Optimization of the k2’ parameter for the pharmacokinetic modelling of dynamic PIB scans using SRTM2

Introduction:
In Alzheimer’s disease (AD), quantitative measures of amyloid-β (Aβ) deposition in the brain can be assessed by the analysis of dynamic 11C-labelled Pittsburgh Compound B (PIB) scans. To this end, pharmacokinetic modelling of the data is required, with the most frequent approach being the simplified reference tissue model 2 (SRTM2). However, this method was originally developed for the analysis of neuroreceptor binding, thus it expects a well-defined receptor-rich region. This assumption might be violated in the case of PIB, especially in healthy subjects, which are not expected to have Aβ deposition. Therefore, the aim of this study was to explore the consequences of violating this assumption and, if possible, to define an optimal brain region for estimating k₂’ (i.e. transfer of the tracer from the reference tissue back to plasma), which is used and fixed in the second iteration of SRTM2.
Materials and Methods:
Thirty subjects underwent a dynamic PIB PET scan and were then classified by visual
inspections as PIB positive (+) or PIB negative (-), with a total of fifteen per group. A set of regions were used to estimate the median value of the k₂’ parameter for SRTM2: seven anatomical regions derived from the Hammer’s atlas (grey matter, frontal lobe, parietal lobe, temporal lobe, white matter, brainstem, and the combination of white matter and brainstem), one SPM voxel-based comparison of statistical differences between the groups, and three using different binding potential (BP_ND) thresholds (0.01, 0.05, 0.1). A sensitivity analysis was also made by fixing a range of k₂’ values (from 0.03 to 0.09) and assessing the effects of these changes on estimated BP_ND.
Results:
The different regions used to estimate k₂’ resulted in distinct values for this parameter. In general, values from grey matter and statistical regions showed a larger difference between the groups, an average difference of 32%. The method that presented the smallest difference was the one where the brainstem was used for the estimation (2% difference between groups), but the estimated median values were, in general, higher than for the other methods. Distribution of median k₂’ values showed that the most common estimation is 0.05 for this set of subjects. The method that presented the lowest difference between groups and was the closest to this estimated value was the threshold approach using BP_ND equal to 0.1. Sensitivity analysis of the effects of different k2’ estimation on BP_ND showed that the larger the k₂’ value, the smaller the change in BP_ND.
Discussion:
The large difference between k₂’ estimations between groups suggests that grey matter VOI based methods and the one based on statistical differences between groups might not be acceptable. Threshold based approaches guarantee that only regions with some binding will be considered when estimating k₂’. Preliminary analysis of the data showed that since the relationship between k₂’ and BP_ND is not linear, an overestimation of k₂’ value might result in a smaller bias in BP_ND estimation than an underestimation. Therefore the best method for estimating the k₂’ is by using a threshold on the BP_ND.