Effect of variation in relaxation parameter value on LOR-RAMLA reconstruction of 18F-FDG PET studies

We tested the impact of different values of the relaxation parameter lambda (lambda) on contrast and noise in line-of-response row-action maximum likelihood algorithm (LOR-RAMLA) in 18F-fluorodeoxyglucose positron emission tomography (PET)/computed tomography (CT) imaging.

Phantom studies were performed on a Gemini XL PET/CT scanner. The NEMA/IEC (National Electrical Manufacturers Association/International Electro technical Commission) torso phantom was used and acquisition data were reconstructed with lambda values ranging from 0.025 to 0.1. Quality of the reconstructed images was evaluated by contrast recovery coefficients and background variability values according to the NEMA NU 2-2001 procedures.

Contrast recovery coefficients and background variability increased significantly when lambda increased. The best noise-versus-resolution trade-off was obtained with lambda in the 0.04-0.06 range. For LOR-RAMLA reconstruction, the manufacturer allows a possible lambda choice from 0.025 to 0.1. We would not advise too small (0.025) or too large (0.1) lambda values which result in too smooth or too noisy images.

We determined optimal lambda values in LOR-RAMLA on a Gemini XL PET/CT scanner. Caution is needed when using lambda values out of that range.