Synopsis
In this study, we developed and
validated a “single-shot” sequential 2D-chemical shift-encoded MRI (CSE-MRI) technique
for motion-insensitive quantification of liver proton density fat fraction
(PDFF). A phantom of 11 vials with varying PDFF demonstrated equivalent PDFF
between 2D- and 3D-CSE-MRI. Fifteen subjects underwent five different
CSE-MRI acquisitions: 3D-single-breathhold (BH), slice-interleaved 2D-single-BH
and free-breathing (FB), and sequential 2D-single-BH and FB. PDFF was
measured and averaged across all nine Couinaud liver segments. Good agreement
was observed in PDFF between all 2D-CSE-MRI acquisitions and 3D-CSE-MRI.
Qualitative motion artifact evaluation demonstrated significantly
superior scores for free-breathing “single-shot” sequential 2D-CSE-MRI compared
to free-breathing slice-interleaved 2D-CSE-MRI.Purpose
Chemical shift-encoded MRI (CSE-MRI) techniques provide accurate and precise quantification of proton density fat-fraction (PDFF), a quantitative imaging biomarker of triglyceride concentration [1-6]. Current CSE-MRI techniques are based on either 3D [1] or slice-interleaved 2D [2] acquisitions. Unfortunately, these techniques require a prolonged breath-hold (BH) to avoid motion-related artifacts. In previous studies, respiratory-gating and navigator-based methods have been shown to enable accurate liver PDFF quantification without the need for breath-holding [7]. Gated methods, however, require several minutes of scan time, and the resulting images contain moderate residual motion artifacts. Therefore, the purpose of this study was to develop and validate a “single-shot” sequential 2D CSE-MRI technique for motion-insensitive quantification of PDFF in the liver.
Methods
Phantom Experiments: A PDFF phantom consisting of 11 cylindrical vials with varying peanut oil concentrations in an agar based emulsion (PDFF=0-100% in 11 vials) was constructed as in previous work [8]. This phantom was scanned (Table 1) using three CSE-MRI acquisitions: (3D, slice-interleaved 2D, and sequential 2D).
Human Subjects: Patients undergoing clinical abdominal MRI (n=11) and healthy volunteers (n=4) were prospectively enrolled in this IRB-approved and HIPAA-compliant single-institution protocol.
Image Acquisition: All imaging was performed at 1.5T using a phased array torso coil (MR450w-v25.0 with GEM, GE Healthcare, Waukesha, WI). Each subject was scanned using five different acquisitions (Table 2), including: 1) a previously validated 3D single-breathold (BH) technique [1], 2&3) a slice-interleaved 2D technique in a single-BH and free-breathing (FB), and 4&5) a sequential 2D technique in a single-BH and FB. Acquisition parameters between corresponding FB and BH acquisitions were identical.
Image Reconstruction: PDFF maps were reconstructed for each CSE acquisition, corrected for all relevant confounding factors, including R2* decay, multi-peak fat, phase errors, and temperature related frequency shifts (for phantom scans) [1,9,10].
Data Analysis: Regions-of-interest (ROIs) of ~2cm in diameter were placed in each of the phantom vials over the three central slices, in order to measure the average PDFF from the 3D, slice-interleaved 2D, and sequential 2D acquisitions. Measurements from the 2D acquisitions were compared to the 3D acquisition using linear correlation analysis.
From the human subject data, one reviewer analyzed all PDFF maps using circular ROIs placed within each of the nine Couinaud liver segments, positioned to avoid large vessels and bile ducts. The PDFF measurements from all segments were averaged to obtain a representative liver PDFF for each acquisition. Measurements from all four 2D acquisitions were compared to the 3D acquisition using Bland-Altman analysis.
Additionally, one reader with 15 years of experience in abdominal MRI scored the PDFF maps for motion artifacts (0=severe artifacts/non-diagnostic; 1=moderate artifacts, interferes with PDFF measurements in liver; 2=mild artifacts present, minimal/no interference with PDFF measurements in liver; 3=no identifiable artifacts). Scores were compared across sequences using a Mann-Whitney U-test.
Results
Figure 1 demonstrates equivalence between 2D and 3D CSE in PDFF phantoms.
Eleven men and 4 women were recruited (average (range)=47.6 (19-75) years). In these subjects, sequential 2D sequences showed good image quality and outstanding robustness to motion (Figure 2).
The qualitative motion artifact evaluation resulted in the following scores (average, (range)) for 3D: 2.2 (1-3), interleaved 2D BH: 2.9 (2-3), interleaved 2D FB: 0.7 (0-3), sequential 2D BH: 2.9 (2-3), sequential 2D FB: 2.9 (2-3). Scores for interleaved 2D BH and sequential 2D (BH or FB) were significantly higher than 3D, and scores for sequential 2D FB were significantly higher than interleaved 2D FB (p<0.01 in all cases).
Good agreement was observed in PDFF quantification between all 2D sequences and the standard 3D sequence (Figure 3), although there appeared to be some positive bias at low PDFF values for the 2D sequential acquisition.
Discussion and
Conclusions
Motion-robust liver PDFF quantification is feasible using a “single-shot” sequential 2D CSE technique. This technique has demonstrated excellent accuracy in phantoms, and good accuracy in liver imaging over a large range of PDFF.
Acquisition of sequential 2D data results in a loss in signal-to-noise ratio (SNR) relative to 3D or slice-interleaved 2D acquisitions, which may be the cause of the positive bias observed at low PDFF values. In future work, this may be partially mitigated through protocol optimization or advanced reconstruction methods that address the noise statistics of PDFF estimation in the setting of low SNR.
In summary, the proposed sequential 2D CSE technique shows excellent promise for motion-robust liver PDFF quantification in patients unable to hold their breath. This technique may prove useful for obtunded patients, as well as pediatric and fetal imaging applications in research and in the clinic.
Acknowledgements
The authors wish to acknowledge support from the NIH (UL1TR00427, R01 DK083380, R01 DK088925, R01 DK100651, K24 DK102595), as well as GE Healthcare.References
[1] Meisamy S, Hines CD, Hamilton G, et al. Quantification of hepatic steatosis with T1-independent, T2-corrected MR imaging with spectral modeling of fat: blinded comparison with MR spectroscopy. Radiology. 2011 Mar;258(3):767-75.
[2] Yokoo T, Bydder M, Hamilton G, et al. Nonalcoholic Fatty Liver Disease: Diagnostic and Fat-Grading Accuracy of Low-Flip-Angle Multiecho Gradient-Recalled-Echo MR Imaging at 1.5 T. Radiology. 2009;251(1):67-76.
[3] Reeder SB, Cruite I, Hamilton G, Sirlin CB. Quantitative assessment of liver fat with magnetic resonance imaging and spectroscopy. J Magn Reson Imaging. 2011 Oct;34(4):729-49.
[4] Hines CD, Frydrychowicz A, Hamilton G, et al. T(1) independent, T(2) (*) corrected chemical shift based fat-water separation with multi-peak fat spectral modeling is an accurate and precise measure of hepatic steatosis. J Magn Reson Imaging. 2011 Apr;33(4):873-81.
[5] Rehm JL, Wolfgram PM, Hernando D, et al. Proton density fat-fraction is an accurate biomarker of hepatic steatosis in adolescent girls and young women. Eur Radiol. 2015 Oct;25(10):2921-30.
[6] Bashir MR, Zhong X, Nickel MD, et al. Quantification of hepatic steatosis with a multistep adaptive fitting MRI approach: prospective validation against MR spectroscopy. AJR Am J Roentgenol. 2015 Feb;204(2):297-306.
[7] Motosugi U, Hernando D, Bannas P, et al. Quantification of liver fat with respiratory-gated quantitative chemical shift encoded MRI. J Magn Reson Imaging. 2015 Nov;42(5):1241-8.
[8] Hernando D, et al, Multi-site, multi-vendor validation of accuracy, robustness and reproducibility of fat quantification on an oil-water phantom at 1.5T and 3T, ISMRM 2015, #86.
[9] Yu H, Shimakawa A, Hines CDG, et al. Combination of Complex-Based and Magnitude-Based Multiecho Water-Fat Separation for Accurate Quantification of Fat-Fraction. Magnetic Resonance in Medicine. 2011;66(1):199-206.
[10] Hernando D, Sharma SD, Kramer H, Reeder SB. On the Confounding Effect of Temperature on Chemical Shift-Encoded Fat Quantification. Magnetic Resonance in Medicine. 2014;72(2):464-470.