The impact of shimming on fat suppression in head-and-neck MRI: current practice vs an image based approach
Tim Schakel1, Jeroen C.W. Siero2, Hans Hoogduin2, and Marielle Philippens1

1Radiotherapy, UMC Utrecht, Utrecht, Netherlands, 2Radiology, UMC Utrecht, Utrecht, Netherlands

Synopsis

In the head-and-neck region, off resonance effects due to magnetic field inhomogeneities can lead to poor fat suppression. In this study we compare the current clinical practice of shimming (volume shim) with the gain of an image based approach. B0 field maps are analyzed using water/fat segmented Dixon images to estimate the fat suppression. Diffusion weighted images are used to verify the estimates for fat suppression. An image based shimming optimization was performed to simulate 1st and 2nd order shim field. Image based shimming is a promising technique to improve subject specific shimming and fat suppression in the head-and-neck region.

Purpose

Off resonance effects caused by magnetic field (B0) inhomogeneities can lead to imaging artefacts. This is especially pronounced in areas such as the head-and-neck where the complex anatomy and large susceptibility variations induce large B0 inhomogeneities. Therefore, frequency selective fat suppression in this region is prone to artefacts. Poor fat suppression can make images harder to interpret and obscure pathology.

The application of shim fields generated by dedicated shimming gradients could compensate for these off resonance effects. In this study we evaluate the current practice of shimming in the head-and-neck region at 3.0 T. We also investigate the potential gain of an image based shimming approach [1,2,3] for effective fat suppression in this region.

Methods

Data was acquired in 10 patients at a 3.0 T Ingenia widebore system (Philips Healthcare, The Netherlands). Patients were scanned in an immobilization mask using flexible receive coils compatible with the mask. B0 field maps were acquired using a 3D gradient echo sequence with 2 echoes: TE/∆TE/TR: 4.6/2.3/9.5 ms; FOV 280 x 280 x 120 mm; voxel size 1.45 x 1.45 x 4.0 mm (AP x LR x FH). These field maps are acquired with the same shim settings (volume shim) as the rest of the sequences in the protocol. Water/fat information was extracted from a processed T2w mDixon sequence, locating the voxels in which fat (red) or water (blue) is dominant. A diffusion weighted TSE based sequence [4] with SPIR was used to assess the fat suppression.

Current clinical practice is to position a rectangular ‘shimbox’ around the area of interest. The shimbox is kept inside the tissue. The shim is optimized within this shimbox, this method is referred to volume shimming and utilizes first order gradients.

An image based shimming algorithm [1] was implemented in Matlab, which uses the second order gradients to generate more complex shim fields. The shimming algorithm minimizes an input field map with calculated shim fields using either first or second order terms. A mask which includes the whole patient contour, containing the center ⅔ slices, was used as an additional input for the shim algorithm in order to have the underlying anatomy align with that used in the rectangular shimbox. Using the shimmed B0 field maps (volshim), water/fat segmentation and the settings of the frequency selective fat suppression pulse, we could estimate the number of unsuppressed fat voxel. The same is done after recalculating the shim fields (IB1 and IB2) in Matlab.

Results

Figure 1 shows data from 2 patients. The first column shows the water/fat segmentation with the water in blue and fat in red overlay-ed over the B0 field map (volshim). The estimated fat suppression is shown in the 2nd column, where failing fat suppression is indicated with red voxels. The estimate is confirmed by the diffusion weighted scan (column 3), where the unsuppressed fat appears bright (red arrow). The simulations of the shim field using 1st and 2nd order shim gradients (IB1 and IB2, last 2 columns) show a large estimated improvement of the fat suppression (red voxels: unsuppressed fat).

The impact of shimming can also be visualized with the frequency spectrum (figure 2). The solid lines (blue, water; red, fat) are from the current clinical practice, volshim. The dashed lines show the same spectra after optimization of the shims (2nd order).

The results for all 10 patients are summarized in table 1, where all patients show a decrease in the amount of unsuppressed fat using the image based shimming approach. Two groups of patients can be identified: one with high numbers of unsuppressed fat voxels using volshim (first 5 patients), and one where one where the initial shim and thus fat suppression is already performing well (last 5 patients). The first group shows the largest improvement for IB1 and IB2 compared to the current clinical practice.

Discussion/Conclusion

Visual inspection showed agreement between areas of estimated and actual failing fat suppression. Simulations of an alternative shimming strategy showed large improvements in terms of fat suppression. The image based shimming approach can be extended by using different cost functions to optimize the shim fields and by incorporating water/fat information in the optimization.

The current clinical practice using volume shim suffers from off resonance effects leading to failing fat suppression. Image based shimming is a promising technique to improve subject specific shimming and fat suppression in the head-and-neck region.

Acknowledgements

No acknowledgement found.

References

[1] J.C.W. Siero et al. Proc. Intl. Soc. Mag. Reson. Med. 18 (2010) 2589

[2] V.O. Boer et al. NMR Biomed. 2011 Nov;24(9):1081-8

[3] M.E.P. Philippens et al. Radiotherapy and Oncology , Volume 92 , S128

[4] F. Schick. Magn Reson Med. 1997 Oct;38(4):638-44

Figures

Figure 1: 2 patients (rows), with from left to right: B0 field map (water (blue)/fat (red) segmentation overlay); B0 field map (volshim); DW-TSE-SPLICE (volshim); B0 field map (simulated 1st order IBshim); B0 field map (simulated 2nd order IBshim). Yellow arrow: positive lymph node.

Figure 2: Frequency spectra of 1 patient. Blue/Red solid: B0 field map (volshim). Blue/red dashed: B0 field map (2nd order shim). Black vertical lines: frequency selective fat suppression pulse.

Table 1: Performance of fat suppression in current clinical practice and simulated performance of an image based approach.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1333