Clinical evaluation of reconstruction and acquisition time for pediatric 18F-FDG brain PET using digital PET/CT.

Abstract

18F-2-fluoro-2-deoxyglucose (FDG) positron emission tomography (PET) plays an important role in the diagnosis, evaluation and treatment of childhood epilepsy. The selection of appropriate acquisition and reconstruction parameters, however, can be challenging with the introduction of advanced hardware and software functionalities. To quantify the diagnostic performance of a block-sequential regularized expectation maximization (BSREM) tool and reduced effective counts in brain PET/CT for pediatric epilepsy patients on a digital silicon photomultiplier system. We included 400 sets of brain PET/CT images from 25 pediatric patients (0.5-16years old) in this retrospective study. Patient images were reconstructed with conventional iterative techniques or BSREM with varied penalization factor (β), at varied acquisition time (45s, 90s, 180s, 300s) to simulate reduced count density. Two pediatric nuclear medicine physicians reviewed images in random order - blinded to patient, reconstruction method and imaging time - and scored technical quality (noise, spatial resolution, artifacts), clinical quality (image quality of the cortex, basal ganglia and thalamus) and overall diagnostic satisfaction on a 5-point scale. Reconstruction with BSREM improved quality and clinical scores across all count levels, with the greatest benefits in low-count conditions. Image quality scores were greatest at 300-s acquisition times with β=500 (overall; noise; artifacts; image quality of the cortex, basal ganglia and thalamus) or β=200 (spatial resolution). No statistically significant difference in the highest graded reconstruction was observed between imaging at 180s and 300s with an appropriately implemented penalization factor (β=350-500), indicating that a reduction in dose or acquisition time is feasible without reduction in diagnostic satisfaction. Clinical evaluation of pediatric 18F-FDG brain PET image quality was shown to be diagnostic at reductions of count density by 40% using BSREM with a penalization factor of β=350-500. This can be accomplished while maintaining confidence of achieving a diagnostic-quality image.