Abstract
516
Objectives: Clinical application of dynamic PET imaging routine is limited by long patient scan time and requirement of blood sampling. We have proposed a simplified dynamic FDG PET imaging protocol that combines shortened scan duration and population-based input function (PBIF) to increase clinical feasibility. The objective of this study is to optimize the simplified protocol and evaluate the performance of voxel-based indirect and direct parametric reconstruction with clinical data. Focusing on a late scan strategy, we evaluate the impact of different scan duration and start time (t[asterisk]) choice.
Methods: Volunteer scans were performed on the United Imaging uMI 510 PET/CT system in the Chinese PLA General Hospital. Firstly, 11 subjects (7 male, 4 female; 7 healthy, 4 with cancer lesion; age 15-73y) were scanned and the data were used to generate a PBIF database. Right after injection of 0.12 mCi/kg FDG, dynamic FDG scans started with 20-min cardiac scan, followed by 2-pass 5-bed scan (each scanned for 4 min) started from 40 min post injection. Image-derived input functions (IDIFs) were measured from the dynamic images to generate the PBIF template. Secondly, another 5 subjects (3 male, 2 female; 2 healthy, 3 with cancer lesion; age 43-65y) received 80-min single-bed cardiac PET/CT scans post injection of 0.12 mCi/kg FDG. These data were used to perform voxel-based dynamic analysis. The parametric Ki images were reconstructed through indirect and direct approaches. In the indirect method, dynamic images were reconstructed with OSEM method (24 subsets, 3 iterations). Attenuation, normalization, random, decay, and scatter corrections were performed. Patlak plot was drawn on each voxel of dynamic images to generate the indirectly reconstructed Ki images. In the direct approach, nested Patlak expectation-maximization algorithm was applied to reconstruct Ki images directly from sinograms. 20 inner loops of parametric image update were nested into each outer loop of dynamic image update (3 outer iterations). A scaling factor is required in PBIF based analysis. In the indirect approach, the scaling factor was directly measured from the aorta region in the dynamic images. In the direct approach, the dynamic image series and an unscaled parametric image were both generated from the direct reconstruction with unscaled PBIF template. A scaling factor was determined from the dynamic image series, and then was applied to the parametric image. To investigate the simplified acquisition protocols, regional analysis with complete IDIF was also performed on the regions of interest (ROIs) as standard Ki. For different combinations of duration and t[asterisk], correlations were calculated between the area under curve (AUC) of scaled PBIF and IDIF. The correlations were also calculated between the estimated mean Ki in ROIs and standard Ki.
Results: Three acquisition protocols were compared: (1) 20-80 min (2) 20-50 min (3) 50-80 min on the 5 single-bed-imaging subjects. The visual image quality of directly reconstructed Ki images were obviously better than that from indirectly reconstruction. The AUC values of PBIF were highly correlated with those of IDIF, and showed the best results for the 20-50 min protocol. In the liver ROIs, Ki values estimated from direct analysis were consistently larger than those estimated from indirect analysis. The correlation coefficients were similar for both direct and indirect approaches. The best correlation was found in Ki values with the 20-50 min protocol.
Conclusions: Simplified acquisition protocols were compared based on indirect and direct Patlak analysis of clinical dynamic FDG PET scans. Good correlation was found between the scaled PBIF and IDIF. Consistent difference was found between indirectly and directly estimated Ki values. Preliminary results showed 20-50 min protocol was the optimal protocol. Further investigation are en route to include more volunteers in the study and compare other combination of scan duration and t[asterisk].