Accurate determination of metabolic rates from dynamic positron emission tomography data with very-low temporal resolution.

Abstract

The graphical approach is widely used for the pixelwise determination of local metabolic rate of glucose from dynamic positron emission tomography (PET) data. In its conventional implementation, measured integrals over time frames are used to approximate instantaneous tracer concentrations at midframe times ("midframe approach"). This is justified in case of high temporal resolution of the PET measurement; that is, if scan protocols with a large number of short frames are used. This requires fast data handling and large amounts of memory. Cardiac gating and three-dimensional (3D) acquisition of dynamic studies is hardly possible with this approach. Therefore, a new variant of the graphical method is proposed which can be used with a very low number of rather long frames. An operational equation of the graphical method was derived which uses measured time integrals only and, thus, avoids the systematic errors of the midframe approximation. This "integral approach" was evaluated in computer simulations based on experimental data. The integral approach enables the use of protocols with 3 frames only without compromising accuracy of the derived metabolic rates whereas the midframe approach leads to bias of about 10% to 20% for these protocols. Furthermore, test-retest stability can significantly be improved when using the integral approach. The integral approach to the graphical evaluation of dynamic PET data yields accurate and precise results using scan protocols with down to only 3 frames. This can be relevant to gating and/or 3D acquisition of dynamic studies. The integral approach is applied most naturally whenever the input function is derived from the dynamic PET data.