Introduction
Parametric CBF, OEF and CMRO2 images can be derived from oxygen-15 labelled H2O, O2 and CO scans. Traditionally, this is performed using a steady state approach. With some cyclotrons, such as the RDS 112, however, these tracers can only be administered as a bolus or short inhalation. Therefore, a basis function method was developed for generating parametric images based on bolus/short inhalation administration and dynamic PET scans [1]. In the present study, several variations of this basis function method were implemented and the effects on parametric image quality assessed.
Methods
The basis function method [1] is based on a modified version of the kinetic model as described by Mintun et al. [2]. In this study various implementations of this basis function method were developed. These methods differed in the number of parameters derived from the water scan that were subsequently fixed in fitting the oxygen data: dispersion, volume of distribution of water (Vd) and CBF. An additional implementation used fixed values of Vd derived from a co-registered and segmented MRI scan. Note that CBF images were generated from the oxygen scan alone, except when CBF was reused from the water scan. First, simulations were performed to assess the effects of fixing none, one or more parameters on the accuracy and precision of CBF (from oxygen scan alone), OEF and CMRO2 as function of noise (0,5,…,25%). Next, using each implementation, parametric CBF, OEF and CMRO2 images were calculated for 10 clinical studies.
Results
Both simulations and clinical images showed that bias (>25%) and poor reproducibility (>100%) of CBF and OEF were substantial when no parameter was fixed. Bias and precision improved to acceptable levels (<10 and 25%, respectively) when Vd was fixed from the water scan. Reusing Vd from the water scan provided slightly poorer precision for all parameters than when using fixed Vd values for grey and white matter (based on MRI scan), but may be more accurate in clinical practice. Bias and precision of CMRO2 was almost independent of specific model implementation. The quality of clinical parametric images, as shown in figure 1, was consistent with the findings of the simulation study, indicating that primarily the quality of CBF and OEF images depend on the number of parameters reused.
Conclusions
Accuracy and precision of CBF and OEF and, the quality of parametric CBF and OEF images depend on the specific basis function method implementation, i. e. ‘better’ results are obtained by fixing or reusing more parameters from the water data within the analysis of the oxygen scan. However, CMRO2 data with sufficient image quality can be obtained without reusing or fixing any parameters.