Several full-field digital mammography (FFDM) systems have been approved for clinical applications. It is important to develop a CAD system that can easily be adapted to images acquired by FFDM systems from different manufacturers. To develop a CAD system that is independent of the FFDM manufacturer's proprietary preprocessing methods, we used the raw FFDM image as input and developed a multi-resolution preprocessing scheme for image enhancement. Our CAD system performed prescreening to identify mass candidates, segmented the suspicious structures, extracted morphological and texture features, and then classified masses and normal tissue. In this study, we investigated the use of a two-stage gradient field analysis to identify suspicious masses, and the effectiveness of a new gradient field feature extracted from each suspicious object for false positive (FP) reduction. A data set of 104 cases with 243 images acquired with a GE FFDM system was collected. Most cases had two mammographic views, except for 12 cases that had three views and 1 case with only one view. The data set contained 106 masses. The true locations of the masses were identified by an experienced radiologist. Using free-response receiver operating characteristic (FROC) analysis, it was found that our CAD system achieved a cased-based sensitivity of 70%, 80%, and 88% at 0.8, 1.3, and 1.7 FP marks/image, respectively. The high performance indicated the usefulness of the new gradient field analysis method.