nonhuman parts chain under idealized often unrealistic conditions, uniform background phantoms target objects with sharp edges. Measuring performance entire including radiologist, using real clinical images requires different methods that fall rubric observer "ROC" analysis, involve collecting rating data on images. The purpose this work review recent developments field, particularly respect free-response method, where location information also collected.