Comparison between regional cerebral blood flow estimates using different MR arterial spin labelling approaches and early-phase amyloid PET on a simultaneous PET/MR

Abstract

Aim/Introduction: Amyloid-PET can be used as dualphase biomarker enabling differential diagnosis between dementias by providing images of amyloid load (late-phase) and functional estimates of regional cerebral blood flow (rCBF) (early-phase) [1]. MRI arterial spin labelling (ASL) can quantify CBF and can provide similar patient classification information as [18F]FDG PET [2]. Here, we compare rCBF obtained with ASL with the early-phase distribution of amyloid-PET using [18F]flutemetamol (FLUT). Materials and Methods: Twenty participants, 70(60-80) years old, 14 cognitively normal/6 mild cognitive impairment, were scanned on a SIGNA PET/MR (GE-Healthcare). PET list-mode data were acquired for 30 minutes following the injection of ~185MBq of FLUT. PET images were reconstructed using GE Q.Clear (β=250) for two separate time frames corresponding to 20s-80s and 120s-300s post-arrival of tracer in the brain. Enhanced-ASL with pCASL labelling and 3D spiral FSE acquisition was acquired with 6 post-labelling delays to obtain voxel-wise arrival-time-corrected CBF maps [3] (scan duration: 9min). CBF maps were calculated using either global (CBFglobal) or local (CBFlocal) correction for equilibrium magnetisation and also using a single time point post-labelling of 2.2s (CBF2.2) (scan duration: ~3min). FreeSurfer brain segmentation on each participant’s T1w MPRAGE image (1mm isotropic resolution) was used to estimate median CBF values for 12 ‘meta-regions’. Linear regression was used to compare intensity normalised early phase PET and ASL. Results: For PET20-80 vs CBFglobal, CBFlocal, CBF2.2 the r (median(range)) values across the 20 participants were 0.56(0.26-0.65), 0.67(0.33-0.82) and 0.59(0.27-0.82) respectively. Removing subcortical ROIs caudate and striatum (clear outliers on the scatter-plots) improved the r values to 0.63(0.36-0.81), 0.82(0.48-0.96), 0.85(0.42-0.96). Lower r values for PET120-300 were observed. Conclusion: For cortical regions, there is a good proportional relationship between ASL and PET measurements (improved using earlier PET (PET20-80) to PET120-300 proposed in [1]) with this proportion not holding for the subcortical ROIs with values underestimated for ASL. ASL CBF maps with local magnetisation correction have a closer association with PET suggesting that spatial inhomogeneity in coil sensitivities cause significant errors if uncorrected. Good association is maintained with single time point ASL suggesting that 2.2s is sufficient delay for complete transit of labelled blood; enabling faster ASL acquisition. These close associations suggest that simultaneous ASL with late-phase amyloid- PET and injection of FLUT off the bed could provide excellent classification of dementia patients in a single scan.