Protocol Harmonisation and in-vivo Comparison of Arterial Spin Labelling MRI for Multicenter Clinical Trials

Background:

Arterial Spin Labelling (ASL) is a technique for measuring cerebral blood flow (CBF)1,2 and has shown promising results for discriminating Alzheimer’s disease patients from mild cognitive impairment and controls1. ASL is commercially available on all major MRI vendors; however, it has been shown that CBF values are sequence and scanner-dependent, due to variability of the acquisition methods used by each vendor. In this study, we compare the latest generation of pseudo-continuous ASL (pCASL) vendor-provided sequences to quantify between-vendor variability of whole-brain and regional CBF estimation.
Methods:
Eight healthy volunteers (4/4 M/F; mean/SD/median age=47/12/48 years) were scanned on three 3T MR systems with pCASL matched acquisition parameters (Table 1) and ADNI-2 3D-T1weighted (MPRAGE). CBF quantification and post-processing was performed using ExploreASL3 in conjunction with SPM124. Post-processing included registration of the T1-MPRAGE and ASL to MNI space, parcellation of the T1-MPRAGE to grey matter (GM), white matter (WM), and regions affected in dementia (ROI_dem). Quantification was performed according to the ASL consensus paper5. Mean-CBF was calculated in ROI_dem and in the whole-brain within a standard (GM+WM>70%) and expanded (GM+WM>5%) mask. Spatial coefficient of variation (sCoV)6 was calculated for GM(>70%) and whole-brain (GM+WM>70%). Statistical significances were calculated using repeated measure ANOVA (P<0.05).
Results:
Whole-brain CBF values using the expanded mask showed better agreement between three vendors than the standard mask (Figure 1). There was no significant difference in mean-CBF of ROI_dem for Philips and Siemens, while GE showed significantly lower mean-CBF in these regions (Figure 2). sCoV also showed a similar trend, with GE having the lowest sCoV relative to Philips and Siemens for both whole-brain and GM (Figure 3).
Conclusions:
Whole-brain CBF was similar across vendors when an expanded mask was applied. However, there was a difference between GE and Philips/Siemens in CBF and sCoV in grey matter regions. We hypothesize that the differences are related to between-scanner differences in effective spatial resolution, in particular the lower effective resolution of GE’s spiral readout7. Future work will investigate whether smoothness equalization8 can account for this. We anticipate that this work will increase the utility of ASL as a perfusion biomarker in multi-center dementia studies.