Free-Breathing Liver Fat Quantification Using an Undersampled Multi-Echo 3D Stack-of-Radial Technique

Tess Armstrong^1,2, Isabel Dregely^1,3, Alto Stemmer⁴, Yutaka Natsuaki⁵, and Holden Wu^1,2

¹Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, ²Physics and Biology in Medicine, University of California, Los Angeles, Los Angeles, CA, United States, ³Biomedical Engineering, King's College London, London, United Kingdom, ⁴Siemens Healthcare, Erlangen, Germany, ⁵Siemens Healthcare, Los Angeles, CA, United States

Synopsis

Non-alcoholic fatty liver disease is the leading cause of chronic liver disease. Multi-echo Cartesian MRI methods can non-invasively quantify liver fat, but are susceptible to motion artifacts and limited by breath hold (BH) imaging. We have developed a new free-breathing (FB) liver fat quantification method using non-Cartesian 3D stack-of-radial imaging. To reduce scan time, we undersampled data up to a factor of R=3. In healthy subjects, mean fat quantification was statistically equivalent among different R. Initial comparisons with spectroscopy show good agreement. Our new technique can potentially achieve accurate whole-liver fat quantification within a fast 1-minute FB scan.

Introduction

Non-alcoholic fatty liver disease (NAFLD) affects 20-30% of the population worldwide and is the leading cause of chronic liver disease.¹ The current gold standard for diagnosis and monitoring of NAFLD is an invasive biopsy to characterize fatty infiltration in the liver, however, biopsy suffers from spatial sampling bias and associated morbidity.^1-4 Multi-echo MRI methods can provide reliable non-invasive 3D quantification of fat in the entire liver.^5-7 However, current multi-echo MRI methods are based on Cartesian sampling, which is susceptible to respiratory motion-induced artifacts. As a result, scans are performed during a single breath hold (BH) and face challenges to achieve full volumetric coverage, high spatial resolution, desirable echo times, and artifact-free images. Furthermore, breath holding may not be possible for certain patients. In this work, we develop and evaluate a new free-breathing (FB) liver fat quantification technique using a multi-echo non-Cartesian 3D stack-of-radial sequence with golden-angle ordering. Various degrees of undersampling are investigated to further reduce scan time.

Methods

Sequence A bipolar multi-echo RF-spoiled gradient echo sequence using the golden-angle-ordered 3D stack-of-radial trajectory was developed (Fig. 1). Bipolar and radial gradient errors can lead to incorrect fat quantification^8-11, thus gradient delay phase errors were measured¹² with additional calibration spokes (40 for each Gx and Gy). Since the golden angle ordering supports flexible selection of any contiguous subset of radial readouts for reconstruction, we emulated accelerated scans by retrospectively reconstructing the first 33% of readouts (undersampling factor R=3), first 50% of readouts (R=2), and 100% of readouts (R=1, fully sampled).

Experiments Liver scans were acquired on a 3T scanner (Skyra, Siemens) using the FB 3D stack-of-radial sequence in n=7 healthy volunteers. In select volunteers, a BH 3D Cartesian sequence¹³ with 4-fold acceleration and CAIPIRINHA¹⁴ reconstruction was also acquired, along with gold standard single-voxel MR spectroscopy (SVS) in the liver and bone marrow. Representative parameters are listed in Table 1. A 32-channel array was used for all acquisitions. T1 bias in fat fraction (FF) calculation was minimized using a low flip angle.¹⁵

Reconstruction Using the calibration data, the Gx and Gy delays were calibrated for each channel and TE, and then used to update the k-space trajectory for 3D gridding reconstruction (Fig. 1d). Parallel imaging reconstruction was not employed. Fat-water separation was performed using a graph cut algorithm with a 6-peak fat model and a single effective R2* per voxel.^16-18 Fat fraction (FF) was calculated as FF(%) = 100% x Fat / (Fat + Water) with magnitude discrimination to eliminate noise bias.¹⁵ Fat, water, and FF for the BH Cartesian acquisitions were calculated by scanner software using a similar signal model.¹³ SVS FF results were calculated by scanner software.¹⁹

Analysis 3D fat and water images were viewed and regions of interest (ROIs) were drawn in the liver and the bone marrow using OsiriX to assess fat-water separation and FF quantification. A one-way ANOVA F-test was performed to evaluate statistical differences in FF between FB 3D stack-of-radial with R=1, 2, and 3.

Results

Representative fat, water and FF images are shown in Fig. 2 for the BH Cartesian and FB 3D stack-of-radial scans with various R. Due to differences in BH and FB positions, slices are not perfectly matched. Linear correlation plots between 3D stack-of-radial at R=1, 2, 3 are shown in Fig. 3. A strong linear relationship exists between the under-sampled data (R=2 and 3) and fully sampled data (R=1). The mean FF values in the ROIs among all 3D stack-of-radial scans at different R are not significantly different (p-value = 0.9977) (Fig. 3c). Bar plots comparing FB 3D stack-of-radial with BH 3D Cartesian and BH SVS in two representative volunteers are shown in Fig. 4. The mean FF obtained by BH SVS and FB 3D stack-of-radial scans is highly consistent.

Discussion and Conclusion

Although the R=2 and R=3 3D stack-of-radial images have increased noise and streaking, the fat-water separation quality is maintained. As a result, mean FF quantification was not significantly different for all volunteers among R for the FB 3D stack-of-radial scans. Additional acceleration may be possible by utilizing non-Cartesian parallel imaging. We performed an initial comparison between FF from FB 3D stack-of-radial, standard BH Cartesian and BH SVS. The FF between all techniques had good correspondence, but had some differences possibly due to ROI placement, subject inter-scan motion, or signal model parameters. Further work is needed to carefully compare different FF measurement techniques. In summary, our technique can potentially achieve accurate 3D whole-liver fat quantification within a 1-2 minute free-breathing scan. This may improve patient compliance and fat quantification for management of NAFLD.

Acknowledgements

Siemens Medical Solutions

References

[1] Bellentani, Stefano, and Mariano Marino. "Epidemiology and natural history of non-alcoholic fatty liver disease (NAFLD)." Ann Hepatol 8.Suppl 1 (2009): S4-S8.

[2] Rinella ME. Nonalcoholic Fatty Liver Disease: A Systematic Review. JAMA. 2015; 313(22): 2263-2273.

[3] Vernon, G., A. Baranova, and Z. M. Younossi. "Systematic review: the epidemiology and natural history of non-alcoholic fatty liver disease and non-alcoholic steatohepatitis in adults." Alimentary pharmacology & therapeutics 34.3 (2011): 274-285.

[4] de Alwis, Nimantha Mark Wilfred, and Christopher Paul Day. "Non-alcoholic fatty liver disease: the mist gradually clears." Journal of hepatology 48 (2008): S104-S112.

[5] Bydder, Mark, et al. "Relaxation effects in the quantification of fat using gradient echo imaging." Magnetic resonance imaging 26.3 (2008): 347-359.

[6] Ma, Jingfei. "Dixon techniques for water and fat imaging." Journal of Magnetic Resonance Imaging 28.3 (2008): 543-558.

[7] Yu H, McKenzie CA, Shimakawa A, et al. Multiecho reconstruction for simultaneous water-fat decomposition and T2* estimation. J Magn Reson Imaging 2007; 26: 1153–1161.

[8] Moussavi, A., Untenberger, M., Uecker, M. and Frahm, J. (2014), Correction of gradient-induced phase errors in radial MRI. Magn Reson Med, 71: 308–312.

[9] Yu, H., Shimakawa, A., McKenzie, C. A., Lu, W., Reeder, S. B., Hinks, R. S. and Brittain, J. H. (2010), Phase and amplitude correction for multi-echo water–fat separation with bipolar acquisitions. J. Magn. Reson. Imaging, 31: 1264–1271.

[10] Lu, W., Yu, H., Shimakawa, A., Alley, M., Reeder, S. B. and Hargreaves, B. A. (2008), Water–fat separation with bipolar multiecho sequences. Magn Reson Med, 60: 198–209.

[11] Peters, Dana C., J. Andrew Derbyshire, and Elliot R. McVeigh. "Centering the projection reconstruction trajectory: reducing gradient delay errors." Magnetic resonance in medicine 50.1 (2003): 1-6.

[12] Block, K. T., and M. Uecker. "Simple method for adaptive gradient-delay compensation in radial MRI." Proceedings of the International Society for Magnetic Resonance in Medicine. 2011.

[13] Zhong, X., Nickel, M. D., Kannengiesser, S. A.R., Dale, B. M., Kiefer, B. and Bashir, M. R. (2014), Liver fat quantification using a multi-step adaptive fitting approach with multi-echo GRE imaging. Magn Reson Med, 72: 1353–1365.

[14] Breuer, F. A., Blaimer, M., Heidemann, R. M., Mueller, M. F., Griswold, M. A. and Jakob, P. M. (2005), Controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA) for multi-slice imaging. Magn Reson Med, 53: 684–691.

[15] Liu, C.-Y., McKenzie, C. A., Yu, H., Brittain, J. H. and Reeder, S. B. (2007), Fat quantification with IDEAL gradient echo imaging: Correction of bias from T1 and noise. Magn Reson Med, 58: 354–364.

[16] Yu, H., Shimakawa, A., McKenzie, C. A., Brodsky, E., Brittain, J. H. and Reeder, S. B. (2008), Multiecho water-fat separation and simultaneous R estimation with multifrequency fat spectrum modeling. Magn Reson Med, 60: 1122–1134.

[17] Hernando, Diego, et al. "Robust water/fat separation in the presence of large field inhomogeneities using a graph cut algorithm." Magnetic Resonance in Medicine 63.1 (2010): 79-90.

[18] ISMRM Fat-Water Toolbox. 2012.

[19] Pineda, Nashiely, et al. "Measurement of Hepatic Lipid: High-Speed T2-Corrected Multiecho Acquisition at 1H MR Spectroscopy—A Rapid and Accurate Technique 1." Radiology 252.2 (2009): 568-576.

Figures

Fig 1: a) 3D stack-of-radial k-space trajectory. Radial readouts are acquired along k_z before rotating. b) Radial readouts are rotated continually by the golden angle (θ_G). c) Six echoes are acquired every TR using a bipolar multi-echo readout gradient. d) Reconstruction pipeline for the 3D stack-of-radial data.

Table 1: The typical imaging parameters for the FB 3D stack-of-radial and BH Cartesian scans.

Fig 2: The water, fat, and FF images for the BH Cartesian the FB stack-of-radial scans for the same volunteer for axial (a) and coronal (b) orientations.

Fig 3: Linear correlation plots for the liver ROI (a) and bone marrow ROI (b) FF measurements for the FB stack-of-radial scans at the undersampling factors R=1 R=2 and R=3. c) Box plot of FB radial FF at the undersampling factors R=1, R=2 and R=3 (n=7, ANOVA F-test).

Fig 4: Bar plots of the FF measurements from two representative volunteers for the BH SVS, BH 3D Cartesian and FB 3D stack-of-radial (R=1, R=2, and R=3) scans in the liver (c) and the bone marrow (d).

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

3837