Three-dimensional mass reconstruction in mammography

In this paper, we present a novel method for reconstructing the 3-D shapes of masses. We first use the Shape from Silhouettes technique to get an approximation to the 3-D shape of the mass. We calculate the centroid of the mass in the 2-D images and back-project them to get their intersection in 3-D. We then apply a novel iterative method, which is derived from ART (Algebraic Reconstruction Technique), to refine the 3-D shape of the mass. The thickness of the masses is calculated according to the h(int) representation. We use the thickness of the masses in the CC and MLO or LM views to refine the 3-D approximation to the reconstructed shape of the mass. We find that the mean deviation rate of the reconstruction of a pair of benign masses is much larger than that of malignant masses, which can be used as a criterion of classifying a mass into malignant or benign.