Modeling Strategies for Quantification of In Vivo 18F-AV-1451 Binding in Patients with Tau Pathology

Aggregation of hyperphosphorylated tau is a major hallmark of many neurodegenerative diseases, including Alzheimer disease (AD). In vivo imaging with PET may offer important insights into pathophysiologic mechanisms, diagnosis, and disease progression. We describe different strategies for quantification of 18F-AV-1451 (T807) tau binding, including models with blood sampling and noninvasive alternatives. Methods: Fifteen subjects (4 controls, 6 AD, 3 progressive supranuclear palsy, 2 cortico basal syndrome) underwent 180-min PET with 18F-AV-1451 and arterial blood sampling. Modeling with arterial input functions included 1-, 2-, and 3-tissue-compartment models and the Logan plot. Using the cerebellum as reference region, we applied the simplified reference tissue model 2 and Logan reference plot. Finally, simplified outcome measures were calculated as ratio, with reference to cerebellar concentrations (SUV ratio SUVR) and SUVs. Results: Tissue compartment models were not able to describe the kinetics of 18F-AV-1451, with poor fits in 33%–53% of cortical regions and 80% in subcortical areas. In contrast, the Logan plot showed excellent fits and parameter variance (total volume of distribution SE R2 = 0.91) and SUVR 100–120 min (R2 = 0.94), with 80–100 min already representing a reasonable compromise between duration and accuracy (R2 = 0.93). Time–activity curves and kinetic parameters were equal for cortical regions and the cerebellum in control subjects but different in the putamen. Cerebellar total volumes of distribution were higher in controls than patients. For these methods, increased cortical binding was observed for AD patients and to some extent for cortico basal syndrome, but not progressive supranuclear palsy. Conclusion: The Logan plot provided the best estimate of tau binding using arterial input functions. Assuming that the cerebellum is a valid reference region, simplified methods seem to provide robust alternatives for quantification, such as the Logan reference plot with 100-min scan time. Furthermore, SUVRs between target and cerebellar activities obtained from an 80- to 100-min static scan offer promising potential for clinical routine application.