RETURN TO ABSTRACT LISTING


Objective:
Q.Clear is a Bayesian blocked sequential penalized-likelihood PET reconstruction algorithm (BSREM) that has shown potential to improve image quality and quantification accuracy for PET imaging. Prior pediatric studies have shown that digital PET scanners with increased z-axis field of view (FOV) may allow for faster scans or reduced administered activity based on simulations of decreased total counts, but these studies have not evaluated the impact of BSREM reconstruction. This study aimed to evaluate the impact of BSREM on simulated whole-body FDG PET images with decreased total counts obtained using a digital PET/CT system. We hypothesize that use of this algorithm will allow for lower acquisition times or administered activities compared to Ordered Subsets Expectation Maximization (OSEM) reconstructions while maintaining image quality.

Materials and Methods:
This retrospective study included 20 (16 men, 4 women) clinically indicated whole-body FDG PET examinations acquired on a 5-ring, 25 cm axial FOV PET/CT system. Standard-of-care clinical images were acquired at 90 seconds per bed position with 21% overlap. PET examinations were reprocessed by retrospective rebinning of list-mode data to simulate reduced scan acquisition times of 60, 50, 40, and 30 seconds per bed (sec/bed). Images were then reconstructed using BSREM as follows: 90 sec/bed, ß = 700; 60 sec/bed, ß = 800; 50 sec/bed, ß = 900; 40 sec/bed, ß = 1000, 30 sec/bed, ß = 1300). Simulated image volumes were independently rated by three board-certified pediatric radiologists for (1) image quality, (2) conspicuity of normal structures, and (3) image noise using a 3-rank Likert scale. Reviews occurred in random order, and reviewers were blinded to reconstruction parameters. Standardized uptake values (SUV), including SUVmean and SUVmax were recorded by a board-certified pediatric radiologist using spherical 3D 3 cm regions-of interest placed in the liver parenchyma and right-mid thigh, and based on segmentation of the most FDG-avid lesion. Analysis of variance (ANOVA), Friedman's test and Dunn's test were used to compare scores and SUVs across acquisition times.

Results:
Mean patient age was 9.0 ± 5.5 years. Qualitatively, as compared to the 90 sec/bed standard protocol, no significant difference was observed for conspicuity of normal structures, image noise, or overall image quality for any reduced count reconstruction (p > 0.99). Quantitatively, no statistical difference was observed for SUVmean in liver or thigh across all reduced count reconstructions (p > 0.99). SUVmean for the most FDG-avid lesion was statistically significantly different starting at 50 sec/bed (p = 0.017). SUVmax for the thigh was not significantly different for any reconstruction (p > 0.99) but SUVmax was statistically significantly different at 30 sec/bed (p = 0.001) for the most FDG-avid lesion and at 60 sec/bed for the liver (p = 0.03).

Conclusion:
For pediatric whole body 18F-FDG PET performed on a 25-cm FOV digital PET/CT, a BSREM reconstruction algorithm allows for even shorter acquisition times without significantly impacting image quality.