Evaluation of error on parameter estimates in the analysis of receptor studies with positron emission tomography

Positron emission tomography dynamic study together with compartment model analysis allows the quantification of radioligand binding to receptor sites in the human brain. In the quantitative analysis, parameter estimate error is affected by noise that depends on the collected count, and the reliability of estimated kinetic parameters is important for assessment of the binding potential. However, evaluation of the reliability for human data is not easy because the true noise level is not precisely known. In this study, we evaluated a method for estimating the reliability of determination of parameters in a region-of-interest analysis of human data by the bootstrap approach. In a simulation study, the mean and variation values estimated by the bootstrap approach agreed with those by each non-bootstrap realization with the simulated time-activity curves. In human data, reasonable values of mean and variation for the estimated binding potential were obtained by these methods with the bootstrap approach. This method is useful for PET quantitative analysis, as it is able to be applied easily to the evaluation of the influence of injection dose and sensitivity of the PET system.