Synopsis

Keywords: Kidney, Quantitative Imaging, T1 mapping, cortex, medulla, transplantation

This work demonstrates rapid free-breathing 3D T1 mapping of the kidney using Magnetization-Prepared Golden-angle RAdial Sparse Parallel imaging (GraspT1). The accuracy and performance of GraspT1was prospectively compared with B1-corrected variable flip angle (VFA) T1 mapping in phantom and in 13 patients with renal transplant. In phantom, both GraspT1 and VFA enabled accurate T1 estimation compared to the gold standard. In patients, free-breathing 3D GraspT1 provided better spatial resolution for improved cortico-medullary differentiation compared to the VFA method, while achieving whole-kidney coverage and similar T1 quantification, at the expense of longer acquisition time.

Introduction

T₁ values in the renal cortex and medulla have been correlated to kidney dysfunction related to fibrosis or inflammation, in native^1-3 and transplanted kidneys^{4, 5}. Inversion recovery (IR) T₁ mapping methods have been found to have better test-retest repeatability in vitro⁶, however, their acquisition times are lengthy, which leads to reduced anatomical coverage of the kidney⁷. IR preparation in combination with a stack-of-stars readout has been shown to provide accurate, whole-organ, free-breathing T₁ mapping in the liver⁸. Here, we aim to extend this approach, called Golden-angle RAdial Sparse Parallel T₁ mapping (GraspT1) to the kidney and compare image quality and T₁ mapping accuracy with variable flip angles (VFA) methods in both phantom and patients with renal transplant.

Methods

Phantom Study: A cylindrical phantom containing 7 vials with MnCl₂ and distilled water solutions (Fig. 1) with typical T₁ values encountered in abdominal organs (300-1500 ms) was imaged on the same 3T magnet (Skyra, Siemens) where the human studies were conducted. The phantom was imaged with an inversion recovery spin-echo (IR-SE) acquisition that provided the reference T₁ values, as well as two (2FA) and three flip angles (3FA) VFA, and 3D GraspT1 acquisitions with similar protocols as used in human studies (Table 1). In both phantom and human studies, a B₁ mapping sequence (TR/TE/FA: 6510ms/1.91ms/8º, FOV 360x360 mm²,10 6-mm slices, 128x128 matrix) was also acquired before the VFA acquisitions, for B₁ correction. An MR physicist delineated ROIs for each of the 7 vials, in the central slice of the phantom, on all acquired T₁ maps, while avoiding artifacts. The T₁ values measured by the GraspT1 and VFA sequences were compared to reference IR-SE T₁ values by Lin’s Concordance Correlation Coefficient (CCC).
Patient Study: In this IRB-approved, bi-center, prospective ongoing study, thirteen patients [M/F: 11/2, age 49.2±9 (30-65) y] with renal transplant had multiparametric MRI on a 3T magnet (Skyra, Siemens). The patients were scanned in fasting conditions with VFA and GraspT1 mapping acquisitions, in coronal orientation, centered on the hilum of the renal allograft (Table 1). Number of slices and spatial resolution was matched between the 3 acquisitions as much as possible (Table 1). Patients were instructed to suspend respiration at end-expiration during the VFA and B1 mapping acquisitions, while they could breathe freely during the 3D GraspT1 acquisition.
B₁ corrected VFA T₁ maps were calculated inline on the scanner, while the GraspT1 T₁ maps were reconstructed offline on a Yarra server, using an in-house MATLAB R2021 algorithm. The T₁ maps resulting from each acquisition were evaluated in Horos by an abdominal radiologist, who rated image sharpness, cortico-medullary differentiation, motion artifacts, streak artifacts, and image noise on a 5-point Likert scale (Figure 2, Table 2). The radiologist also selected 6 circular ROIs in the renal allograft cortex and medulla (2 each in the upper, mid- and lower pole) at the allograft hilum slice⁴, which were then copied onto the T₁ map; average T₁ for cortical and medullary ROIs was reported.
The T₁ map image quality scores and T₁ values were compared between sequences using paired Wilcoxon tests. Cortical and medullary T₁ was then compared between sequences using Bland-Altman analysis.

Results

Phantom Study: 3D GraspT1 and VFA acquisitions achieved comparable T₁ measurements to the IR-SE reference measurements (Fig.1, CCC>0.99), although the B1-corrected VFA T₁ maps had subjectively lower image quality in vials within the T₁ value range (1000-1500 ms; V1-3) of the kidney.
Patient Study: 3D GraspT1 had higher cortico-medullary differentiation than both VFA sequences (Table 2), but higher image noise compared to 2FA VFA. 3FA VFA had higher image noise than 2FA VFA. There were slight discrepancies between the 3D GRASP T1 and the VFA measurements (Table 3; bias 4%-10%, BALA of 20-30%), with significantly higher cortical and medullary 2FA VFA T₁ (Table 3).

Discussion

VFA and 3D GraspT1 achieve similar T₁ mapping accuracy in phantoms. In patients, 3D GraspT1 showed improved cortico-medullary differentiation compared to VFA methods, although T₁ values were similar (<10% Bland-Altman bias). Previous T₁ mapping studies in renal allografts were performed at 1.5T⁴, or with a modified Look-Locker method⁵ known to systematically under-estimate T₁⁶, which prevents a direct comparison. Cortical T₁ values measured in our study with VFA and GraspT1 methods were equivalent to those measured in patients with chronic kidney disease at 3T (range: 1600-2200 ms)³ with an IR method in a recent study. However, they were higher than cortical and medullary T₁ in patients with chronic and acute kidney disease, measured at 3T with a respiratory-triggered IR-SE EPI method in two other studies^{1, 2}. The differences in T₁ could be due to biological differences between transplanted and dysfunctional native kidneys.

Conclusions

Free-breathing 3D GraspT1 provides improved cortico-medullary differentiation, whole-kidney coverage, and equivalent kidney T₁ measurements in vivo compared to the VFA method, although image acquisition time is longer. Future work will address the longer acquisition time of GraspT1 compared to VFA by exploring the stability of T₁ values calculated from fewer radial trajectories per inversion time and/or fewer inversion times. We will also correlate T₁ measurements with renal allograft histopathology, given its potential to serve as an imaging marker of fibrosis and inflammation.

Acknowledgements

This work was supported by the NIH grants R01DK129888, R01EB030549 and R21EB032917.