INTRODUCTION

Arterial Spin Labeling (ASL) is a promising technique for renal perfusion assessment. Different imaging techniques at independent magnetic filed strengths (1.5T and 3T) have been used to assess renal perfusion¹. However, there is a notable absence of comparative research at both field strengths. Therefore, this study aims to compare the outcomes and reproducibility of renal ASL at both 1.5T and 3T. These findings might contribute to ASL reliability in assessing renal perfusion, supporting its potential integration into clinical practice.

METHODS

Subjects and Study Design:
Written informed consent was obtained from subjects. Urine and blood were obtained to measure the estimated glomerular filtration rate (eGFR) with the CKD-EPI equation². Eight healthy volunteers (eGFR>90ml/min/1.73m², mean age±standard deviation (SD)=32±7.9 years) were scanned twice within the same day on 1.5T Aera and 3T Skyra MRI scanners (Siemens Healthcare), using 18-channel body and spine coils. The same protocol was repeated with a median of seven days (range=1-17 days).

MRI Protocol:
Pseudo-continuous ASL (PCASL) sequence employing six different post-labeling delays (PLDs), with background suppression (BS) and SE-EPI readout. Pre-saturation pulses were applied before the labeling pulses. BS was achieved using a slice-selective FOCI pulse before labeling and two non-selective tangent/hyperbolic tangent pulses after labeling. Pulse timings were optimized for each PLD to suppress static tissue signal to 15%. This optimization was performed for each field strength considering the expected kidney T1 values. PCASL were acquired with and without BS pulses for the PLD=1.3s, to estimate the efficiency of the BS pulses.
T1 Inversion Recovery sequence employing 16 inversion times and SE-EPI readout. Other sequence parameters are described in Table 1.

Image processing:
All the images underwent registration in Elastix³ using a groupwise registration method. Subsequent analysis was conducted using custom scripts in Matlab (Mathworks). T1 maps were generated by fitting the signal to the inversion-recovery equation [1]:
$$S_k=M_0(1-2e^{-TI_k/T1)}+M_0e^{-TR/T1})$$
To compute Renal Blood Flow (RBF) and Arterial Transit Time (ATT) maps, a voxel-by-voxel approach was employed, fitting the data to the one-compartment Buxton⁴ model, with the assumption that by the time of image acquisition, labeled spins remained in the blood [2]:
$$(SI_C-SI_L)_t=\begin{cases}0&\text{if }0<\tau_t+\text{PLD}_t<\text{ATT}\\\frac{2\alpha\text{T1}_b}{6000\lambda}M_0RBFe^{-\frac{\text{ATT}_t}{\text{T1}_b}}(1-e^{-\frac{(\tau_t+\text{PLD}_t-\text{ATT})}{\text{T1}_b}})&\text{if }\text{ATT}<\tau_t+\text{PLD}_t<\tau_t+\text{ATT}\\\frac{2\alpha\text{T1}_b}{6000\lambda}M_0RBFe^{-\frac{\text{PLD}_t}{\text{T1}_b}}(1-e^{-\frac{\tau_t}{\text{T1}_b}})&\text{if } \tau_t+\text{ATT}<\tau_t+\text{PLD}_t \end{cases}$$
where $$$(SI_C - SI_L)$$$ is the signal difference between control and label images for each PLD, $$$\alpha=0.75\times0.9\times0.9$$$ is the labeling efficiency, considering PCASL efficiency and the effects of 2 BS pulses, $$$\lambda=0.9$$$mL/g is the tissue-blood partition coefficient, $$${T1_{b}}=1.48$$$s and $$$1.65$$$s are the longitudinal relaxation times of blood at 1.5T and 3T, respectively and $$$\tau$$$ is the labelling duration. Regions of Interest (ROIs) in the cortex and medulla were manually drawn in the T1 maps and used to obtain mean values by averaging both kidneys.

Statistical analysis:
Differences between field strengths, were tested by paired t-test. Reproducibility was assessed by Bland-Altman plots and by the inter-session within subject coefficient of variation (wsCV), calculated as the root mean square of the squared ratio of the SD to the mean for each subject of the two repeated measurements⁵.

RESULTS AND DISCUSSION

RBF, ATT, and T1 maps for a representative subject are depicted in Figure 1.Table 2 provides averaged cortical and medullary values across the group, for each visit and field strength. T1 measurements aligned with literature values⁶, and as expected, T1 values were longer at 3T compared to 1.5T. Inter-session wsCVs for T1 in the cortex and medulla were low, indicating excellent reproducibility (<2% for 1.5T and <1.2% for 3T). RBF and ATT measurements obtained during both visits and at both field strengths yielded remarkably similar results, consistent with the literature^1,6-7, with no statistically significant differences between field strengths, except for the RBF values in the medulla (p<0.001), which were higher at 1.5T. It is worth noting that cortical RBF values were also 10% higher at 1.5T (although the differences did not reach statistical significance). These data were also reproducible, with cortical and medullary wsCVs<11% for RBF and ATT at both 1.5T and 3T, except for the medulla at 3T (wsCV=18.3%). Bland-Altman plots (Figure 2) showed good reproducibility between visits for both the cortex and medulla. Perfusion weighted signal (PWS), temporal Signal-to-Noise Ratio (tSNR), and BS efficiency for PLD=1.3s were also obtained (Table 3). Results showed a higher BS efficiency in the 1.5T scans, which could be attributed to a more uniform magnetic field, which partly explain the observed differences in RBF, that could also be due to a higher labeling efficiency, at 1.5T compared to 3T. Although this parameter was not measured.

CONCLUSION

Similarity in results across both field strengths highlights the potential for wider adoption of PCASL-based renal perfusion assessment and support its use in multicenter clinical studies.