Access to eBooks is limited to institutions that have purchased or currently subscribe to the SPIE eBooks program. eBooks are not available via an individual subscription. SPIE books (print and digital) may be purchased individually on SPIE.Org.

Contact your librarian to recommend SPIE eBooks for your organization.
Abstract
Computer-aided detection or CAD refers to automated screening systems that localize suspicious regions in an image for a radiologist to consider (e.g., the computer marks the location of a suspicious mass or clustered calcifications using symbols on the screen). As the saying goes, "X marks the spot." These techniques can serve as a “second reader” to improve sensitivity such as to detect subtle lesions in mammography that might otherwise be missed by the radiologist. There has been a considerable body of work in this area over the past few decades. In fact, commercial detection systems are now available that reportedly identified 77% of overlooked breast malignancies and increased screening sensitivity by 20%. These automated detection systems are not constrained by the limits of human vision, should be more consistent, and have the potential to improve the performance of less experienced radiologists. Taking this one step further, predictive computer models have also been developed to characterize the status or make some recommendation for a lesion already detected by a radiologist or CAD system. Such diagnoses may go beyond a lesion to apply to a whole image, imaging study, or patient. For example, the model may predict whether a lesion is benign or malignant, or recommend a cancer patient for some therapy or not based on the likelihood of complications. This predictive modeling work is often called computer-aided diagnosis (CADx) or computer-aided classification or characterization (CAC). This is the focus of this chapter. Unfortunately, this characterization work is sometimes abbreviated to CAD, the same acronym as for detection. To add to this confusion, most detection and characterization algorithms actually include elements of the other, such that there is no clear delineation between them anymore. For example, an algorithm to detect suspicious breast masses must not only find regions of suspicion in a mammogram, but also consider features describing each region in order to predict which ones have the highest likelihood of cancer, so as to reduce the number of false positives. Both CAD and CADx often share the same tools for image processing, machine learning, and statistical analysis. Many people simply refer to the general field encompassing both areas as computer-aided diagnosis.
Online access to SPIE eBooks is limited to subscribing institutions.
CHAPTER 27
30 PAGES


SHARE
Back to Top