Misregistration correction of data-driven gated (DDG) PET/CT by self-gated PET and deformable image registration CT

Introduction: Mis-registration between CT and PET data can result in mis-localization and inaccurate quantification of a functional uptake in whole body PET/CT imaging. This problem is exacerbated in situations when an abnormal inspiration occurred during the free-breathing CT scan for attenuation correction (AC) of the data-driven gated (DDG) PET data, which is typically derived from the expiratory phase of a breathing cycle. The aims of this study were (1) to design a deformable image registration (DIR) between the mis-registered CT and PET to improve DDG PET registration and quantification based on the respiratory motion model derived from the self-gated PET data and (2) to compare the results of DIR with the ground-truth DDG PET/CT data (DDG PET and DDG CT¹) in quantification and localization.

Methods: Ten whole-body ¹⁸F-FDG PET/CT scans with 17 clinically identified lesions in the misregistered regions were acquired for 2 min per 25-cm on the GE Discovery MI in this study. List-mode data were first analyzed with a principal components analysis (PCA)-based DDG and separated into the end-inspiratory (EI) and end-expiratiory (EE) PET data, derived from -10 to 15% and 30 to 80% of the breathing cycle, respectively. A multi-resolution deformable registration with spatio-temporal priors was used for estimate of a motion model from the EE to EI phase of the PET data, which was then used to generate the PET images at any phase of up to four times the amplitude of EE to EI for correlation with the misregistered CT. Once a matched phase of the DIR PET was determined, the misregistered CT was deformed according to the motion model to a pseudo CT at the EE phase, which was compared with the ground truth DDG CT without misregistration for quantification and localization of the DDG PET data.¹

Results: SUV increases of 10.2±9.7% and 18.1±14.6% for DIR CT and DDG CT, respectively, over the baseline misregistered CT for quantification of the DDG PET data. There was a statistically significant difference between DIR CT and DDG CT (p=0.0063), suggesting DIR CT can improve quantification of the DDG PET data by reducing but not completely removing misregistration between CT and DDG PET. For localization, DIR could improve registration, however DIR was not effective in misregistration correction in the areas where an uptake was nearby the border of two organs such as the lungs and the rib cage, the lungs and the liver, or the gall bladder and the liver.

Conclusions: DIR based on the PET motion model from the gated PET data could improve quantification and localization of the DDG PET data over the baseline DDG PET data. However, in both quantification and localization, DIR was not as effective as the DDG CT in the areas where an uptake was nearby the border of two organs.

• Download figure
• Open in new tab
• Download powerpoint