3D registration of SPECT and MR images by a multi-resolution approach

Summary form only given. A surface fitting algorithm has been extended in multiresolution space and a sequentially terminated distance function defined to be minimized in this multiresolution hierarchical search space. The brain surface was used to register SPECT (single photon emission computered tomography) (HMPAO) data with MRI (magnetic resonance imaging). For brain surface where a large lesion is present, cubic interpolation methods facilitate reconstructing and recapturing the normal appearance of the defective area of the brain surface by approximating parameter curve segments in 2-D, or surface patches in 3-D space. Several experiments were performed to assess the proposed technique and to evaluate the required sampling and threshold criteria using arbitrary and known misregistered SPECT data, with very satisfactory results.<>