Atlas-based attenuation correction for small animal PET/MRI scanners

Small animal PET/MRI scanners producing anatomically co-registered simultaneously-acquired images of morphology, function and metabolic activity have become available and are expected to have a huge positive impact on the pre-clinical imaging field. Attenuation correction (AC) necessary for accurate quantification of PET signals is challenging in these scanners because the measured anatomical map from MRI is based on proton densities and not on electron densities. We propose a deformable mouse atlas-based registration method for AC in small animal PET/MRI. In this method, we first match the posture of the atlas to the posture of the mouse being imaged using landmark and elasticity constraints. The asymmetric L 2 pseudo-distance between the atlas surface and the extracted mouse surface is then minimized in order to register the two surface data. A Sobolov prior is used to ensure smoothness of the warping field. This warping field is then extended to the entire mouse volume using elastic deformations. The computed transform is then applied to the CT of the mouse atlas. Mutual information-based rigid registration is maximized to spatially normalize the warped atlas CT to the mouse MRI. We evaluated the proposed method for co-registered MRI and CT data acquired for a normal nu/nu mouse. Four organs were manually segmented from the animal MRI. Dice coefficients of organ overlap were measured for these organs between the manually assigned labels and those estimated by the proposed method. Dice coefficients were also measured between the skull and skeleton extracted directly from the animal CT and those estimated using the proposed method. Organ overlap of > 90 % in the head and torso region of the animal and > 60 % in the abdominal region are observed. For the skull and skeleton, the measured overlap was > 90 %. Based on published values, we derive a detailed attenuation map at 511 keV for the mouse using the warping field estimated by the proposed method.