M.A. Rossman, F. Candocia, M. Adjouadi, P. Jayakar, and I. Yaylali (USA)
Affine Transformation, SPECT, Registration, Medical Images
This paper describes an intensity-based image registration algorithm using the affine transformation. The images considered are Single Photon Emission Computed Tomography (SPECT) brain images provided by the Neuroscience group at Miami Children’s Hospital. Registration must be performed on time-series images for which any meaningful pixel-to-pixel comparisons will be made. This is done so that the images are brought into spatial alignment and made to share a common coordinate system. It will be demonstrated that the algorithm implemented here gives comparable results to those obtained from an image registration program currently in clinical use, and yet utilizes a less constrained mathematical approach. The basis of the clinical registration program is the AIR (Automated Image Registration) software package designed by Roger Woods and colleagues at the UCLA Medical School. The comparisons made between the registration results obtained from the affine algorithm and the commercial package will be demonstrated visually with 3-D rendered graphics and numerically by the use of mean square error. It will be argued that even though the rigid-body transformation is often suggested for intra-modality medical image registration, the affine transformation is a more general mathematical solution to the misalignment problem that gives comparable results without the additional complexity.
Important Links:
Go Back
