METHODS AND APPARATUS FOR COMPUTER AUTOMATED DIAGNOSIS OF MAMMOGRAM IMAGES

A system, method, and computer program product for computer analysis of lesions in digitized film-based and/or digital mammograms is described, wherein diagnostic information is combined from two different 2-D mammographic views with the information obtained from one view (or field of view) or mammographic position is processed with information obtained from a second (or a plurality of) related mammographic views to reduce false-positive findings (increase specificity) while preserving or improving diagnostic sensitivity. The digital mammograms or digitized film-based mammograms used, are those that are in current use, and those that conform to the requirements of the American College of Radiology and the Mammography Quality Standards Acts. In a preferred embodiment, a line constructed at the location of the chest wall (or parallel to the chest wall), the location of the nipple, and a line constructed perpendicular to the chest wall datum line and passing through the location of the nipple serve as reference datum across mammogram views. An algorithm locates suspicious lesions in each mammography view and evaluates the concordance of the 3-D spatial locations to rule out physically impossible false-positive findings, based on calculations of spatial relationships. Concordant findings are detected using anatomic landmarks and such findings are reported using terms that are currently in use by physicians and other health care providers in the field of mammography.