Computer Aided Detection (CAD) Systems for Mammography and the Use of GRID in Medicine

Abstract

It is well known that the most effective way to defeat breast cancer is early detection, as surgery and medical therapies are more efficient when the disease is diagnosed at an early stage. The principal diagnostic technique for breast cancer detection is X-ray mammography. Screening programs have been introduced in many European countries to invite women to have periodic radiological breast examinations. In such screenings, radiologists are often required to examine large numbers of mammograms with a double reading, that is, two radiologists examine the images independently and then compare their results. In this way an increment in sensitivity (the rate of correctly identified images with a lesion) of up to 15% is obtained.1,2 In most radiological centres, it is a rarity to find two radiologists to examine each report. In recent years different Computer Aided Detection (CAD) systems have been developed as a support to radiologists working in mammography: one may hope that the “second opinion” provided by CAD might represent a lower cost alternative to improve the diagnosis. At present, four CAD systems have obtained the FDA approval in the USA. † Studies3,4 show an increment in sensitivity when CAD systems are used. Freer and Ulissey in 2001 5 demonstrated that the use of a commercial CAD system (ImageChecker M1000, R2 Technology) increases the number of cancers detected up to 19.5% with little increment in recall rate. Ciatto et al.,5 in a study simulating a double reading with a commercial CAD system (SecondLook‡), showed a moderate increment in sensitivity while reducing specificity (the rate of correctly identified images without a lesion). Notwithstanding these optimistic results, there is an ongoing debate to define the advantages of the use of CAD as second reader: the main limits underlined, e.g., by Nishikawa6 are that retrospective studies are considered much too optimistic and that clinical studies must be performed to demonstrate a statistically significant benefit from the use of CAD.