Room: AAPM ePoster Library
Block sequential regularized expectation maximization (BSREM), as opposed to OSEM reconstruction, has garnered significant interest for its ability to reach full image convergence while suppressing noise. Our goal was to design a realistic phantom experiment which can be used to optimize reconstruction parameters (ß) of the BSREM reconstruction for quantification of prostate cancer metastasis with [18F]F-DCFPyL. We also evaluated the image quality of the images reconstructed with this regularized algorithm.
18 simulated-lesions (diameters 3mm-16mm) were casted using Epoxy resin infused with 2 activity concentrations of [22Na]NaCl. These concentrations were chosen based on an analysis of prostate cancer patients images with [F-18]-DCFPyL.
The background of the phantom and the compartment simulating the bladder were filled with [18F]FDG to achieve target concentrations determined from the patient segmentations, 1.8kBq/mL and 176.4kBq/mL, respectively.
An 8-hours dynamic scan using a GE Discovery D690 PET/CT scanner was performed. Images were reconstructed with BSREM algorithm (?=2, ß=100,300,400,500,650,800, and 25 iterations). Lesions were segmented using the PET Edge+ method available in MIM software. The ground truth of activity concentrations in the 22Na lesions was determined from a scan with fully decayed background; performed 3 days after the dynamic scan. Segmentation bias, contrast, and activity recovery values were calculated for each size of sphere.
Bias in lesion volume ranged from 700% (smallest spheres) to 0% (largest sphere). ß=300 produced the best contrast recovery, 86.4%, while ß =100 overestimated to 214%. Recovery coefficients were 97.8% and 75.5% for ß =100 and 300, respectively measured on the 16mm sphere. Signal-to-noise ratios were similar for ß=100 and 300, with maximum values of 2683 (ß =300) and 2584 (ß=100).
Our results suggest that ß=100 and 300 were optimal for image quality and quantitation of [18F]F-DCFPyL, with ß=300 having a good compromise between image quality and quantitation.