Abstract
1349
Objectives: Physiological respiratory motion during PET data acquisition degrades PET image quality and accuracy which may influence clinical decision making. Therefore, motion-compensated image reconstruction (MCIR) has been proposed to overcome this problem. In this study, a new ultrafast elastic motion compensation approach based on motion deblurring (MCMD) is evaluated on patient data and compared to standard MCIR technique.
Methods: CT and whole-body PET data (6-8 bed positions, duration 120s or 360s) were acquired on an integrated PET/CT (Siemens Biograph mCT), 1 h after injection of 18F-FDG (4MBq/kg body weight). During PET data acquisition the respiratory signal was recorded using the Anzai belt system (Anzai Medical Co.). Motion compensated images were reconstructed using both correction techniques, MCMD and MCIR. In the MCMD method, motion is corrected using a deblurring kernel during iterative image reconstruction, where the motion between the reference image (optimal gate) and the rest of the data is estimated using mass-preserving optical flow. In the MCIR method, 10 respiratory gated images were generated using the ANZAI signal. Motion vector fields between each respiratory gate and a predefined reference image were calculated using optical flow methods. Finally, a motion corrected image was reconstructed using a standard EM-based image reconstruction that incorporates elastic motion correction. To evaluate the effect of motion correction on quantitative data analysis, regions-of-interest defined on multiple lung lesions were analyzed and changes in SUVmax, SUVmean and metabolic 18F-FDG volume (both defined on 60% threshold of SUVmax) were reported.
Results: Data of 5 patients (57±7.5 y) with a total of 26 identified lung lesions were included in the comparative analysis. SUVmax of all lesions was 6.5 (median) with a range between 3.0 and 27.1. SUVmax values increased by (16.5±13.5)% (mean±SD) for the MCMD method and by (11.2±12.3)% by the MCIR technique. SUVmean values increased similarly by (16.7±14.9)% (MCMD) and (8.0±20.2)% (MCIR). The metabolic volumes of the lesions after MCMD motion compensation decreased to (68.9±23.2)% of the volume without motion compensation, whereas the metabolic volumes after MCIR motion compensation decreased to (81.9±26.7)%.
Conclusion: Motion compensation based on MCMD shows a clear quantitative improvement on PET images with an increase of SUV values and a decrease of the metabolic volumes. Results are comparable to those of MCIR motion compensation techniques. Effects are smaller using MCIR technique which may be related to non-optimized parameters in optical flow motion estimation. Research Support: 1. Hong, Inki, Judson Jones, and Michael Casey. "Ultrafast Elastic Motion Correction via Motion Deblurring." 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), 2014. 2. Koesters, Thomas, Klaus P. Schaefers, and Frank Wuebbeling. "EMrecon: An expectation maximization based image reconstruction framework for emission tomography data." Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), 2011.