Abstract
In this paper, we propose a registration method for cardiac CT and SPECT. There are fewer correlations between CT and SPECT. Hence it is difficult to register CT to SPECT directly by referencing pixel values. Consequently we propose 2-step registration method using model data which is called “Atlas data". Atlas data has two kinds of information, the first one is normalized SPECT data which is made by averaging some of patient's SPECT. The other is cardiac anatomical data. They include a long and short axes of left ventricle. In the first step of 2-step registration, CT and atlas data are aligned. User needs to set up the axis that is the same as that of atlas data. When it is set, the alignment problem of CT and atlas is solved easily because it just transforms CT axes onto atlas axes. In the second step of the registration, atlas data is aligned to SPECT. In this step, we use general registration method that uses correlation coefficient as metric, and rigid transformation as transformation. Through these steps, CT images are registered to SPECT images via atlas. All case of experimental registration worked well for diagnosis of ischemic heart disease.
