Speaker
Description
(2) Material and Methods
For the registration, the breast is modeled with a 4-node tetrahedron mesh extracted from a volume modality (USCT or MRI). The model can be derived either from the segmented MRI volume with assigned material stiffness parameters or from a stiffness distribution model based on the USCT sound speed images. In the case of MRI to USCT and USCT to X-ray mammography registration, the buoyancy effect respectively the compression are simulated on the model within a FEM framework. Registration is done by optimizing the most influencing parameters. After applying the deformation field to the volume modality, the co-registered volumes are rigidly aligned at their centres of mass.
(3) Results
An average registration error below 5 and 12 mm for MRI to USCT and USCT to mammography registration, respectively, allowed us to evaluate the diagnostic performance of USCT. It was shown that regions of high sound speed corresponded well with the tumor position indicated in MRI from a contrast kinetic analysis. Moreover, the quantitative analysis of sound speed and attenuation parameters with respect to segmented mammograms revealed that sound speed gives a better distinction between breast tissues than attenuation, whereas their combined information further improved the classification. A clear distinction was observed between lesions and other breast tissue.
(4) Discussion and Conclusion
The first evaluation of USCT's diagnostic value with the proposed registration approach already showed promising results. Further evaluations are planned with additional patient datasets and with more complex modelling, e.g. dealing with the MRI breast coil deformations and comparing the performance of stiffness distribution models derived from speed sound images to the models derived from segmented MRI. Additionally, registration to the emerging X-ray tomosynthesis modality is of high interest.
