Introduction

The goal of this work is to extend cardiac MR Fingerprinting (cMRF) with a rosette-based readout for T₁, T₂, and fat fraction mapping. cMRF is a novel multi-parametric mapping method that was initially shown to enable simultaneous myocardial T₁ and T₂ quantification¹. While many MRF approaches do not distinguish water from fat, elimination of the signal from fat may reduce mapping errors caused by water-fat partial volume effects. Moreover, quantitative proton density fat fraction (PDFF) mapping may provide additional value in diagnosing disease^2,3. Previously, the Dixon method has been incorporated in the cMRF framework using multi-echo radial acquisitions to enable T₁, T₂, and PDFF quantification⁴. Previous work has demonstrated that water-fat separation can be achieved along with myocardial T₁ and T₂ mapping using cMRF with rosette trajectories⁵. In this work, rosette cMRF is extended to enable simultaneous myocardial T₁, T₂, and PDFF mapping from a single scan.

Methods

A rosette trajectory with eight lobes and a readout duration of 7.7ms (Figure 1a) was designed to suppress signals at -220Hz (the main resonance frequency of fat at 1.5T). Simulation studies show that this trajectory suppresses 94.7% of the signal at -220Hz (Figure 1b). This readout trajectory was incorporated into a previously reported 15-heartbeat 5-segment ECG-triggered cMRF sequence structure⁵ with acquisition window ~250 ms at late diastole, resulting in a total of 390 highly undersampled images for pattern matching. The spatial resolution was 1.56×1.56mm² and the slice thickness was 8mm in all phantom and in vivo experiments.
All experiments were performed on a 1.5T scanner (Siemens Sola, Erlangen, Germany). Rosette cMRF data were collected in the T₂ layer of the ISMRM/NIST MRI system phantom ^6,7 and an in-house fat fraction phantom⁸. T₁ and T₂ values obtained using rosette cMRF were compared with gold standard values measured using inversion recovery and single-echo spin echo methods. Three-point Dixon GRE measurements with optimal echo times at 1.5T (1.9/3.4/4.9ms) were used as the gold-standard PDFF values. 14 healthy subjects were scanned after written informed consent in this IRB-approved study. Mid-ventricular level short axis slices in the heart were acquired using the proposed rosette cMRF sequence and the original spiral 15-heartbeat cMRF sequence.⁵ Conventional T₁ and T₂ mapping methods (MOLLI and T₂-prepared bSSFP) were also performed in 10 of the subjects.
Undersampled water and fat images and a B₀ map were generated using the rosette cMRF approach, as described in previous work⁵. T₁ and T₂ maps for both water and fat were generated by matching undersampled images to a dictionary that models the subject’s cardiac rhythm⁵ and includes corrections for slice profile, preparation pulse efficiency⁹, and B₀⁵. To generate quantitative PDFF maps, the 8-lobe trajectory was divided into 9 segments (Figure 1a). Data from each segment were projected onto a low-dimensional subspace of rank 6 derived from the singular value decomposition (SVD) of the dictionary. Subspace images corresponding to the first singular value from each of the 9 rosette segments served as multi-echo images and were used to generate a PDFF map using Hierarchical IDEAL¹⁰ with noise correction¹¹.
ROIs in each vial in the phantom study, and on T₁ and T₂ maps in segments 7-12 of the standardized AHA model for in vivo measurements, were drawn manually. The mean and standard deviation in T₁ and T₂ values were calculated for each ROI, as well as over the entire myocardium in the in vivo maps. A student’s t-test was used to compare T₁ and T₂ measurements using different methods. Significant difference was considered with P<0.05.

Results

In the ISMRM/NIST system phantom, T₁ and T₂ measurements using rosette cMRF are in excellent agreement with the reference values (Figure 2). In the fat fraction phantom, water and fat specific T₁ and T₂ maps, proton density images, and the PDFF map generated by Hierarchical IDEAL using the rosette cMRF data themselves are shown in Figure 3a. Quantitative PDFF measurements using rosette cMRF agree well with 3-point Dixon GRE measurements (Figure 3b).
Representative maps and images from one healthy subject are shown in Figure 4. The averaged T₁ and T₂ values of all subjects in each segment as well as in the entire myocardium are shown in Figure 5. Over the myocardium, spiral cMRF yielded similar T₁ values (1004.7 ± 53.6 ms) compared with MOLLI (994.8 ± 18.8 ms) while rosette cMRF generated significantly higher T₁ values (1081.8 ± 33.6 ms); both spiral and rosette cMRF yielded lower T₂ values (spiral 37.3 ± 2.9 ms; rosette 40.4 ± 1.4 ms) compared with the conventional method (45.8 ± 2.4 ms), which is consistent with previous findings⁵. Averaged PDFF in the entire slice was 0.6% among all volunteers.

Discussion and Conclusions

This study shows that rosette cMRF can enable efficient cardiac tissue characterization through the simultaneous quantification of myocardial T₁, T₂, and PDFF. Note that in segment 7 (anterior wall) which is surrounded by epicardial fat, the difference in T₁ measured by rosette cMRF and spiral cMRF is more pronounced than in other segments; spiral cMRF yielded higher T₂ values than rosette cMRF, which is opposite to the trend in the other segments (Figure 5). Given that fat has T₁ of 300~370 ms (lower than myocadium) and T₂ of ~53ms (higher than myocardium) at 1.5T¹², these measurements are possibly due to reduced fat contamination in rosette cMRF. Studies on cardiac patients with abnormal fat deposition will be performed in the future.

Acknowledgements

This work is supported by NSF/CBET 1553441, NIH/NHLBI R01HL094557, and Siemens Healthineers (Erlangen, Germany).

References

1. Hamilton, J. I. et al. MR fingerprinting for rapid quantification of myocardial T1, T2, and proton spin density. Magn. Reson. Med. 77, 1446–1458 (2017).

2. Farrelly, C., Shah, S., Davarpanah, A., Keeling, A. N. & Carr, J. C. ECG-gated multiecho Dixon fat-water separation in cardiac MRI: Advantages over conventional fat-saturated imaging. Am. J. Roentgenol. 199, (2012).

3. Kellman, P. et al. Multiecho dixon fat and water separation method for detecting fibrofatty infiltration in the myocardium. Magn. Reson. Med. 61, 215–221 (2008).

4. Jaubert, O. et al. Water–fat Dixon cardiac magnetic resonance fingerprinting. Magn. Reson. Med. 00, 1–17 (2019).

5. Liu, Y., Hamilton, J., Eck, B., Griswold, M. & Seiberlich, N. Myocardial T1 and T2 quantification and water–fat separation using cardiac MR fingerprinting with rosette trajectories at 3T and 1.5T. Magn. Reson. Med. (2021) doi:10.1002/mrm.28404.

6. Keenan, K. et al. Multi-­site, multi-­vendor comparison of T1 measurement using ISMRM/NIST system phantom. in Proc Int Soc Magn Reson Med. 24 3290 (2016).

7. Russek, S. et al. Characterization of NIST/ISMRM MRI system phantom. in Proc Int Soc Magn Reson Med. 20 2456 (2012).

8. Hines, C. D. G. et al. T1 independent, T2* corrected MRI with accurate spectral modeling for quantification of fat: Validation in a fat-water-SPIO phantom. J. Magn. Reson. Imaging 30, 1215–1222 (2009).

9. Hamilton, J. I. et al. Investigating and reducing the effects of confounding factors for robust T1 and T2 mapping with cardiac MR fingerprinting. Magn. Reson. Imaging 53, 40–51 (2018).

10. Tsao, J. & Jiang, Y. Hierarchical IDEAL: Fast, robust, and multiresolution separation of multiple chemical species from multiple echo times. Magn. Reson. Med. 70, 155–159 (2013).

11. Liu, C. Y., McKenzie, C. A., Yu, H., Brittain, J. H. & Reeder, S. B. Fat quantification with IDEAL gradient echo imaging: Correction of bias from T1 and noise. Magn. Reson. Med. 354–364 (2007) doi:10.1002/mrm.21301.

12. Rakow-Penner, R., Daniel, B., Yu, H., Sawyer-Glover, A. & Glover, G. H. Relaxation times of breast tissue at 1.5T and 3T measured using IDEAL. J. Magn. Reson. Imaging 23, 87–91 (2006).