Compensation du mouvement respiratoire en TEP/TDM à l'aide de la super-résolution. (Application of super-resolution to respiratory gated positron emission tomography)

POSITRON Emission Tomography (PET) is widely used today for diagnosis and therapy follow up assessment in oncology. Novel applications like image guided radiotherapy are flourishing. However, PET images are of poor quality due to noise and blur. Respiratory motion is a major source of reduced quality, particularly when considering thoracic imaging. In order to minimize its effects, it has been suggested to synchronize the PET acquisition with the respiratory cycle. Data thus acquired can be separated into several bins, depending on the respiratory state during which they were acquired. Data acquired during the same cycle can then be summed and reconstructed. The resulting frames, however, are of low quality as each contains only a fraction of the overall acquired information. Motion compensation approaches have been developed, combining these synchronized datasets to a particular part of the respiratory cycle, hence making use of all the available statistics. Most existing compensation techniques follow a register-and-sum approach, which can be performed either before, during or after reconstruction. The resulting image is free of motion and is similar to the one that could have been reconstructed in the absence of motion. Super-resolution techniques aim at enhancing an image belonging to a sequence of images representing different looks at the same scene. By making use of the additional spatiotemporal information available in the original sequence, these techniques produce images with wider bandwidth than that of any of the individual original frames and free of field aliasing. The goal of this thesis is to apply such a technique to respiratory motion compensation. First, we developed a novel application of a known super-resolution algorithm by applying it to a respiratory gated sequence, and demonstrated that this algorithm can yield significant image quality enhancement of the individual respiratory gated PET frames. Since image-based motion correction algorithms often yield sub-optimal results, this algorithm was then incorporated within PET reconstruction. Images thus reconstructed are of higher quality than the ones on which super-resolution was applied after reconstruction. Finally, we showed that super-resolution motion correction leads to increased accuracy when planning radiotherapy treatment.