Prospective Image Alignment for Time-Resolved Renal BOLD MRI
Inge Manuela Kalis1, David Pilutti1, Axel Joachim Krafft1,2,3, and Michael Bock1

1Dept. of Radiology - Medical Physics, University Medical Center Freiburg, Freiburg, Germany, 2German Cancer Consortium (DKTK), Heidelberg, Germany, 3German Cancer Research Center (DKFZ), Heidelberg, Germany

Synopsis

Renal function can be analyzed by time-resolved BOLD MRI before, during and after a functional challenge. Inconsistent kidney positions from one measurement to another hamper the analysis of renal parenchyma and medulla over time. Here, a new method, Kidney ALIgnment for BOLD Renal Imaging (KALIBRI), with prospective rigid image registration of each kidney is proposed.

Introduction

Kidney oxygenation can be measured with time-resolved renal BOLD MRI where changes in renal oxygenation are induced by a functional challenge (1–5) (e.g., by drinking water) which is detected as an R2* change in the renal tissue. Renal R2* measurements are usually performed within individual, subsequent breath holds before, during and after the challenge. Because of inconsistencies in the kidney locations from one measurement to another, motion compensation is required. Ideally, displacements would be corrected prospectively to prescribe the correct slice positions for each measurement. Here, a new method, Kidney ALIgnment for BOLD Renal Imaging (KALIBRI), is presented for prospective 3D motion correction of the kidneys.

Material and Methods

For prospective motion correction, KALIBRI acquires a 3D VIBE data set before the acquisition of a 2D multi-echo gradient echo (mGRE) sequence for BOLD imaging. Both data sets (3D VIBE, 2D BOLD) are measured in a single breath hold. Data are acquired repeatedly, and the 3D VIBE data from the very first breath hold serves as a reference data set. VIBE image data are transferred automatically to an external PC where an ITK-based (6) rigid image registration (RIGR) of the current to the reference VIBE data set is done. To save computation time, the registration is performed only within pre-defined regions for each kidney. Position updates for both kidneys are then sent to the MR system to realign the imaging slices of the subsequent 2D mGRE acquisition (Fig. 1). The sequence employs a vendor-specific feedback mechanism and is executed once the slice update is received. The whole procedure is designed to be completed within a single breath hold of about 20 seconds. To remove residual in-plane displacements and deformations, the dynamic 2D BOLD images are retrospectively registered to the reference BOLD images using translational and non-rigid image registration using MIRT (7). The KALIBRI method was implemented on a 3T whole body system (Siemens PRISMA, Erlangen, Germany) using the following imaging parameters: 3D VIBE: TR = 3.34 ms, TE = 1.19 ms, a = 10°, matrix: 256x174, voxel size: 1.6x1.6x4.0 mm3, partitions: 22, GRAPPA with 24 reference lines, TA = 3.4 s; 2D BOLD: 2 coronal slices, TR = 35 ms, TE1 = 2.42 ms, ΔTE = 2.66 ms, 12 echoes, matrix: 192x192, voxel size: 2.2x2.2x5.0 mm3, a = 25°, GRAPPA with 24 reference lines, bandwidth = 810 Hz/Px, TA = 7.6 s. Time-resolved renal BOLD MRI was performed in 6 healthy volunteers who had no food or water at least 5 h before the exam. After 4-7 baseline R2* measurements, the volunteers drank 1000 ml tap water in the magnet, and another 25-40 BOLD data sets were collected over about 50 min. The KALIBRI method was evaluated by calculation of the Mutual Information (MI) (8,9), the spatial overlap between two segmentations as Dice Coefficient (DC) (10) as well as the Standard Deviation (SD) of each kidney before and after registration.

Results

The total acquisition time including motion correction was 17 s. Prospectively corrected images showed better alignment of internal renal structures than data without correction: improvements are illustrated in Fig. 2 as difference images between the reference data and the images acquired without registration, with prospective and retrospective registration, respectively. On average, MI values improved after the prospective registration by up to 25%; and an additional 10% after retrospective correction. Prospective registration alone did not change DC and SD, but the combined registrations led to an improvement of up to 4% for DC and 10% for SD. Results of the R2* analysis of time-resolved BOLD MRI using small ROIs in medullar and cortical regions are summarized in Fig. 3. The average R2* baseline value was about 30.6 ± 4.3 s-1 in medullar regions and 17.7 ± 1.5 s-1 in the cortex. In 4 out of 6 volunteers, the medullar R2* values showed a decrease during water challenge of up to 40%, and a recovery afterwards. For all volunteers, the cortical values did not change.

Summary

The proposed method KALIBRI for motion correction showed to improve the quality of time-resolved renal BOLD MRI, since local regions in the medulla and the cortex are better aligned and thus can be compared consistently over many breath holds. We demonstrated its functionality with in vivo experiments during a water challenge performing also a semi-automatic full time-resolved analysis. The KALIBRI method could also be applied to other serial measurements in the kidney, e.g. for MR-guided radiotherapy, for longitudinal studies, and, after adaptation, for other abdominal organs.

Acknowledgements

This work was funded (in part) by the Helmholtz-Alliance ICEMED – Imaging and Curing Environmental Metabolic Diseases, through the Initiative and Network Fund of the Helmholtz Association, and by the DFG-Project HA 7006/1-1.

References

1. Prasad PV. Evaluation of Intra-Renal Oxygenation by BOLD MRI. Nephron Clin. Pract. 2006;103:c58–c65.

2. Hofmann L, Simon-Zoula S, Nowak A, Giger A, Vock P, Boesch C, Frey FJ, Vogt B. BOLD-MRI for the assessment of renal oxygenation in humans: Acute effect of nephrotoxic xenobiotics. Kidney Int. 2006;70:144–150.

3. Li L-P, Halter S, Prasad PV. Blood Oxygen Level-Dependent MR Imaging of the Kidneys. Magn. Reson. Imaging Clin. N. Am. 2008;16:613–625.

4. Liss P, Cox EF, Eckerbom P, Francis ST. Imaging of intrarenal haemodynamics and oxygen metabolism. Clin. Exp. Pharmacol. Physiol. 2013;40:158–167.

5. Epstein FH, Veves A, Prasad PV. Effect of Diabetes on Renal Medullary Oxygenation During Water Diuresis. Diabetes Care 2002;25:575–578.

6. Ibanez L, Schroeder W, Ng L, Cates J. The ITK Software Guide. Kitware Inc 2005.

7. Myronenko A. Medical Image Registration Toolbox. [Online]. Available: https://sites.google.com/site/myronenko/research/mirt.[Accessed: Mar-2015].

8. Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P. Multimodality image registration by maximization of mutual information. IEEE Trans. Med. Imaging 1997;16:187–198.

9. Maes F, Vandermeulen D, Suetens P. Medical image registration using mutual information. Proc. IEEE 2003;91:1699–1722.

10. Zou KH, Warfield SK, Bharatha A, Tempany CMC, Kaus MR, Haker SJ, Wells WM, Jolesz FA, Kikinis R. Statistical Validation of Image Segmentation Quality Based on a Spatial Overlap Index. Acad. Radiol. 2004;11:178–189

Figures

Fig. 1: Timing diagram showing the proposed prospective motion correction method. Initially, reference data (3D VIBEref and BOLDref) are acquired, which are used in all subsequent registration calculations. In each subsequent measurement, 3D VIBEcurr data are acquired which are co-registered to the reference data set. Each kidney is registered independently within cROIs. The resulting translations Trig are used to realign the slice positions of the next 2D BOLDcurr images.

Fig. 2: BOLD images before registration, after prospective and after retrospective registration, and the corresponding difference images relative to the reference image (normalized).

Fig. 3: Cortical and medullar R2* values of the left and the right kidney, plotted over the measured time points. The vertical yellow bar defines the drinking period.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0711