Investigation of the quantitative accuracy of low-dose amyloid and tau PET imaging.

With the increasing incidence of dementia worldwide, the frequent use of amyloid and tau positron emission tomography imaging requires low-dose protocols for the differential diagnoses of various neurodegenerative diseases and the monitoring of disease progression. In this study, we investigated the feasibility to reduce the PET dose without a significant loss of quantitative accuracy in 3D dynamic row action maximum likelihood algorithm-reconstructed PET images using [11C]PIB and [18F]THK5351. Eighteen cognitively normal young controls, cognitively normal elderly controls, and patients with probable Alzheimer's disease (n = 6 each), were included. Reduced doses were simulated by randomly sampling half and quarter of the full counts in list mode data for one independent realization at each simulated dose. Bias was evaluated between the reduced dose from the full dose of standardized uptake value ratio (SUVR), distribution volume ratio (DVR) from reference Logan, and non-displaceable binding potential (BPND) from simplified reference tissue model (SRTM). DVR yielded the least bias at low dose compared to SUVR and BPND, and thus, is highly recommended. The dose of [18F]THK5351 and [11C]PIB can be reduced to a quarter of the full dose using DVR for evaluation, whereas the dose can only be reduced to half and a quarter of the full dose for [18F]THK5351 and [11C]PIB using SUVR. BPND showed inconsistent trend and large bias at low dose. The feasibility of dose reduction was dependent on the selected parameters of interest, reconstruction algorithms, reference regions, and to a lesser degree by motion effects.