Smooth transformation of medical images by competition and consensus of displacement vectors

A technique of fitting the MR images of different subjects to a standard brain image is proposed. Proposed neural network will automatically find out the best displacement vectors, keeping a consensus within the neighboring segments. In the mapping procedure, if necessary, some of the coordination points, e.g., anterior and posterior commissural points can be manually marked to force them fit exactly. This method can be applied to other sorts of medical images to get better interpretation of the images, both for clinical and for scientific purposes. (author)