Automatic breast CAD results for seven categories of abnormalities.

5 Background: This prospective study was conducted to determine efficacy of preclinical use of automatic computer-aided cancer detection (CAD) system. Patients were women of high risk referred for MRI of breast with goals of assessing sensitivity/specificity of an automatic CAD platform which used kinetic and morphological (spiculation, volume, solidity, surface area, elongation, maximum extent) tools. We assessed classification of abnormality using histopathology as our gold standard.,1,976 patients with 2,149 abnormalities included were scanned from 2007 to 2010. Patient inclusion criteria were women: with dense breasts; BRCA1 or BRCA2 genes; family history of cancer; suspected of having recurrent cancer; anomalous mammogram/ultrasound; pain; or clinical suspicion. All patients included gave institutionally IRB approved HIPAA compliant informed consent. CAD software post-processing was automatically applied to dynamic contrast-enhanced MRI (DCE-MRI). One pre- and 4 post-scan series were used to calculate wash-in/out parameters. System's automatic detection was applied to entire breast(s) volume. No manually placed regions of interest were used.,Of 1,976 women scanned, 361 had mastectomies, 17 had bilateral mastectomies, 488 had lumpectomies/excisional biopsies, 623 had core biopsies, 166 had non-surgical approaches; radiation or chemotherapy. Mean tumor resected was 2.95 cm; SD=2.88 cm, while CAD detected breast tumor calculated size was 3.20 cm; SD=2.66 cm.,Automatic CAD applied to breast DCE-MRI demonstrated high sensitivity, reasonable specificity, and accurate estimates for location within breast, and volume of malignancy. Automatic 3D CAD translucent display will aid oncology planners in tumor localization, volume and extent estimations. [Table: see text].