Kinetic modeling using a two-tissue compartment model and an additional irreversible vascular component improves the quantification of [11C]PBR28 brain PET data

Objectives We propose a novel kinetic model for the neuroinflammation ligand [11C]-PBR28. This model hypothesizes the existence of an additional irreversible component from the blood to the endothelium, similar to that previously demonstrated for [11C]-(R)-PK11195

Methods The model (2TCM-1K) was tested on a dataset of ten healthy subjects. A simulation was also performed to quantify the error generated by the standard 2TCM when the presence of the irreversible component is not taken into account

Results Compared to standard 2TCM, 2TCM-1K improved the curves fit to the tissue data points in all the regions and increased the randomness of the weighted residuals. 2TCM-1K poorly identified only one brain region in one subject. In contrast, 2TCM showed a poor fit in 4% of the brain regions. Akaike score was lower for 2TCM-1K in 94% of the regions. Importantly, the VT values obtained with 2TCM and 2TCM-1K were also poorly correlated (R2 < 0.4). VT 2TCM-1K estimates were found to be better correlated region-wise with mRNA TSPO gene expression than 2TCM (r = 0.54 and r = -0.42 respectively). Simulated data showed that 2TCM is not very sensitive to variations in tissular VT, as differences greater than 50% translated into estimated VT variations of 15% or less

Conclusions A kinetic model that accounts for the endothelial irreversible binding of TSPO tracers improves the quantification of [11C]PBR28 brain PET data