Consistency of standardized uptake value with variation in PET reconstruction parameters

Objectives As PET/CT is increasingly being used for biomarker-like assessments, both adequate image quality and robust quantitative measurement are essential. Variations in reconstruction parameters affect both aspects and are again becoming relevant as technology evolves. We assessed and quantified how different reconstruction settings impact standardized uptake value (SUV).

Methods Using a current release time of flight system (Gemini 64 TOF PET/CT Philips) we reconstructed listmode data under all available reconstruction settings. Iterative reconstructions labeled as high quality (3 iterations, 33 subsets, 23 cm kernel width), normal (3 iterations, 33 subsets, 14.1 cm kernel width), and fast (3 iterations, 20 subsets, 14.1 cm kernel width) were evaluated. "Sharpness" was varied as the relaxation setting from a normal setting of 1, smooth of 0.7, smooth A of 0.6 to smooth B of 0.5, the smallest with the least refinement per iteration. The 12 resulting image sets per patient were analyzed, comparing changes in SUV_max for regions of interest (ROIs) drawn in target lesions as well as organ regions. A total of 11 patient studies were evaluated.

Results Using normal speed, normal sharpness as a baseline, SUV_max was found to be highest (on average 9.7% higher) for the "high quality" images reconstructed with 3 iterations, 33 subsets, 23 cm kernel width and relaxation parameter of 1, and lowest (on average 8.2% below baseline) for "fast" settings with 3 iterations, 20 subsets, 14.1 cm kernel width and relaxation parameter of 0.5. As the SUV_max increased per ROI, variance in SUV diminished, with "fast" settings becoming more consistent with baseline (on average 2.4% below), while "high quality" settings continued to increase (on average 11.3% above).

Conclusions The consistency of SUV is affected by altered reconstruction approaches such as subset settings and kernel width. While they lead to perceived improvement in image quality, the impact on SUV needs to be considered.

Research Support OSUMC Imaging Signature Progra