Positron emission tomography (PET) imaging of tau aggregates plays an increasingly important role in the diagnosis of Alzheimer's disease (AD) and other tauopathies, and has provided new insights into the pathophysiology of these diseases. Promising tau radiotracers have been developed, however they all suffer from important limitations due to off-target binding either in brain regions directly implicated in the disease process or in adjacent regions. We propose to develop a new approach based on factor analysis of dynamic sequences to remove the contribution of this contaminating signal to the PET measurements. We expect our approach to improve our ability to quantify small signal changes in tau aggregates in key brain regions, which in turn should improve our ability to diagnose AD in early stages, to detect changes in longitudinal studies and to monitor responses to future therapeutics. We will characterize the performance of the proposed method in computer simulations and will evaluate the technique in experimental measurements acquired in healthy controls, subjects with mild cognitive impairment (MCI) and AD patients.