Abstract
370
Objectives To reduce motion blur and attenuation map mismatch in thoracic and abdominal PET.
Methods Simultaneous PET/MR examinations were performed on a Biograph mMR scanner (Siemens Healthcare, Erlangen, Germany) in 2 patients with metastatic cancer (lung and abdominal manifestations). Single-bed PET list-mode data were acquired for 5 minutes 3 hours after injection of 370 MBq FDG. During the first 3 min of the PET scan the respiration-related tissue deformation in the PET field of view was recorded using sagittal 2D multi-slice MRI. Each slice was measured 12 times at an acquisition time of 0.4 s per frame (4.8 s per slice) and a spatial resolution of 2 x 2 mm2. Slices were reordered to obtain 4D tissue data under free breathing. These frames were co-registered using non-rigid registration to obtain a respiration motion model. PET data were gated into 4 respiratory gates according to amplitude of a continuously acquired MR navigator as the respiration signal. The MR-based end-expiratory patient attenuation map was deformed to match the 4 gates and images were reconstructed using an OSEM algorithm (21 subsets, 3 iterations). The gated images were co-registered using the motion model and summed up to obtain the final image. Uptake, contrast, volume (50% isocontour) and delineation (slope steepness and full width at half maximum) of seven tumor lesions were compared with and without motion correction.
Results Image quality and uptake quantification were significantly improved. The mean increase in maximum tracer uptake was 30.7 %, the mean contrast enhancement was 34.3 %. Lesion volume was reduced by an average 37.4 %. Lesion delineation was enhanced by 64.6 % (slope) and 24.7 % (FWHM).
Conclusions MR-based correction of respiratory motion of PET data is clinically feasible and provides significant improvement of image quality and uptake quantification.
Research Support Supported by fortüne grant no. 1891-0-0