Effect of a Bayesian Penalized Likelihood PET Reconstruction Compared With Ordered Subset Expectation Maximization on Clinical Image Quality Over a Wide Range of Patient Weights

Abstract

Objective: A study was performed to compare background liver signal-to-noise ratio (SNR) and visually assessed image quality of clinical PET/CT studies from the same PET acquisition data reconstructed by Bayesian penalized likelihood (BPL) and ordered subset expectation maximization (OSEM) over a range of patient weights.

Materials and methods: The effect of a BPL PET reconstruction algorithm on liver SNR and visually assessed image quality over a range of patient weights (41-196 kg; n = 108) was retrospectively compared with standard-of-care OSEM reconstruction on the same PET acquisition data after IV administration of ¹⁸F-FDG (4 MBq/kg).

Results: BPL showed no significant change (p > 0.05) in liver SNR with increasing weight and body mass index (BMI), whereas OSEM showed increasing noise with increasing weight and BMI. The liver SNR was significantly higher using BPL than a standard OSEM reconstruction (p < 0.0002 for all BMI groups). Visually assessed image quality declined at a greater rate with increasing weight and BMI in the OSEM images than with BPL images.

Conclusion: BPL provides a more consistent visually assessed image quality and liver background SNR than does OSEM, with the greatest benefit for the heaviest patients.

Keywords: Bayesian penalized likelihood; FDG; PET/CT; ordered subset expectation maximization; signal-to-noise ratio.