Room: Exhibit Hall | Forum 8
Purpose: To quantify the diagnostic performance of regularized reconstruction and reduced effective counts in brain PET/CT for pediatric epilepsy patients.
Methods: Twenty-five pediatric patients (1â€“16y) with clinical brain PET/CT were included in this retrospective study. Research ethics approval was received. Administration of 3.7 MBq/kg 18-F-FDG per the clinical standard-of-care and institutional imaging protocol was delivered (uptake time = 44Â±7 minutes). 3D imaging was acquired on a SiPM PET/CT (GE Discovery MI) in a single list-mode, 300 second bed-position with time-of-flight. Images were reconstructed with a fixed matrix size (384x384) at 2.78 mm thickness and variable field-of-view based on habitus (22â€“25 cm). CT was performed for attenuation correction and anatomic localization. Volumes were reconstructed with ordered-subset-expectation-maximization or Q.Clear, a proprietary block-sequential-regularized-expectation-maximization algorithm with varied penalization factor (Î²), each at variable acquisition time (45s, 90s, 180s, 300s) to simulate reduced injected dose. Two pediatric nuclear medicine physicians reviewed PET volumes blinded and random to patient\reconstruction\time and scored quality (noise, spatial resolution, artifacts) and clinical parameters (image quality of the cortex, basal ganglia and thalamus) on a five-point scale (1 = non-diagnostic, 3 = clinically adequate, 5 = excellent). Significance was determined through Wilcoxian signed rank.
Results: Reconstruction with Q.Clear improved quality and clinical scores across all count levels, although no statistically significant change was identified when effective acquisition times were highest (300s). The greatest improvement in quality due to Q.Clear was shown in low-count conditions (45s). No statistically significant improvement in overall image quality was identified between 90, 180 and 300 seconds with Î²â‰¥350, with all scores considered adequate or better (â‰¥3).
Conclusion: Clinical evaluation of pediatric PET brain image quality with simulated reduced time\dose reconstructions was shown to be clinically adequate at reductions of up to 70% when applying Q.Clear with an appropriately implemented penalization factor.