Robust kidney perfusion mapping in pediatric chronic kidney disease using single‐shot 3D‐GRASE ASL with optimized retrospective motion correction

Abstract

Purpose To develop a robust renal arterial spin labeling (ASL) acquisition and processing strategy for mapping renal blood flow (RBF) in a pediatric cohort with severe kidney disease. Methods A single‐shot background‐suppressed 3D gradient and spin‐echo (GRASE) flow‐sensitive alternating inversion recovery (FAIR) ASL acquisition method was used to perform 2 studies. First, an evaluation of the feasibility of single‐shot 3D‐GRASE and retrospective noise reduction methods was performed in healthy volunteers. Second, a pediatric cohort with severe chronic kidney disease underwent single‐shot 3D‐GRASE FAIR ASL and RBF was quantified following several retrospective motion correction pipelines, including image registration and threshold‐free weighted averaging. The effect of motion correction on the fit errors of saturation recovery (SR) images (required for RBF quantification) and on the perfusion‐weighted image (PWI) temporal signal‐to‐noise ratio (tSNR) was evaluated, as well as the intra‐ and inter‐session repeatability of renal longitudinal relaxation time (T1) and RBF. Results The mean cortical and/or functional renal parenchyma RBF in healthy volunteers and CKD patients was 295 ± 97 and 95 ± 47 mL/100 g/min, respectively. Motion‐correction reduced image artefacts in both T1 and RBF maps, significantly reduced SR fit errors, significantly increased the PWI tSNR and improved the improved the repeatability of T1 and RBF in the pediatric patient cohort. Conclusion Single‐shot 3D‐GRASE ASL combined with retrospective motion correction enabled repeatable non‐invasive RBF mapping in the first pediatric cohort with severe kidney disease undergoing ASL scans.

Consensus-based technical recommendations for clinical translation of renal ASL MRI

“…Issues such as repeatability and validation of ASL with alternative methods to measure renal perfusion were extensively addressed therewith and, therefore, will not be covered in the present manuscript. The previous literature review [8] as well as 12 studies published in the intervening period [11][12][13][14][15][16][17][18][19][20][21][22], the results of a series of surveys and the discussions that took place at the in-person meetings form the basis for the final recommendations presented here.…”

“…Therefore, breath-hold scans are not recommended for clinical renal ASL [R7.3] and scans under free breathing are preferred [R7.4]. Respiratory triggering methods have been widely used in renal ASL [16,18,43,44,58,73] (among others) and can reduce motion artefacts in free breathing scans but at the expense of scan time [R7.5]. Respiratory coaching, where subjects are instructed to breathe during the periods when no data are being acquired (i.e.…”

“…Other preprocessing methods in renal ASL aim to remove or downweight outlier measurements or correct corrupted individual perfusion weightedimages, most frequently caused by subtraction errors due to kidney motion (see Figure S2 in the supplementary material). Outlier rejection methods, including retrospective sorting of renal ASL data, have relied on manual [44,56,62] or automatic [18,58] approaches, including using data from external sensors such as respiratory bellows [55,63]. Outlier rejection methods have been shown beneficial to improve the data quality and, therefore, are recommended for renal ASL [R8.2].…”

“…Another class of preprocessing methods are motion correction techniques typically based on image registration, which can be used in conjunction with outlier rejection. These methods are widely used in renal ASL [11,18,63,75,76] (among others) to realign the kidneys and ensure their position is consistent throughout the ASL time series and, therefore, reduce subtraction errors. They have been shown to reduce image artefacts, reduce partial volume effects between cortex and medulla, and improve the SNR of the perfusion-weighted signal and the repeatability of perfusion measurements.…”

“…In addition, these methods should be used to not only align ASL control and label images but also any calibration scans [18,21], such as M 0 data (essential for quantification) and optional T 1 data (acquisition of which is currently not recommended as part of the default renal ASL protocol). Extra care should be taken when aligning the ASL data to the calibration data (particularly if BGS is used) due to differences in image contrast which can reduce the accuracy of registration algorithms.…”