Radiology

Pattern recognition (psychology)

Mathematics

Statistics

Magnetic resonance imaging

Calibration

Perfusion

Artificial intelligence

Computer science

White matter

Arterial spin labeling

Medicine

Reproducibility

Magnetic Resonance in Medicine

Purpose To 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 Methods Absolute 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. Results Significant 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. Conclusion Poor 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.

Calibration of arterial spin labeling data—potential pitfalls in post‐processing