Computational modeling of the breast during mammography for tumor tracking

Mammography is currently recognized as the gold standard for screening and diagnosis of breast cancer. A number of non-rigid registration algorithms have been used to track regions of interest across 2D mammographic images (cranio-caudal and mediolateral-oblique views). However, such techniques typically rely solely on the image properties A modeling framework is presented to potentially improve tumor tracking by constraining the image registration using physical laws of soft tissue mechanics. A simplified phantom model was constructed using an incompressible, homogeneous and isotropic silicon gel, modeled as a hyperelastic neo-Hookean material. The material constant was estimated using a nonlinear least-squares optimization technique to minimize errors between predicted displacements of material points in a large deformation finite element (FE) model and the corresponding experimentally observed displacements under gravity loading. The gel phantom was compressed between two plates to mimic a typical mammographic procedure and the deformed surfaces were scanned. Contact constraints were used to simulate compression in the FE model and the predicted displacements agreed well with the experimentally observed deformation. We also found that the effects of gravity markedly affected the accuracy of the compression model results. We conclude that modeling the soft tissue mechanics of the breast can provide a useful tool for tracking possible tumors from the compressed state (during mammography) to other configurations for further examination.