Kinetic modeling of brain FDG data with input function derived from images by independent component analysis

Blood (or plasma) input function is mandatory in quantitative Positron Emission Tomography (PET) imaging. The input function (IF) is usually determined by blood sampling from heated patient arm, by means of continuous blood sampling, or from population based IFs normalized with one or more samples. However, external blood sampling is invasive and is always subject to uncertainties due to the manipulations. IF could also be extracted by means of algorithms such as factor analysis and independent components especially in heart studies. In this work, we show that IF can be determined from the images of brain arteries in a study of regional cerebral metabolic rates of glucose (rCMRG) with 18F-fluorodeoxyglucose (FDG) in normal volunteers using Independent Component Analysis (ICA). A region of interest (ROI) was drawn around the artery in the image and decomposed into blood and tissue using ICA. The calculated IF is therefore corrected for spillover of radioactivity from tissue, then it is normalized with a single plasma sample to correct for partial volume and finally it is denoised with spectral analysis. The calculated IF with ICA is found comparable to IF manually determined, and rCMRG values obtained with ICA-IF and sampled IF differ by nearly 3%.