A novel MRI-SPECT registration scheme using eight similarity measures

In this paper, we propose a novel image registration scheme for MRI/SPECT registration. First, MRI is decomposed into some different brain maps by fuzzy c-means algorithm. Then, Gaussian convolution is performed with the map represented gray and white matter to simulate the radiation phenomenon of SPECT. Finally, the MRI/SPECT registration is carried out by optimizing a similarity measure. Finally, the best transformation for registration is reached by Powell's direction set method. In order to determine the most appropriate similarity measure, eight similarity measures are evaluated by performing the proposed registration scheme. The results show that Pattern Intensity has best results, and Mean Square Difference of Intensities is the worst.