MR-SPECT coregistration using intensity matching and a CBF simulation model

Image coregistration helps to assign anatomic location to relatively fuzzy images obtained by single photon emission computed tomography (SPECT). We present a method for SPECT coregistration to magnetic resonance (MR) volume based on the MR image segmentation into a simple four-compartmental model including extracranial tissue followed by labeling with ideal blood flow values and filtering that simulate the generation of idealized SPECT data. The model and the real CBF volume are matched iteratively with the cost function based on intensity correlation and amount of matching pixels above preset noise level. The method finds-the largest model-conforming part of the image independently of variation in localized disturbances. It is efficient in speed and insensitive to wide variation of model parameters.