The work covered in this thesis deals with the problem of automatically aligning 3D MR images of the breast. The detection of breast lesions using MR scans can be assisted by subtraction of pre-and post-contrast injection images.  The occurrence of patient movement between the scanning periods can introduce artefacts in the subtracted image.  The approach taken was to use two registration software packages, Automated Image Registration (AIR) and the registration algorithm used in the Statistical Parametric Mapping (SPM) software, for the correction of any misalignment. These packages, which are based on different principles, are typically used by the research community in the registration of brain and other images. The first hypothesis was that the correction of the mis-registration between two MR breast data sets produced by these software packages is the same. The evaluation of the registration is often done by the observation, on a slice-by-slice basis, of the subtraction image.  Since this is a laborious task, and one that results in a purely qualitative assessment, the Joint Histogram (JH) was used as a two-dimensional representation of the difference between the two sets of images. A quantitative evaluation of the dispersion of JH values from the diagonal gives an indication of the total extent of the misalignment. Image data from two patients was used to test the registration algorithms, one with only minor deformation between scans, the other with larger deformations. In all the situations tested it was found that the AIR method produced a better alignment than that achieved using the SPM method, showing the first hypothesis to be false. The second hypothesis was that the validation using both breast models produce the same result. The first model indicates that AIR produce better results. However, the second model, which has the potential to introduce non-linear deformations that are anatomically more realistic, still has some limitations that need to be resolved before it is possible to arrive at an overall conclusion.