Abstract
BACKGROUND: (15)O-water positron emission tomography (PET) is considered the gold standard method for non-invasive measurement of cerebral blood flow (CBF). However, previously published average CBF values in healthy subjects have varied greatly and the cause of these variations remains unclear. This study investigates how image reconstruction methods and spatial resolution affect CBF measurements with (15)O-water PET. METHODS: Eight healthy subjects each underwent dynamic (15)O-water PET scans with continuous arterial blood sampling. Images were reconstructed using two different algorithms; ordered subset expectation maximisation and block sequential regularised expectation maximalisation with varying reconstruction parameters. CBF was estimated for the whole brain, grey matter, and central white matter. Reconstruction-specific effective spatial resolution was estimated using phantom measurements and simulations. RESULTS: The mean whole brain CBF was 0.48 mL/cm(3)/min and showed little dependence on the image reconstruction method. Grey matter CBF varied between 0.52 and 0.57 mL/cm(3)/min, and central white matter CBF between 0.20 and 0.28 mL/cm(3)/min. Regional CBF showed great dependence on effective spatial resolution with a negative correlation between grey matter CBF and resolution (r = -0.96) and a positive correlation between central white matter and resolution (r = 0.93). CONCLUSION: This study concludes that grey matter and central white matter CBF, but not whole brain CBF measured with quantitative (15)O-water PET is reconstruction method dependent, mainly due to varying spatial resolution with consequent partial volume effects. Variations in published CBF values cannot be explained solely by reconstruction methods or spatial resolution.