Human vs Machine -- Another Chapter in CAD for Mammography Debate

Background: Computer-aided detection (CAD) systems for breast imaging were designed to make suspicious annotations on a mammogram in order to prompt the radiologist to determine if the finding is significant. Objective: To determine the sensitivity of CAD systems for mammography when used alone without a radiologist's input and to compare it to the sensitivity of digital and film-screen mammography as derived from the Digital Mammographic Imaging Screening Trial (DMIST). Methods: Film-screen and digital mammograms were obtained from the American College of Radiology Imaging Network (ACRIN) image archive. Only cases that had a known cancer were evaluated, and both screen-film and digital mammograms were obtained for each case since the patients involved were part of the DMIST initiative. The CAD systems tested included: iCAD SecondLook, R2 ImageChecker Cenova version 1.0 for digital film, and R2 ImageChecker version 8.0 for screen-film. The default sensitivities of the programs were u more...