Comparison of Motion Correction Methods Incorporating Motion Modelling for PET/CT Using a Single Breath Hold Attenuation Map

Introducing motion models into respiratory motion correction methods can lead to a reduction in blurring and artefacts. However, the pool of research where motion modelling methods are applied to combined positron emission tomography and computed tomography is relatively shallow. Previous work used non-attenuation corrected time-of-flight data to fit motion models, not only to motion correct the volumes themselves, but also to warp a single attenuation map to the positions of the initial gated data. This work seeks to extend previous work to offer a comparison of respiratory motion correction methods, not only with and without motion models, but also to compare pair-wise and group-wise registration techniques, on simulation data, in a low count scenario, where the attenuation map is from a pseudo-breath hold acquisition. To test the methods, 4-Dimensional Extended Cardiac Torso images are constructed, simulated and reconstructed without attenuation correction, then motion corrected using one of pair-wise, pair-wise with motion model, group-wise and group-wise with motion model registration. Next these motion corrected volumes are registered to the breath hold attenuation map. The positron emission tomography data are then reconstructed using deformed attenuation maps and motion corrected. Evaluation compares the results of these methods against non-motion corrected and motion free examples. Results indicate that the incorporation of motion models and group-wise registration, improves contrast and quantification.