Improved olefinic fat suppression in skeletal muscle DTI using a magnitude‐based dixon method

    loading  Checking for direct PDF access through Ovid


In diffusion‐weighted MRI acquisitions, complete fat suppression is essential for obtaining correct quantitative measurements of diffusion coefficients 1, especially in applications in which there is a significant fat content within the region of interest. Inaccurate values result in incomplete fat suppression in voxels containing both fat and water, as well as tissue voxels in which the fat signal appears due to the chemical shift displacement. An example is in muscular dystrophies 3; one of the main manifestations of the disease is the replacement of muscle tissue by fat. The most commonly used fat‐saturation methods, such as spectrally adiabatic inversion recovery (SPAIR), can achieve robust suppression of the main methylene (−CH2) and methyl (−CH3) peaks; however, the olefinic peak at 5.3 ppm with a resonance frequency close to the water signal is not suppressed by SPAIR, which means that there is some residual fat signal 4. Because the olefinic peak contributes only between 5% and 10% of the overall fat signal, in many applications this is an acceptable residual that can be ignored 2. However, this is not true in diffusion‐weighted acquisitions because the diffusion coefficient of fat is two orders of magnitude lower than that of water 7. Thus, even at low‐diffusion weighting factors the residual olefinic fat signal can be a significant fraction of the overall signal, thus confounding quantitative measurements 8. As an example, in a voxel containing equal amounts of water and total fat signal, olefinic fat would increase the signal intensity by approximately 5%. However, at a diffusion weighting of b = 400 s2/mm (typical for muscle imaging) in the same situation, the water signal would decrease by approximately 55%, whereas the olefinic fat signal would remain almost unchanged, contributing around 20% to the overall signal in the voxel.
It should also be noted that standard Dixon‐based methods cannot be readily used in diffusion‐weighted acquisitions because they rely on phase data, which is changed to a pseudo‐random value by the diffusion gradients and cannot be deconvoluted when the chemical shift of the main fat peak is larger than a few millimeters 2; therefore, a different approach is necessary.
Williams et al. proposed a method for fat‐suppressed diffusion‐weighted imaging (DWI) in which the main peak is suppressed using a combination of SPAIR and slice‐selection gradient reversal (SSGR), whereas the olefinic fat peak is suppressed with a second spectral fat saturation method 11. This technique offers good suppression for most of the fat spectrum, although it is susceptible to main field (B0) inhomogeneities and causes the loss of approximately 10% of the water signal due to the spectral proximity of the olefinic fat and water peaks 11. For brevity, through the rest of this work we refer to this method as spectral olefinic fat suppression (SOFS).
Hernando et al. also used SPAIR to suppress the −CH2 and −CH3 peaks and described how magnitude‐only diffusion‐weighted data from a six‐echo (commonly described as 6‐point) acquisition can be separated into images corresponding to water and olefinic fat with a multiparameter fit, using the diffusion‐free (b = 0) image for initialization of the fitting parameters 2. Because this method involves a nonlinear multi‐parameter fit performed on the data to separate water and fat, it has a potential to be unstable if only a few echoes are used. In addition, it requires either using a partial Fourier acquisition to reconstruct complex data, which decreases the resolution in the phase domain 12, or making a full k‐space acquisition, which in a typical acquisition means an echo time (TE) increase of 30 ms.

Related Topics

    loading  Loading Related Articles