Simone Angela Winkler1, Andrew Alejski2, Trevor Wade2, Charles McKenzie2, and Brian K Rutt1
1Dept. of Radiology, Stanford University, Stanford, CA, United States, 2Robarts Research Institute, The University of Western Ontario, London, ON, Canada
Synopsis
We hypothesized that
ceramic inserts will reduce sound pressure levels (SPLs) in high performance
head gradient coils. We used realistic multi-physics modeling methods (previously
validated by experiments) to investigate this hypothesis, and in particular to
evaluate vibroacoustic reductions as a function of ceramic insert geometry and
frequency of excitation. Averaged over the range 0-3000Hz, we demonstrate a
maximum overall SPL reduction of 10.9dB, with a substantially higher reduction
in the high frequency regime (2000-3000Hz) of 20.7dB. We show that a uniform 15mm
thick cylindrical insert is a practical design that yields the majority of the
acoustic reduction benefit.Background
The conductors of MR gradient coils are subject to large Lorentz forces.
These forces can cause sound pressure levels (SPLs) inside the gradient coil in
excess of 100dB, which can become a serious safety issue for ultra-high-field high
performance head gradient coils. It has been found that increased stiffness of
the coil material decreases SPLs and vibration levels significantly [1]. However,
it is impractical to fabricate the entire coil from a stiffer material due to
mechanical/weight considerations. We propose here the approach of using ceramic inserts attached to the inner
surface of the head gradient coil, to strategically stiffen the coil (this
inner surface acts as the main source of acoustic wave generation). Our goal in
this work was to evaluate the validity of this strategy and to optimize the
ceramic insert design. Toward this end, we used our previously developed and
experimentally validated methods for accurate modeling of MR gradient
vibroacoustics [2-4].
Methods
We conducted realistic numerical modeling using COMSOL.
Our head gradient design was a novel folded shielded gradient design for human
brain imaging [5]. The coil structure was modeled as an epoxy cylinder (inner/outer
diameter 338/490mm, length 450mm) with accurate conductor wire patterns (Fig.1).
We analyzed three different ceramic insert configurations
(Fig.2): (a) a cylindrical ceramic layer of constant thickness (thickness
range 5-20mm); (b) a 20mm thick cylinder in combination with a stepped section of 50mm thickness extending over 200mm at the
service end of the gradient coil; and (c) a 20mm thick cylinder in combination with
a 200mm thick “end-cap” completely filling the bore at the service
end. Both (b) and (c) leave an imaging region of 250mm in length – adequate room
to center the brain at the gradient isocenter. Ceramic (98%
alumina) was modeled as a linear elastic material:
E=300GPa, ρ=3900 kg/m3, ν=0.22. Epoxy was modeled as:
E=10 GPa, ρ=1600 kg/m3, ν=0.4.
The analysis followed all aspects of our previously developed method for
realistic gradient coil vibroacoustic modeling [2-4], including the effect of
Lorentz damping due to vibration-induced eddy currents in the conductors using
copper wire of radius 3mm [4]. The simulation was carried out at 3T using a
harmonic excitation with an AC current of amplitude 50A over a frequency range
of 0‑3000Hz, which spans the frequency content of most pulse sequences.
Results
Fig.3 shows the SPLs for the proposed ceramic inserts. A
frequency-averaged SPL reduction of 10.9dB can be achieved by using a 20mm
ceramic layer insert alone, with the first 15 mm contributing the majority (10.0dB).
The largest frequency-averaged SPL reduction is achieved with a plugged insert:
16.8dB. Fig.3 also shows the frequency dependence of SPL reductions for three
different frequency bands. We observe that the greatest SPL reductions are
achieved in the high frequency band (Fig.3; green), with a maximum SPL reduction
of 20.7dB, reached mostly by adding the first 15mm of ceramic inner layer. This
high frequency regime also benefits the most from a ceramic layer alone without
the added stepped or plugged insert. In contrast, the added stepped or plugged
ceramic feature reduces SPLs most strongly in the lower frequency ranges (Fig.3;
blue, red), with SPL reduction in the range of 8dB contributed by these
features.
Discussion
Our results show the potential for significant SPL reduction in MR
gradient coils by use of ceramic inserts. One practical concern is the added
weight – for example, 5mm/20mm thick straight cylinders would add 20.4kg/35.1kg,
respectively, and the stepped/plugged configurations would add 61.5kg/89.5kg,
respectively. In this regard, we note that the additional plugged/stepped insert
mostly benefits the low-/intermediate-frequency bands, where acoustic levels start
out 10-25dB lower than in the high-frequency band. Thus the most important reduction
(in the high-frequency band) is achievable without stepping/plugging, using a 15mm
straight cylindrical layer, with only moderately increased weight. The added ceramic
layer is also expected to improve thermal heat conduction and therefore
minimize thermal hotspots on the all-important inner gradient surface. Another
limitation of ceramic is mechanical stability; these properties will be
carefully investigated, although ceramic potting compounds and castable
ceramics are available with greatly improved mechanical properties. Further experiments
will validate our proposed concept. With the combination of the proposed
concept with other noise reduction solutions such as the addition of a horn [2]
or acoustic barrier foam, we anticipate SPL reductions in the range of 30dB.
Conclusion
A new framework for comprehensive vibroacoustic analysis of MR gradient
coils has been used to study the SPL reductions achievable using ceramic
inserts. A maximum spectrally-averaged SPL reduction of >10dB has been
demonstrated, with >20dB within a high-frequency band.
Acknowledgements
Research
support from NIH (P41 EB015891, 1 S10 RR026351-01Al) and GE Healthcare. The
authors thank Graeme McKinnon and Scott Hinks for their valuable input.References
[1] Doty FD, et al, Proc Intl Soc Mag Reson Med 13:406 (2005). [2] Winkler SA, et al, Proc Intl Soc Mag Reson Med 22:4852 (2014). [3] Winkler SA, et al, Proc Intl Soc Mag Reson Med 23:3089 (2015). [4] Winkler SA, et al, Proc Intl Soc Mag Reson Med 23:1020 (2015).[5] Wade TP, et al, Proc Intl Soc Mag Reson Med 22:4851 (2014).