A major benefit of the three-dimensional (3D) PET imaging technique in neuroscience, as well as in clinical applications, is that it offers the possibility of dynamically quantifying metabolic processes with a sensitivity of up to 10−12 mol L−1 for the tracer concentration. However, all positron emission tomographs provide biased data with complex dependencies, which means that to obtain quantitative activity distributions in 3D, it is necessary to make several corrections. For example, inhomogeneous detector efficiencies, photon attenuation, Compton scattering, and random coincidences need to be corrected. Furthermore, dynamic imaging represents a challenge, because a high temporal resolution requires short acquisition time frames with rather poor statistics of recorded events from the radioactive decay. Apart from the necessary corrections, the applied reconstruction method has an important impact on the achievable image quality in PET. In this respect, iterative reconstruction methods are becoming the state-of-the-art techniques as they offer superior image quality when compared to analytical methods. Although iterative reconstruction is associated with higher computational demand, the higher calculation effort can be moderated by using a range of optimisation strategies and has been further helped by the remarkable boost in computational resources over the last two decades.