X-ray Mammography - MRI Registration Using a Volume-Preserving Affine Transformation and an EM-MRF for Breast Tissue Classification

T. Mertzanidou, J. H. Hipwell, M. J. Cardoso, C. Tanner, S. Ourselin, D. J. Hawkes
Proc. Int. Workshop on Digital Mammography
2010

---

Abstract

Registration of MR volumes to X-ray mammograms is a clinically valuable task, as each modality provides complementary information on normal and abnormal breast tissue structure and function. We propose an intensity-based technique with a 3D volume-preserving affine transformation. An important part of our framework is the use of an Expectation-Maximization (EM) algorithm, with a Markov Random Field (MRF) regularization, that is used for breast tissue classification and subsequently the mapping of the MR intensities to X-ray attenuation. Initially, the proposed framework was tested on simulated X-ray data, where the goal was to register the original undeformed MRI to a simulated X-ray that was produced using a real compression image, acquired from volunteers in the MR scanner (8 cases). Since the ground truth in this case can be estimated from individually defined landmarks, we have evaluated the mean reprojection error, which was 3.83mm. The algorithm was then applied and evaluated visually on 5 cases that had both X-ray mammograms and MRIs.

---

Link to publisher's page

---

@InProceedings{eth_biwi_00808,
  author = {T. Mertzanidou and J. H. Hipwell and M. J. Cardoso and C. Tanner and S. Ourselin and D. J. Hawkes},
  title = {X-ray Mammography - MRI Registration Using a Volume-Preserving Affine Transformation and an EM-MRF for Breast Tissue Classification},
  booktitle = {Proc. Int. Workshop on Digital Mammography},
  year = {2010},
  pages = {23-30},
  volume = {6136},
  series = {LNCS},
  publisher = {Springer},
  keywords = {}
}