Abstract
156
Objectives To evaluate and apply a novel image derived plasma input function (PIF) algorithm for unbiased quantification of [11C]PiB binding.
Methods A modeling-based PIF iterative estimation algorithm we recently developed (Zhou et al., SNM 2011 abstract) was first verified by 4 [11C]PiB (amyloid-β) human dynamic PET scans with arterial blood sampling for PIF measurements, and then applied to 60 (50 controls, and 10 mild cognitive impairment (MCI) patients) [11C]PiB human dynamic PET scans without blood sampling. Fifteen ROIs were manually drawn on the co-registered MRI images and then copied to dynamic PET images for ROI kinetics. For comparison, both estimated and measured PIFs were normalized by area under the curve. Distribution volume (DV) was estimated by applying Logan plot with PIF to the measured ROI kinetics. The DV ratio (DVR)=DV(ROI)/DV(cerebellum) was calculated after DV estimation. DVR was also estimated by a simplified reference tissue model for comparison. The Cohen's effect size was calculated for DV and DVRs.
Results The estimated PIFs were similar to the measured PIFs in terms of curve shape. Linear correlations between the DVs from estimated PIF (estPIF) and measured PIF (mPIF) were obtained as: DV(estPIF) = 1.11DV(mPIF) - 1.94, R2=0.99, and DVR(estPIF)=1.01DVR(mPIF) - 0.01, R2=0.99. The DVs in the MCI group were higher than DVs in controls for all ROIs, but only significant in frontal, lateral temporal, parietal, cingulate, striatum, and thalamus (p<0.02), marginal in cerebellum (p=0.06), and no difference in mesial temporal, occipital, pons, and white matter. The effect size for DV was larger than DVR for all ROIs.
Conclusions Estimation of DV by using estimated PIF is an unbiased noninvasive method for quantification of [11C]PiB PET. The DV obtained by estimated PIF has higher statistical power than DVR estimates to discriminate MCI from controls, which is able to detect accumulation of amyloid-β in cerebellum, and has potential to improve the accuracy in early diagnosis of AD