Artifacts caused by patient breathing and movement during PET data acquisition affect image quality. Respiratory gating has been proposed to gate the list-mode PET data into multiple bins over a respiratory cycle. Non-rigid registration of respiratory-gated PET images can reduce the motion artifacts and preserve the count statistics, but it is time consuming. In this work, we propose an unsupervised non-rigid image registration framework using deep learning. We use a differentiable spatial transformer layer to warp the source image to the target image and use a stacked structure for deformation field refinement. Estimated deformation fields were incorporated into an iterative image reconstruction algorithm to perform motion compensated PET image reconstruction. We validated the proposed method using simulation and clinical data and showed its ability to reduce the motion artifact in PET images.
The image quality needed for CT-based attenuation correction (CTAC) is significantly lower than what is used currently for diagnostic CT imaging. Consequently, the X-ray dose required for sufficient image quality with CTAC is relatively small, potentially smaller than the lowest X-ray dose clinical CT scanners can provide. Operating modes have been proposed in which the X-rays are periodically turned on and off during the scan in order to reduce X-ray dose. This study reviews the different methods by which X-rays can be modulated in a CT scanner, and assesses their adequacy for lowdose acquisitions as required for CTAC. Calculations and experimental data are provided to exemplify selected X-ray pulsing scenarios. Our analysis shows that low-dose pulsing is possible but challenging with clinically available CT tubes. Alternative X-ray tube designs would lift this restriction.