ABSTRACT
One of the main advantages of acquiring radiologic images digitally and displaying them on cathode ray tube (CRT) monitors is that image-processing tools and computer-aided detection (CAD) schemes can be readily implemented. Thus, there has been a proliferation of image processing (1-2) and CAD (3-5) tools developed in the past 20 years. With the advent of digital acquisition technologies in chest imaging (i.e., computed radiography [CR]) and the recent progress in digital mammography, clinical implementation of CAD seems even more probable than in the past. Image-processing tools are often used regularly in certain clinical situations now (e.g., in computed tomography [CT], magnetic resonance imaging [MRI], and ultrasonography [US], where images are digitally acquired and are often displayed on and diagnosed from a CRT monitor), but for the most part CAD tools are used only in experimental situations. Part of the problem with implementing CAD and image-processing tools in the clinical environment is that only a few, limited studies have been conducted to demonstrate reliably their usefulness in improving observer performance. For the most part, results from the use of CAD and image processing in the clinical environment have been equivocal (6-10). Sometimes CAD or image processing aids performance, sometimes it has little or no effect, and sometimes it actually hurts overall performance. Before getting into the specific applications of CAD and the clinical implementation of CAD, a brief explanation of why and how CAD might be useful to the radiologist is in order.
