Ed Boskamp1
1Hyperfine, Guilford, CT, United States
Synopsis
Keywords: Physics & Engineering: Hardware
Motivation: This course is intended for engineers and physicists who want to design RF coils
Goal(s): After this course you will be able to make simple transmit and receive coils for low and high field
Approach: We will review design requirements for low and high field Tx and Rx coils, SNR, SAR, EM field-tissue interaction, decoupling
Results: High field and low field require very different designs
Impact: Coil designers will appreciate the differences in low and high field RF coil designs and will understand the difference in design requirements
Introduction
When I
started building RF coils for MRI in 1983 it was for the strong magnetic field
of 0.3 T. In the 41 years that have passed I built a large number of coils and
arrays between 0.064 and 9.3T. Each
of these field strengths pose unique challenges to RF coil design. We will be
addressing:
- ·
The
difference in requirements for Transmit and Receive coils
- ·
EM
field and tissue interaction at UHF
- ·
SNR
- ·
Decoupling
between Transmit and Receive coils
- ·
SAR
- ·
Ultra
low field coil design and its challenges
Transmit (Tx) coil versus receiver (Rx) coil requirements
These coils have very different design requirements. During
the RF transmit pulse we want to make sure we are creating a uniform flip angle
distribution across the excitation volume, and we want to do it efficiently and
without exceeding SAR limits and without creating image artifacts from areas
where the gradient fields are non linear.
This leads
to the following requirements for
Tx coils:
- ·
The
RF magnetic field in the imaging volume needs to be uniform but needs to drop
off rapidly outside the imaging volume, defined by Magnetic field uniformity
and gradient linearity.
- ·
All
RF energy should be used towards the creation of a circularly polarized RF
magnetic field without waste.
- ·
The
input impedance of the Tx coil shall be equal to the characteristic impedance
of the transmission line (coax cable) to prevent
reflection of power at the input. This is typically 50 ohm and is achieved by
impedance matching.
- ·
We
shall not waste RF energy in the gradient coil, hence body size RF transmit
coils are typically surrounded by an RF shield. Such a
shield will keep the RF fields out of the gradient coil environment but will be
transparent for the gradient pulses.
- ·
Capacitors
in the resonant Tx coil shall be distributed so that we do not create large E fields.
During the
read out phase, the receiver coils (typically an array) pick up the MRI signal
and turn it into a voltage which will then be processed to create
the image. The MRI signal can be very small, so it is important to keep noise to a minimum. The #1 requirement is: maximize
SNR. Since the
SNR is proportional to √Q, maximizing coil Q is one important aspect of
optimizing SNR. Here is where high field and low field coils demand different
approaches: At high field, the receiver coil Q is determined by RF losses in
the patient and when coil Q is improved to a point where the patient losses are
dominant, it does not pay putting a lot of money and effort into
improving the Q of the coil further, since the gains will be marginal. At low
field and ultra low field, the patient load hardly affects the Q, hence any gain in coil Q will improve the SNR.
This leads
to the following requirements for
Rx coils:
- ·
The
Q of the coil by itself should be increased until patient losses dominate the Q. In Ultra low field systems the Q of the Rx
coil needs to be as high as possible, given other constraints.
- ·
Every
Rx coil has a circuit board attached containing T/R switches, tuning and
matching networks, varactor diodes, Low Noise amplifier (LNA) and so on. The
dielectric losses in the circuit board need to be kept to a minimum. This can be achieved
through layout and substrate choices.
- ·
The
LNA should be low noise, and the amplification of that LNA should
be high enough so that the Noise figure of that first stage dominates the NF of
the entire receive chain.
- ·
The
impedance of the Rx coil shall be matched to the impedance the LNA wants to be
connected to for lowest Noise Figure performance.
- ·
The
coil shall be as close to the anatomy of interest as possible and should be
shaped like the anatomy.
RF Transmit field uniformity
The transmit field needs to be uniform and preferably
circularly polarized (B1+). There are
several ways to create circularly polarized fields, but the most popular is the
birdcage resonator introduced by Hayes at all in 1985 (1). A typical birdcage RF current distribution is shown in Fig 1. Other types of
circularly polarized coils are needed for scanners that have a vertical magnetic
field. At low to medium fields i.e. 1.5T and below the uniformity of the RF
magnetic field is not affected much by the patient tissue, but that all changes
when the RF wavelength gets to be in the same order of
magnitude as the patient size. In tissue the wavelength is shorter by 1/√ε in which
ε is the permittivity of the tissue at the frequency of interest. The result of the short wavelength is that we get standing waves and non uniform images. The only way to improve this is
to create more degrees of freedom in the transmit coil design like an array of transmit coils (Fig 2) each with independent control of amplitude
and phase. Obviously such a system is not mainstream in today’s
commercially available MRI systems. The higher the field strength, the more Tx channels we need to shim
the RF field.RF transmit coil impedance matching
The RF
power amplifier is connected to the transmit coil via a coaxial cable with a
characteristic impedance of typically 50 ohms. To prevent reflections the cable
needs to be terminated with its characteristic impedance, and thus the transmit
coil needs to look like a 50 ohm real impedance. There are many good books on
Impedance matching (2), so I am not going to spend any time on it here.SAR
SAR or
specific absorption ratio is a byproduct of transmitted RF energy dissipated in
human tissue. MRI's radio waves, especially at
very high frequencies cause tissue heating. There are legal limits to how
much energy can be dissipated in the patient’s tissue. The SAR is given by $$SAR=\sigma E^{2}/2\rho$$, in which σ
is the conductivity of the tissue and ρ is the density of the tissue. E is the
E field magnitude. The E field patterns are closely linked to where all the
capacitors in the transmit coil are located. For example, a birdcage resonator
can be high pass with all the capacitors in the endrings or low pass, with all
the capacitors in the rungs. Both coils have the same RF current distribution
and the same RF magnetic field but they have entirely different E field
patterns. Lowest SAR is obtained by
distributing the capacitance needed to resonate the coil. Lowest E fields are
produced by a birdcage that has capacitors in both the rungs and the rings. It
is called a bandpass birdcage.Receiver coil and array design
In fig 3
we see several Rx coil designs for different anatomies. Note that these coils
are all small, and only big enough to cover the anatomy of interest. We can increase SNR over a large
volume coil if we apply an array of small coils that covers the same volume,
but then we must make sure that the noise between these array elements is not
correlated or we will lose our SNR advantage. We can do this with the help of
low input impedance LNAs as described in Roemer’s paper(3). We can increase the number of channels in the array leading to smaller
individual coil elements but beyond a certain diameter the coil noise becomes
dominant again so it is not worth the effort and cost. There are other reasons to go to large arrays, like
acceleration using a method called Sense, but it is beyond the scope of this
lecture.Decoupling
During the
RF transmit pulse, the transmit coil generates a strong RF magnetic field that
gets picked up by the receiver coil(s) and that induces a strong current in the
receiver coil. The induced current destroys the uniformity of the
transmit field, and can cause local component and patient heating. The way to
prevent this is to put a switch in series with the receiver coil. The switch
has to be low impedance or the Q of the receiver coil will suffer. It is
usually done as a parallel resonator (tank) where the inductor is in series
with a PIN diode or a GaN FET.
During the
data acquisition phase the Rx coil is resonant but since it couples to the
large excitation coil, it picks up all the noise from the Tx coil (RFPA and
patient). If
we do not block the noise from the TX coil then the image SNR would suffer. We
therefore put T/R switches in the Tx coil. They can be PIN diodes directly
in the path of the resonator or as part of a tank. Currents in the large Tx
coil as well as the voltages are high, so the components of the T/R switch need
to be able to handle the stress.
The
switches in the Rx coil also see significant current during the Tx pulse. This is a concern, since
this leads to a local temperature rise of the T/R switches, close to the
patient.
Array
coils need to be decoupled in an additional way: They also need “soft” blocking between coil
elements to prevent Array element coupling and coherent noise. Roemer’s paper (3)
describes how to do this with low input impedance LNAs.Ultra low field
Recently
ultra low field portable MRI scanners were introduced with magnetic fields
significantly below 0.1T. At these larmor frequencies of
< 3MHz, the coil noise is dominant. In this case it is extremely important
to increase the Q of the coil to the maximum. We could do the following to
improve Q:
- ·
Increase
the number of turns on the coil to increase the sensitivity. B1/I will
increase, and since the Q is reactance / resistance, the Q will also increase.
There is a limit to this however. Once the number of turns becomes so big that
the individual windings are touching each other, then the proximity effect
kicks in and this will affect the Q. Q is now affected by the losses due to
interactions from the EM fields of one coil winding with its neighbor.
- ·
Reduce
the coil resistance. 1.Theoretically
we could increase the size of the conductor making up the coil windings. The
skin effect pushes all the current to the circumference of the conductor, where
it flows in a surface layer $$$\delta=\sqrt{(2\rho/\omega\mu)}$$$ called the skin
depth, so a larger diameter conductor gives lower loss. In practice this does
not work, because there are size constraints to the Rx coil. 2. Cool
the coil: This will definitely help to cut down on coil loss, but is a huge
engineering issue. We would have to cool the coil with liquid nitrogen, so we either
blow super cooled gas over the coil or attach a dewar. Both are challenges with
respect to patient safety and cost. In many cases the ULF system is also an
economy system, so it is too expensive to go the cooling route. 3. Wind
the coil out of Litz wire (4, 5). If the wire consists of large number of
filaments, each with a diameter smaller than the skin depth, and the filaments
are well insulated from each other, then the RF current density will spread out
over all the filaments, greatly reducing resistivity.
In Fig 4
we see a typical ultra low field (ULF) Head Coil.
If we
succeed in bringing the Q up, then we still need decouple the coil elements
from the Tx coil and from each other in case of an array. The parallel resonant
tank circuits we apply for this should be low loss, the preamp should be low
loss, and lastly if the 3dB bandwidth of the coil is less than what is required
for data acquisition in the presence of a read out gradient then we would have
to broaden that bandwidth without adding loss. An early paper from David Hoult(6)
shows us how to do that by means of capacitive feedback around the LNA.
Acknowledgements
No acknowledgement found.References
1. C
E Hayes et al, An efficient highly homogeneous RF coil for whole body MRI at
1.5T, JMR 63,1985, p622
2. W.N. Caron, Antenna Impedance Matching, American Radio Relay League, 1989, ISBN 0872592200,
9780872592209
3.
Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. Magn Reson Med. 1990 Nov;16(2):192-225.
doi: 10.1002/mrm.1910160203. PMID: 2266841.
4.
B. A. Reese and C. R. Sullivan, “Litz wire in
the MHz range: Modeling and improved designs,” IEEE 18th Workshop on Control
and Modeling for Power Electronics (COMPEL 2017), Stanford, CA, 2017
5.
Y. Otomo, H.
Igarashi, H. Sano and T. Yamada, "Analysis of Litz Wire Losses Using
Homogenization-Based FEM," in IEEE Transactions on Magnetics, vol. 57, no. 8, pp.
1-9, Aug. 2021, Art no. 7402409, doi: 10.1109/TMAG.2021.3081819
6.
D. I. Hoult. Fast recovery, high sensitivity
NMR probe and preamplifier for low frequencies. Review of Scientific
Instruments 50, 193 (1979)