Accurately reconstructing the three-dimensional mass shapes in mammographic images is important for classifying the abnormality into malignant or benign. In this paper, a partial volume compensated reconstruction technique for mass shapes is presented. The two-dimensional shapes of the masses are first automatically segmented using a region growing approach. The 3D mass shapes are then iteratively refined according to an algebraic reconstruction technique. Partial volume estimation is applied on the boundary to get a smoother 3D shape. Evaluation results show that the proposed algorithm improves the accuracy of the mass reconstruction.