Positron-emission tomography (PET) is a non-invasive technology increasingly used in medical treatment and drug discovery. Tracer Kinetic Modelling (TKM) of dynamic PET images uses a mathematical compartment model of tracer behavior to quantify tracer concentrations in various regions. Parameter estimation of this tracer kinetic model essentially relies on the knowledge of initial approximations to the parameters. This could be extremely difficult if prior information is not available. In this paper, an efficient hybrid parameter estimation algorithm is proposed to fit compartmental model to measurements in order to overcome the difficulty of locating such initial approximation. This hybrid method is a combination of the Quantum-Behaved Particle Swarm Optimization (QPSO) and the Levenberg-Marquardt (LM) methods. The accuracy of results is validated by comparing to results obtained by other optimization techniques in the literature.