Skip to Main Content
The purpose of this study is to improve CT-based attenuation correction of 3D PET data using an attenuation map derived from non-rigidly transformed 3D CT data. Utilizing the 4D NURBS-based cardiac-torso (NCAT) phantom with a realistic respiratory model based on high-resolution respiratory-gated CT data, we develop a method to non-rigidly transform 3D CT data obtained during a single breath hold to match that of 3D PET emission data of the same patient obtained over a longer acquisition time and many respiratory cycles. For patients who underwent 3D CT and PET (transmission and emission) studies, the 3D anatomy of the NCAT phantom was first fit to that revealed through automatic segmentation of the 3D CT data. From the 3D PET emission data, a second body outline was segmented using an automatic algorithm. Using the 4D NCAT respiratory model, the morphed 3D NCAT phantom was transformed such that its body outline provided the best match with that obtained from the 3D PET emission data. The other organs of the NCAT followed the corresponding transformations provided by the 4D respiratory model. The transformations were then applied to the 3D CT image data to form the attenuation map to be used for attenuation correction. For eight preliminary sets of patient data, the NCAT respiratory model allowed excellent registration of the 3D CT and PET transmission data as visually assessed by 3 independent observers. Minor registration errors occurred near the diaphragm and lung walls. The 4D NCAT phantom with a realistic model of the respiratory motion was found to be a valuable tool in a non-rigid warping method to improve CT-PET image fusion. The improved fusion provides for a more accurate CT-based attenuation correction of 3D PET image data.