Spatially constrained kinetic modeling with dual reference tissues improves ¹⁸F-flortaucipir PET in studies of Alzheimer disease.

Abstract

Recent studies have shown that standard compartmental models using plasma input or the cerebellum reference tissue input are generally not reliable for quantifying tau burden in dynamic ¹⁸F-flortaucipir PET studies of Alzheimer disease. So far, the optimal reference region for estimating ¹⁸F-flortaucipir delivery and specific tau binding has yet to be determined. The objective of the study is to improve ¹⁸F-flortaucipir brain tau PET quantification using a spatially constrained kinetic model with dual reference tissues. Participants were classified as either cognitively normal (CN) or cognitively impaired (CI) based on clinical assessment. T1-weighted structural MRI and 105-min dynamic ¹⁸F-flortaucipir PET scans were acquired for each participant. Using both a simplified reference tissue model (SRTM2) and Logan plot with either cerebellum gray matter or centrum semiovale (CS) white matter as the reference tissue, we estimated distribution volume ratios (DVRs) and the relative transport rate constant R₁ for region of interest-based (ROI) and voxelwise-based analyses. Conventional linear regression (LR) and LR with spatially constrained (LRSC) parametric imaging algorithms were then evaluated. Noise-induced bias in the parametric images was compared to estimates from ROI time activity curve-based kinetic modeling. We finally evaluated standardized uptake value ratios at early phase (SUVR_EP, 0.7-2.9 min) and late phase (SUVR_LP, 80-105 min) to approximate R₁ and DVR, respectively. The percent coefficients of variation of R₁ and DVR estimates from SRTM2 with spatially constrained modeling were comparable to those from the Logan plot and SUVRs. The SRTM2 using CS reference tissue with LRSC reduced noise-induced underestimation in the LR generated DVR images to negligible levels (< 1%). Inconsistent overestimation of DVR in the SUVR_LP only occurred using the cerebellum reference tissue-based measurements. The CS reference tissue-based DVR and SUVR_LP, and cerebellum-based SUVR_EP and R₁ provided higher Cohen's effect size d to detect increased tau deposition and reduced relative tracer transport rate in CI individuals. Using a spatially constrained kinetic model with dual reference tissues significantly improved quantification of relative perfusion and tau binding. Cerebellum and CS are the suggested reference tissues to estimate R₁ and DVR, respectively, for dynamic ¹⁸F-flortaucipir PET studies. Cerebellum-based SUVR_EP and CS-based SUVR_LP may be used to simplify ¹⁸F-flortaucipir PET study.