Daiki Tamada^{1}, Hiroshi Onishi^{1}, and Utaroh Motosugi^{1}

An MR Fingerprinting (MRF) simultaneously combining the three-point DIXON (3P-DIXON) method for the fatty liver was proposed. The MRF-FISP sequence with multi-TR/TE/flip angle was developed. The six-peak fat model was used to calculate a dictionary for the MRF. Template matching using the acquired signal evolutions and the rough fat fraction map estimated by 3P-DIXON provided quantification of T1, T2, and

**ACQUISITION:**

The MRF-FISP sequence with 800 acquisition using spiral readout (2048 sampling point with dwell time of 2 microseconds) was used. The spiral trajectory with zero-moment compensated rotates 31 degrees every repetition. Variable flip angles (Fig .1 (b)) and TRs (Fig .1 (c)) were generated based on sinusoidal curves and a Perlin noise approach. TEs of 2.2 and 3.3 ms were used for the first 200 and the subsequent 400 echoes, respectively. The rest of 200 echoes were acquired with fixed FA (5°) and TR (11 ms) and variable TE (1.1, 2.2, and 3.3 ms) for the 3P-DIXON reconstruction. The scan time was 11 second, which is acceptable for one breath-hold acquisition. The phantom and volunteer studies were performed using 3 tesla clinical MRI (Discovery MR750, GE Healthcare) with 32-ch torso body coil. The phantom consists of 5 acrylic tubes with varying concentration of gadolinium and agarose.

**DICTIONARY:**

A dictionary containing 230600 signal evolutions was calculated. The signal evolution in the liver can be expressed as a linear combination of water and fat with six-peak model components as below.

$$$S(T_1, T_2) = (1-\alpha)W(T_1, T_2) + \alpha \sum_{i=1}^n {F_i e^{-j \phi_i t}},$$$

where α is the fat fraction, W is the signals of water, F_{i} and Φ_{i} are the signal and off-resonance frequency for i_{th} fat peak. The coefficients and off-resonance in our study were adopted from a literature which introduced 6-peak fat model8. The dictionary contained possible T1 of water (20~2000 ms), T2 of water (10~500 ms), B1 (80%-110%), and fat fraction (0-100 %) for the liver. Fixed T1 and T2 for the fat were used, that were measured in peanuts oil using multi-TR/TE STEAM. To accelerate the signal matching, the dictionary was compressed using the SVD approach^{9}.

**RECONSTRUCTION:**

The signal matching was achieved using the simple template matching combining with a rough fat fraction map derived using the 3P-DIXON. Each echo was reconstructed using the non-uniform Fourier transform with parallel imaging. The signal evolutions, which was projected on to SVD space, was matched to the compressed dictionary. As describe in Fig. 2, the matching was performed using the dictionary which has fat fractions around the roughly estimated value (±5 %). For the comparison, quantitative values of T1, T2, and fat fraction were measured using multi-TR/TE STEAM^{10} , IDEAL-IQ with variable flip angles (IDEAL-VFA)^{11}, and multi-echo/IR-spin echo (SE) sequences. The voxel for the STEAM was placed on the right lobe.

The multi-TR/TE STEAM sequence used in this study was provided by Dr. Gavin Hamilton at University of California San Diego.

This study was funded by GE Healthcare Japan (Hino, Japan).

[1] Reeder, Scott B., and Claude B. Sirlin. "Quantification of liver fat with magnetic resonance imaging." Magnetic Resonance Imaging Clinics 18.3 (2010): 337-357.

[2] Chen, Yong, et al. "MR fingerprinting for rapid quantitative abdominal imaging." Radiology 279.1 (2016): 278-286.

[3] Koolstra , Kirsten, et al. “Water-Fat Separation in Spiral Magnetic Resonance Fingerprinting using Conjugate Phase Reconstruction.” Proc. ISMRM 2018: 0681

[4] Ostenson, Jason, et al. “Fat Signal Fraction Determination Using MR Fingerprinting”. Proc. ISMRM 2017: 0134

[5] Hines, Catherine DG, et al. "T1 independent, T2* corrected chemical shift based fat–water separation with multi‐peak fat spectral modeling is an accurate and precise measure of hepatic steatosis." Journal of magnetic resonance imaging33.4 (2011): 873-881..

[6] Jiang, Yun, et al. "MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout." Magnetic resonance in medicine 74.6 (2015): 1621-1631.

[7] Hardy, Peter A., R. Scott Hinks, and Jean A. Tkach. "Separation of fat and water in fast spin‐echo MR imaging with the three‐point Dixon technique." Journal of Magnetic Resonance Imaging 5.2 (1995): 181-185.

[8] Yu, Huanzhou, et al. "Multiecho water‐fat separation and simultaneous R estimation with multifrequency fat spectrum modeling." Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine 60.5 (2008): 1122-1134.

[9] McGivney, Debra F., et al. "SVD compression for magnetic resonance fingerprinting in the time domain." IEEE transactions on medical imaging 33.12 (2014): 2311-2322.

[10] Hamilton, Gavin, et al. "In vivo breath‐hold 1H MRS simultaneous estimation of liver proton density fat fraction, and T1 and T2 of water and fat, with a multi‐TR, multi‐TE sequence." Journal of Magnetic Resonance Imaging 42.6 (2015): 1538-1543.

[11] Tamada, Daiki, et al. "Multiparameter estimation using multi‐echo spoiled gradient echo with variable flip angles and multicontrast compressed sensing." Magnetic resonance in medicine (2018).

[12] Ostenson J et al. “Multi-Frequency Interpolation in Spiral Magnetic Resonance Fingerprinting for Correction of Off-Resonance Blurring.”, Magnetic Resonance Imaging. 2017;41: 63–72.

Fig. 1 Pulse sequence for the proposed method which includes 800 excitations. The MR signal evolution used for the matching was acquired in first 600 echoes. The rest of 200 echoes was used for the 3P-DIXON reconstruction used to estimate the rough fat fraction and B0 maps.

Fig. 2 Overview of the signal matching in this study. To achieve fast and stable template matching, the dictionary which has fat fractions around (±5 %) the estimated value was used.

Fig.3 Measured T1, T2 and fat fraction of the phantom using the proposed MRF method (FOV=(18 cm)^{2}, matrix = (256)^{2}). T1 and fat fraction measured with MRF well agreed with STEAM (TR =150-2000 ms, TE = 10-110 ms, voxel size = (2 cm)^{3} , 32 spectra) and IDEAL-VFA (IDEAL-VFA, TR/TE/ΔTE = 6.7/1.0/0.9 ms, FA = 3 and 19°, FOV = (18 cm)^{2} matrix = 160 × 160 × 24, slice thickness = 4.0 mm, 6 echoes) although underestimation of T2 in MRF was observed.

Fig. 4 Measured T1, T2, and fat fraction maps measured with the MRF (FOV=(42 cm)^{2}, matrix = (256)^{2}).

Tab. 1 The quantitative values in right robe. The values measured with MRF agreed with those of STEAM (TR =150-2000 ms, TE = 10-110 ms, voxel size = (2 cm)3 , 32 spectra) and IDEAL-VFA (IDEAL-VFA, TR/TE/ΔTE = 6.7/1.0/0.9 ms, FA = 3 and 19°, FOV = (42 cm)2 matrix = 160 × 160 × 24, slice thickness = 4.0 mm, 6 echoes).