Reliable estimation of kidney filtration rate with DCE-MRI using motion-robust high spatiotemporal resolution Radial VIBE
Sila Kurugol1, Onur Afacan2, Deborah R Stein3, Michael A Ferguson3, Richard S Lee4, Reid Nichols2, Ravi T Seethamraju5, Jeanne S Chow2, and Simon K Warfield6

1Radiology, Boston Children's Hospital and Harvard Medical School, Bosotn, MA, United States, 2Radiology, Boston Children's Hospital and Harvard Medical School, Boston, MA, United States, 3Pediatrics, Boston Children's Hospital and Harvard Medical School, Boston, MA, United States, 4Surgery, Boston Children's Hospital and Harvard Medical School, Boston, MA, United States, 5Siemens Healthcare, 6Boston Children's Hospital and Harvard Medical School, Boston, MA, United States


Chronic kidney disease poses a significant health burden, and patients benefit from early detection of kidney function. Serum-creatinine based estimation is insensitive to early changes and nuclear medicine studies expose patients to radiation. In this work, we evaluated a novel technique, motion-robust high spatiotemporal resolution Dynamic Radial VIBE (DRV) with compressed sensing, to reconstruct high-quality dynamic image series, and to precisely estimate filtration rate per kidney and per voxel. Our results suggest that, compared to conventional Cartesian VIBE, DRV reconstructs higher quality motion-robust images and results in improved the goodness-of-fit to the tracer kinetic model, reducing RMSE and increasing the precision of filtration rate parameter.


To evaluate if dynamic contrast enhanced (DCE)-MRI using motion-robust high spatiotemporal resolution Radial VIBE can improve estimation of kidney glomerular filtration rate (GFR) compared to conventional Cartesian VIBE in pediatric patients.


GFR is a clinically important quantitative measure of renal function. Serum creatinine based estimation is insensitive to early changes and nuclear medicine studies expose patients to radiation. DCE-MRI uses the filtration of contrast agent through the kidneys to directly measure GFR. Conventional DCE-MRI using Dynamic Cartesian VIBE (DCV) is often corrupted by respiratory and cardiac motion which reduces image quality and affects GFR estimation accuracy, especially for per voxel or per kidney segment. Another limitation is the spatial and temporal resolution required for accurate acquisition of the arterial input function (AIF) peak needed to correctly estimate GFR. Decreasing temporal resolution causes systematic underestimation of GFR. Recently, radial “stack-of–stars” sequence acquisitions have been used to reduce the effect of motion in abdominal MRI1. Radial VIBE continuously acquires radial lines including the center of k-space throughout the scan. Therefore, each sampled line contains equally important information, especially the contrast information. Balanced sampling of k-space makes the acquisition motion-robust. Compressed-sensing reconstruction of dynamic image series further improves image quality by reducing streaking artifacts2, which arise from undersampling. In this work, we assess the ability of DCE-MRI using Dynamic Radial VIBE (DRV) to obtain motion-robust images of kidneys with high spatial and temporal resolution and compare its performance for estimating filtration rate parameter to the standard DCV.


We imaged six pediatric patients at 3T for six minutes after injection of Gadavist using a radial “stack-of-stars” 3D FLASH sequence (TR/TE/FA 3.56/1.39ms/12o, 32 coronal slices, voxel size=1.25x1.25x3mm). We achieved a mean temporal resolution of 3.3 sec for the arterial phase (2 minutes) and 13 sec for the remaining phases (4 minutes). 4D dynamic image series were reconstructed offline using compressed-sensing3 to improve temporal resolution and image quality, effectively reducing the streaking artifacts. We also searched the hospital database and gathered data from six patients recently imaged with the standard DCV. Using in-house software, the volumes acquired in two different phases were aligned and resampled to the same resolution. The kidney parenchyma was segmented semi-automatically. An ROI was drawn inside the aorta to determine arterial input function. The tissue enhancement curves of kidney parenchyma were converted to concentration curves and a separable two-compartment tracer kinetic model by Sourbon et al.3 was fitted to estimate the filtration rate parameter (FT). GFR is calculated by multiplying FT with renal parenchyma volume.


An experienced radiologist visually evaluated the DRV images (Fig.1) in six patients with various conditions affecting their kidneys. DRV consistently improved image quality by eliminating ghosting artifacts and limiting streaking artifacts when using compressed sensing reconstruction while achieving high temporal resolution (3.3sec) for all patients compared to clinical images acquired using the DCV sequence (11sec). DRV minimized the effect of motion in tissue contrast-enhancement curves (Fig.1b). Multiple high-quality post-contrast images were also generated using increased number of radial lines for additional renal morphology evaluation. We compared the goodness-of-fit of contrast-enhancement curves to the tracer kinetic model for each kidney acquired using DRV and DCV sequences using root-mean-square error (RMSE) and the R-square measures of fit (Fig.2a,b). DRV reduced the mean RMSE from 0.45±0.57 to 0.08±0.05 and increased the R-square from 0.94±0.11 to 0.99±0.02. We also used wild-bootstrap analysis4 and computed the coefficient of variation percent (CV%=standard deviation/mean x100) of filtration rate parameter (FT) as a measure of precision. DRV improved the precision of parameter estimation by reducing the average CV% from 29.84±19.72 to 10.97±6.41. In addition to per kidney FT estimation, we also estimated filtration rate parameter per voxel. A sample set of images of one patient showing different stages of contrast enhancement and the corresponding voxel-wise filtration rate parameter map is shown in Fig.3. Fig.4 reports GFR values per kidney.


We demonstrated that DCE-MRI using DRV improved image quality, limited motion artifacts, obtained high spatial and temporal resolution and accurately acquired the AIF when compared to DCV. DRV improved the two-compartment tracer kinetic model fitting quality and improved the precision of the estimated filtration rate parameter. DRV also achieved successful computation of per voxel filtration rate parameter. Thus, DRV is potentially a useful method to estimate GFR, especially in pediatric patients for whom motion often poses as an obstacle for quality imaging. The possible use of this novel method is enormous and allows children already undergoing an MRI, such as oncology patients with impaired renal function or undergoing nephrotoxic chemotherapy, to have GFR tested simultaneously.


This work is supported by the Young Investigator Award from the Society of Pediatric Radiology.


1.Chandarana, Hersh, et al. "Free-breathing radial 3D fat-suppressed T1-weighted gradient echo sequence: a viable alternative for contrast-enhanced liver imaging in patients unable to suspend respiration." Investigative radiology 46.10 (2011): 648-653.

2. Feng L, Grimm R, Tobias Block K, et al. Golden-angle radial sparse parallel MRI: Combination of compressed sensing, parallel imaging, and golden-angle radial sampling for fast and flexible dynamic volumetric MRI. Magn Reson Med. 2014;72: 707–717.

3. Sourbron, Steven P., et al. "MRI-measurement of perfusion and glomerular filtration in the human kidney with a separable compartment model." Investigative radiology 43.1 (2008): 40-48.

4. Freiman, M, Voss SD, et al. "Quantitative body DW-MRI biomarkers uncertainty estimation using unscented wild-bootstrap." In International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 74-81. Springer Berlin Heidelberg, 2011.


Fig.1. DCE-MRI images acquired using a) Dynamic Cartesian VIBE (DCV) (time resolution=11sec) and b) Radial VIBE (time resolution=3.3sec). Ghosting artifacts are present in the DCV image due to motion. Radial VIBE improved image quality. Corresponding average tissue enhancement curves of kidney parenchyma are plotted for c) Cartesian VIBE and d) Radial VIBE. The effect of motion is minimized using the radial VIBE sequence, which resulted in improved fitting of the model to the measured signal.

Fig.2. Goodness-of-fit measures to the tracer kinetic model are compared for DCV and DRV. a) shows the root-mean-squared error (RMSE) between the measured signal and the fitted model. b) shows the R-square measure of fit computed over 10 kidneys from six patients using DCV and a separate set of 11 kidneys from six patients using the DRV. Three patients in the study had single kidneys. DRV achieved improved goodness-of-fit reducing RMSE and increasing R-square. c) shows the CV% of parameters over multiple bootstrap resampling repetitions. DRV had a smaller coefficient of variation, indicating improved precision of parameter estimation.

Fig.3. DCE-MRI images acquired using the motion-compensated dynamic radial VIBE sequence. The contrast enhancement of the kidney is shown at different time points indicating passage of contrast from cortex to medulla and to the renal pelvis. The rightmost image shows the estimated map of the filtration rate parameter. Improved image quality obtained by DRV allowed for voxel wise parameter fitting and identification of regions with dysplasia indicated by delayed enhancement and reduced filtration rates (dark blue regions in the superior segment of the kidney).

Fig.4. Per kidney and total GFRs of six pediatric patients computed by multiplying filtration rate parameter from DRV imaging by volume of the kidney parenchyma are reported. Patient 5 and 6 had single kidneys. We also normalized GFR values for each patient's body surface area and reported GFR values per 1.73m2 of body surface area.

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)