E to the suppression of outliers by discarding PSF modelling reconstruction algorithms. It really should be emphasized that the RC measurements were obtained with in-house computer software, which was developed with excellent care to make sure reproducibility and precision. In certain, the VOIs weren’t drawn straight around the PET WAY-200070 manufacturer pictures but were automatically segmented on CT photos and optimized with know-how from the phantom specifications.Enhancing quantification and variability in little brain structures and image uniformityWe checked the PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19954917 quantification in the striatum of the Hoffman brain phantom and observed an increase in activity concentration measured in the putamen and caudate, when compared to non-optimized pictures. We also significantly reduced after the optimization the dispersion of values for putamen, caudate, and grey-matter-to-white-matter ratios. Finally, the good uniformity of the photos across the field of view was confirmed by the ratio of right-to-left activity concentration and anterior-to-posterior activity concentration measured around the Hoffman brain photos. Additionally, as expected, the choice of the reconstruction mode parameters had no impact on the ratio of activity concentration in the right-to-left and frontal-to-occipital regions of interest. Those ratios need to not be affected by the spatial resolution. However, they are likely affected by low frequency variation resulting from scatter, uniformity, and attenuation corrections. Attempts to take into account these effects before pooling data sets from multiple centres had been suggested by Joshi et al. [1], but the results were not considered as convincing by the authors. In the present work, the compensation for scatter, random coincidences, and attenuation was performed using the manufacturer’s latest techniques, with slight differences in the implementation for the different scanner models. The Hoffman phantom also enabled us to verify that the quantification in typical small structures of the brain, such as the caudate, was improved after the reconstruction optimization.Image spatial resolution and signal-to-noise ratioOn average, image spatial resolution was slightly improved, despite a marked decrease for the centres equipped with reconstruction algorithms including TOF and PSF modelling. This improvement is likely due for the increase in the number of iterations in the other centres. As expected, this improvement is counterbalanced by a decrease of the signal-to-noise ratio. We checked with Hoffman phantom acquisitions that this reduction did not impact the visual aspect of the photos.Variability across centresThe variability across centres was reduced but not minimized by this work. For that, it would have been necessary to apply post-filtering tailored per site according to ISR. WeHabert et al. EJNMMI Physics (2016) 3:Page 15 ofdeliberately chose not to do this in order to preserve the spatial resolution. This latter goal was achieved for the 18 centres without PSF modelling reconstruction. In order to account for the residual variability across centres, we are currently designing a statistical model that takes into account the centre effect (as assessed by the phantom studies). Several other parameters have been shown to influence the variability of the measurements across centres, including injected dose per kilogram, delay between the injection and the scan acquisition, patient positioning, weight, and glycaemia, and these have been extensively discussed by Boellard [17,.