PET, CT, and MR image registration of the rat brain and skull

Spatially registered positron emission tomography (PET), computed tomography (CT), and magnetic resonance (MR) images of the same small animal offer potential advantages over PET alone: CT images should allow accurate, nearly noise-free correction of the PET image data for attenuation; the CT or MR images should permit more certain identification of structures evident in the PET images; and CT images provide a priori anatomical information that may be of use with resolution-improving image-reconstruction algorithms that model the PET imaging process. However, image registration algorithms effective in human studies have not been characterized in the small-animal setting. Accordingly, we evaluated the ability of the automated image registration (AIR) and mutual information (MI) algorithms to register PET images of the rat skull and brain to CT or MR images of the same animal. External fiducial marks visible in all three modalities were used to estimate residual errors after registration. The AIR algorithm registered PET bone-to-CT bone images with a maximum error of less than 1.0 mm. The registration errors for PET brain-to-CT brain images, however, were greater, and considerable user intervention was required prior to registration. The AIR algorithm either failed or required excessive user intervention to register PET and MR brain images. In contrast, the MI algorithm yielded smaller registration errors in all scenarios with little user intervention. The MI algorithm thus appears to be a more robust method for registering PET, CT, and MR images of the rat head.