Calibration of arterial spin labeling data-potential pitfalls in post-processing.

Joana Pinto,Melvin Mezue,Pedro Vilela,Thomas W Okell,Irene Tracey,Andrew R Segerdahl,Patrícia Figueiredo,Michael A Chappell

doi:10.1002/mrm.28000

Abstract

PurposeTo assess the impact of the different post‐processing options in the calibration of arterial spin labeling (ASL) data on perfusion quantification and its reproducibility.Theory and MethodsAbsolute quantification of perfusion measurements is one of the promises of ASL techniques. However, it is highly dependent on a calibration procedure that involves a complex processing pipeline for which no standardized procedure has been fully established. In this work, we systematically compare the main ASL calibration methods as well as various post‐processing calibration options, using 2 data sets acquired with the most common sequences, pulsed ASL and pseudo‐continuous ASL.ResultsSignificant and sometimes large discrepancies in ASL perfusion quantification were obtained when using different post‐processing calibration options. Nevertheless, when using a set of theoretically based and carefully chosen options, only small differences were observed for both reference tissue and voxelwise methods. The voxelwise and white matter reference tissue methods were less sensitive to post‐processing options than the cerebrospinal fluid reference tissue method. However, white matter reference tissue calibration also produced poorer reproducibility results. Moreover, it may also not be an appropriate reference in case of white matter pathology.ConclusionPoor post‐processing calibration options can lead to large errors in perfusion quantification, and a complete description of the calibration procedure should therefore be reported in ASL studies. Overall, our results further support the voxelwise calibration method proposed by the ASL white paper, particularly given the advantage of being relatively simple to implement and intrinsically correcting for the coil sensitivity profile.

Highlights

Arterial spin labeling (ASL) is a non‐invasive MRI technique that provides quantitative images of tissue perfusion, by using magnetically labeled blood water protons as an endogenous blood flow tracer.[1,2,3] ASL is acknowledged to have great potential as a completely non‐invasive quantitative perfusion imaging technique, but its implementation has been challenging because of the intrinsically low SNR
For pulsed ASL (PASL) control averaging (CtrAvg), cerebral blood flow (CBF) values were significantly different across methods except between reference tissue (RT)‐CSF and RT‐white matter (WM) methods
For the PASL saturation recovery (SatRec) approach, only RT‐CSF and RT‐WM methods were significantly different from the Voxelwise method (Voxel) method, for the pseudo‐continuous ASL (pCASL) Long TR calibration scan (LongTR) approach, significantly lower CBF values were obtained when using the CSF as a reference tissue, compared with other strategies for deriving M0a from M0t

Summary

Introduction

Arterial spin labeling (ASL) is a non‐invasive MRI technique that provides quantitative images of tissue perfusion, by using magnetically labeled blood water protons as an endogenous blood flow tracer.[1,2,3] ASL is acknowledged to have great potential as a completely non‐invasive quantitative perfusion imaging technique, but its implementation has been challenging because of the intrinsically low SNR. This has motivated the development of a multitude of signal acquisition and processing strategies that aim to overcome this limitation as well as the publication of a ASL implementation consensus paper.[4] pseudo‐continuous ASL (pCASL) is the recommended labeling strategy, pulsed ASL (PASL) is still a commonly used technique.[5,6,7,8,9]. It has been demonstrated through theoretical analysis that the 2 factors that CBF quantification using ASL is most sensitive to are M0a estimation and labeling efficiency.[16]

Methods

Results

Discussion

Conclusion