Distributed versus compartment models for PET receptor studies

Although distributed models are generally accepted as being more realistic than compartment models, use of simpler compartment models is pervasive in nuclear medicine applications, particularly in positron emission tomography (PET). Here, we report on comparisons made between distributed and compartment model outputs to address the question of whether differences between them are sufficient to justify distributed models for analysis of PET receptor experiments. For both two- and three-injection experiments, "data" sets were obtained by simulation using a distributed model and a wide range of parameter values. Optimal fits of the compartment model output to these "data" were achieved with three strategies in which values of different groups of parameter were estimated. Compartment model outputs yielded good fits to all the distributed model outputs and the values of the corresponding parameters were in close agreement. Given the temporal resolution typically available with PET, the use of a distributed model has no advantage over a compartment model for PET receptor quantification.