Xiaowan Li^{1}, Conlin C Christopher^{1}, Christopher Hanrahan^{1}, Vivian S Lee^{2}, and Jeff Lei Zhang^{1}

Intramuscular fat is an
important indicator of the health of the lower extremities. It can be difficult
with conventional Dixon techniques to acquire high resolution fat-fraction maps,
which may be helpful for detailed
analysis of the muscle tissue. We propose to use a linear model to generate a
high resolution fat-fraction map from high resolution T_{2}-weighted images.
The results show that the proposed method and the conventional Dixon method
have good agreement. The high resolution
fat fraction maps generated with the proposed method reveal anatomic details of
intramuscular fat not visible with the conventional method.

**Introduction**

**Methods**

MRI data acquisition

In this IRB-approved study, 10 datasets from different axial calf slices were
acquired from an obese volunteer (female, age: 54 years, BMI: 36) with significant
fat infiltration of the calf muscles. Each dataset consisted of a two-point Dixon acquisition and a multi-echo T_{2}-weighted
acquisition centered on the
same slice. Imaging was
performed using a 3T clinical scanner (Prisma^{fit}; Siemens) with an
8-channel receiver coil. Conventional two-point Dixon MRI was performed using a
gradient echo sequence: TR 4.12ms, TE 1.23, 2.46ms, flip angle 9°, FOV
140×346mm, resolution 1.35×1.35mm, slice thickness 3mm. From each Dixon
acquisition, a fat-fraction map ($$$FF_{Dix}$$$)
was calculated using the fat-only and water-only images. High-resolution two-dimensional T_{2}-weighted images
were acquired using a turbo spin-echo sequence: TR 1000ms, TE 12, 81, 151ms,
flip angle 180°, FOV 140×346mm, resolution 0.34x0.34mm, slice thickness 3mm.
Each
series of T_{2}-weighted images was fit voxel-wise to an exponential-decay function to
generate a T_{2} map. This
imaging procedure was repeated on 5 different axial slices of the left and
right calf.

Linear regression between intramuscular
fat fraction and T_{2}

For each dataset, the voxel-wise T_{2} map was
first normalized using the T_{2} of subcutaneous fat. A linear relationship
was proposed to characterize the normalized T_{2} and its corresponding
fat fraction:

$$FF_{T2} = a*T_2+b$$

where *a* and *b* are the slope and intercept of the regression line. We chose a linear model based on a previously-reported linear correlation between T_{2} and fat fraction^{4}. Optimal values for *a* and *b* were determined using one acquired dataset, by minimizing the mean difference between $$$FF_{Dix}$$$ and $$$FF_{T2}$$$. The optimized regression equation was then used to derive $$$FF_{T2}$$$ maps from the T_{2} map of the other 9 datasets, which were compared to the corresponding $$$FF_{Dix}$$$ for assessment.

**Results**

**Discussion**

1. Lovitt S, Moore SL, Marden FA. The use of MRI in the evaluation of myopathy. Clin Neurophysiol. 2006;117(3):486-495.

2. Grimm A, Meyer H, Nickel MD, et al. Evaluation of 2-point, 3-point, and 6-point Dixon magnetic resonance imaging with flexible echo timing for muscle fat quantification. Eur J Radiol. 2018;103:57-64.

3. Diaz-Manera J, Llauger J, Gallardo E, Illa I. Muscle MRI in muscular dystrophies. Acta myologica : myopathies and cardiomyopathies : official journal of the Mediterranean Society of Myology. 2015;34(2-3):95-108.

4. Gloor M, Fasler S, Fischmann A, et al. Quantification of fat infiltration in oculopharyngeal muscular dystrophy: comparison of three MR imaging methods. Journal of magnetic resonance imaging. 2011;33(1):203-210.

Figure 1. Comparison between representative
fat-fraction maps obtained using a conventional two-point Dixon method and the proposed T_{2}-regression method. (a) High
resolution T_{2} map. (b) Fat-fraction map estimated from the T_{2} map using the proposed method, after
downsampling to the same resolution as the Dixon acquisition. (c) Fat fraction map from the conventional Dixon technique. (d) Fat-fraction map estimated from the high resolution
T_{2} map using the proposed method.

Figure 2: Bland-Altman plot showing the agreement between the
conventional Dixon method and the proposed T_{2}-regression method using the fat-fraction
of all muscle groups across all test cases. The dashed blue line corresponds to the mean difference between
fat-fraction estimates from the two methods, and the dashed red lines mark
the 95% confidence interval.