Detections of Microcalcification Clusters Using Multiple Mammographic Views

To reduce the false positive hits, a new method for the computer aided detection of microcalcification clusters is proposed in the paper by joint analysis of two views of the same breast. The novelty of the scheme includes a consequent two steps of matching processes: spatial and feature matching. The former links a suspicious cluster located on the MLO view with a corresponding location on the CC view using their spatial information to form a paired cluster, and then in the latter stage, each cluster candidates are characterized by its single-view features such as size, shape and intensity. Finally a similarity function is calculated between the pair to determine if they were true microcalcification clusters. The experiments show that the proposed method has advantages of lower FP rate compared to the one on a single view.