Non-rigid Registration of Functional to Anatomical Brain MR Images Using Constrained Optimization
Briggs R, Devous M, Haley R, Gholipour A, Kehtarnavaz N, Gopinath K
University of Texas Southwestern Medical Center, University of Texas Southwestern Medical Center
Accurate registration of functional to anatomical brain MR images is needed for brain functional localization. Due to the effect of nonlinear distortions in fast functional echo planar imaging, even after rigid registration and resampling, mismatches of a few millimeters may occur in overlaying the activation maps from the functional image space to the anatomical image space. A non-rigid registration algorithm based on a bound constrained optimization technique, a Bspline free-form deformation model, and the mutual information similarity measure has been developed in this study. The experiments show that this algorithm provides practically significant enhancement in the registration of functional to anatomical brain MR images.