Incorporation of MRI-AIF information for improved kinetic modelling of dynamic PET data

Abstract

Pharmacokinetic analysis of dynamic PET data requires estimation of the arterial input function (AIF). An alternative method to collecting blood samples during scanning is the simultaneous estimation approach in which AIF and kinetic parameters are estimated at once [1]. We have investigated how information from a simultaneous MRI-AIF can be incorporated into this model and evaluated its effect on estimated kinetic parameters by comparing absolute bias and coefficient of variation (CV) values on influx constants (Ki). As the delivery of the tracer or contrast agent is dominated by vascular flow dynamics, PET-AIF and MRI-AIF boluses will have similar peak shapes if the same injection speed is used [2,3]. Hence it can be assumed that the values of λ1 of eq. 1 [1] can be fixed to reduce the number of parameters in fitting. Simulated data was generated using two tissue compartment model with rate constants from a clinical brain study [5]. Noise was added using eq. 2 where α set to 0.5,2 and 4. AIF scaling was done using one blood sample obtained at the end of the study, or by fixing A3 and λ3 to their true values assuming they can be determined using multiple blood samples.