Standard Uptake Values (SUVs) from Position Emission Tomography (PET) imaging are used clinically to track tumour response to chemotherapy. The values are dependent on the reconstruction method, and must be measured consistently and accurately to be used clinically. GE Healthcare’s Q.Clear software uses a noise reduction and edge enhancement algorithm within their iterative reconstruction to improve image quality. The Q.SUV values calculated using this software are not the same as those using Ordered Subset Expectation Maximisation (OSEM), with GE Healthcare stating that their values are more accurate (Ross, 2014).
The two reconstructions will be compared on SUV’s and noise properties, using phantom data. The analysis of data has been inspired by Bailey and Kalemis (2005) and Schmidtlein et al. (2010), in that the data will be binned into multiple frames, and the reconstructions compared on noise variables. List-mode acquisition will be used, to allow the data to be re-binned off-line into a specified number of statistically independent projections, which can be combined to estimate differing noise and activity levels. This partitioning of the data and reduced count rate will amplify noise.
The aim of this project is to investigate the effect of the noise in the image, when propagated through the reconstruction algorithm, by comparing image quality and quantification. There will be an investigation into whether there is a relationship between the noise in the image and the uptake value measured.