[Validation of an entropy-based algorithm for registration of serial 3D MR mammography data]

The aim of this study was to develop and implement an algorithm for the co-registration of 3D breast MRI sets acquired at two slightly different patient positions (repetitive examination). Combined translation and rotation with locally varying parameters were applied for the purpose of coordinate transformation. A phantom allowing selective changes of the volume of the glandular tissue model was developed, in order to prove the robustness of the proposed matcher against local changes. Serial 3D data sets of phantoms and volunteers were acquired to validate the routines. Co-registration was performed using mutual information (MI) as a similarity measure of the matching of the acquired images. In the phantom study, the phantom was deliberately shifted and rotated around horizontal and vertical axes. Starting the registration with global translations using a rigid matcher, the horizontal (phi) and vertical (theta) rotation angles were optimized in an iteration loop for each slice. This method was then applied to the breast data sets. Application of the algorithm on serial 3D MR data sets improved the co-registration especially in consideration of varying local tissue volumes. The algorithm represents a compromise between a pure rigid and an elastic 3D matcher.