Material and methods
Currently, micro-Tesla NMR research is mostly being conducted by
groups with experience in sensitive SQUID bio-magnetic measurement
since the technique requires specific know-how. However, micro-Tesla
NMR is also challenged by the switching on and off of the magnetic
fields. Most groups who conducted such bio-magnetic measurements
were still using MSRs designed for MEG or magneto-cardiography
(MCG). Recently, people noticed that such an MSR is not suitable for a
micro-Tesla NMR application because of the generation of eddy current
loops along the closed metallic wall. An eddy current along the MSR
wall generates a nT-level magnetic field inside the MSR persisting for
a second or more. This is much stronger than the expected strength of
an NMR signal. Meanwhile, waiting for enough decay of the eddy current
would result in decay of the sample magnetization beforehand. Another
problem with an MSR is the magnetization of its walls. To prevent
these problems, several researchers suggested the introduction of a
compensation coil. Recently, a compensation coil placed inside the
MSR in such a way that its magnetic field could neutralize the magnetic
field on the MSR wall has been suggested, where the compensation coil
could be designed numerically (Hwang et al., 2011, 2012), or analytically
(Nieminen et al., 2011). The compensation coil, if properly designed
and implemented, could significantly reduce eddy currents around the
MSR by neutralizing the magnetic field on the MSR wall generated by
the strong Bp coil. For our research, in addition to the cancelation coil,
we built a specially designed MSR to further reduce the eddy current
problem. The inner-most shell of the aluminum panels of our MSR
is separated into small panels to prevent the generation of an
electrically-closed circuit (Kim et al., 2013). Of course, the outermost
aluminum shell forms a closed surface to play its role as the conventional
RF shield. Between the aluminum shells, Mu-metal layers are placed
to shield from low-frequency magnetic noises. The Mu-metal shells
are designed to be effectively demagnetized wall-by-wall using an orthogonal,
magnetic-flux-circuit scheme (Kim et al., 2013) (Fig. 2a).
We adopted a DC-SQUID (CE2Blue; Supracon AG, Germany), with a
second-order gradiometric pickup coil made of a 125 μm Nb wire with
65 mm diameter and 50 mm baseline, as the NMR signal detector for
the micro-Tesla NMR/MRI system. The second-order gradiometer is
wrapped with island-aluminized Mylar film to reduce RF interference,
and the DC-SQUID is additionally shielded with a superconducting Nb
cast can of 99.9% purity to protect the detector from the strong magnetic
field generated by the Bp coil. The total environmental noise floor of the
system was about 2.2 fT/ ffiffiffiffiffiffi
Hz p at 100 Hz.
In a conventional NMR system, the magnetization and relaxation
characteristics of a sample are decided only by its main external magnetic
field. However, in the micro-Tesla NMR, the main field can be separated
into two sorts of fields, the Bp and the Bm. The micro-Tesla NMR
system is operated in the lower strength field of the Bm (micro-Tesla
range). The Bp is supplied by a coil separate from the Bm coil, and should
be turned off after providing the sample with the magnetization needed
to produce an NMR signal. Fig. 2a shows the coil configuration of our
micro-Tesla MRI system. A 240-turn, copper-wire-wound, solenoid Bp
coil (outer diameter 38 mm, length 62 mm) was used to generate ~52 mT. The homogeneity of the Bm coil was improved by making it a
double Helmholtz coil (Franzen, 1962). The current source was connected
to the Bm coil through two different switchable power resistors
for the K-step (following paragraph) and solid state relays (SSR). The
switchable resistors determined two different magnetic field strengths,
one corresponding to the Larmor frequency of the simulated brainwave
(SBW) and the other for the NMR measurement. A Maxwell-type coil
was used for the Gz gradient. Gx and Gy gradient coils were constructed
with four-paired, rectangular coils. Bipolar power supplies are used as
the current sources for the gradient coils. The currents of all the coils
were controlled by SSRs or mechanical relays. These relays were remotely
switched by a timing board and connected by optical fibers to
prevent interference from outside electronic noises, and the formation
of a ground loop. Bidirectional, transient-voltage-suppressing diodes
and non-inductive resistors were connected in parallel with all coils
for shunting the dark current noise during switch-off. A two-channel,
arbitrary-function generator was used as an AC current source for
each dipole in the phantom. High-pass filters were used to remove the
DC-offset. Since the phases of the applied AC currents at each dipole must be constant during experiments, we precisely controlled the duration
to make an integer multiple of the frequency of the SBW.
Fig. 2b shows a two-dipole phantom for the micro-Tesla BMR experiment.
We made two current dipoles with 0.5 mm copper wires of 9 mm
length. The center of one dipole was placed 22 mm away from that of
the other. The phantom was made of a glass bottle of 27 mm outer diameter
and 73 mm length. In order to generate an ionic volumecurrent
effect, the bottle was filled with 0.8% saline.
Fig. 3 shows the pulse sequences used in this study. Initially the Bp
was applied to form a net sample-magnetization toward the direction
of Bm in the case of the BMR experiment. After Bp was turned off, Bm
and the AC current corresponding to the SBW were applied for the
duration of the tSBW. During this process, AC local currents flowed
through the current dipoles in the phantom and generated AC magnetic
fields. Then the spins around the current dipoles resonated with the AC
magnetic fields of Larmor frequency, corresponding to the Bm, and
began to be tilted with an angular velocity proportional to the strength
of the AC magnetic field. After the tSBW, the Bm was stepped up to a measurement
frequency range and produced free precession decay (FPD) or
echo signals. In the case of the MRI experiment (Fig. 3b), gradient fields
were turned on simultaneously with the step-up of Bm, after the time
tSBW. During this process, the spins precessed about the direction of
the Bm with different frequencies and phases generated by additional
gradient fields. After the time tPW, the polarity of the Gx was reversed.
Then the SQUID measured the echo signal. In the case of the usual MRI
experiment (Fig. 7a), however, the time duration of tSBW was removed.
By this we mean that after the Bp, with direction perpendicular to that
of the Bm, is turned off, the Bm and gradient fields are turned on
simultaneously.
The step-up of Bm is an essential technique for BMR and we call it K-step
(Kim, 2012). There are two major advantages to using K-step. One advantage
is that the brain signal is continuous in the BMR scheme. It is
impossible to control the spontaneous brain signal. Therefore, it is necessary
to separate the NMR/MRI signal from the brain MEG signal of the
same frequency. The main purpose of the K-step is to decouple the FPD
or spin echo signal from the MEG signal by changing the frequency of the detection signal. For example, once a magnetization component
projected into the plane orthogonal to the Bm direction was formed by
the BMR tipping process with a 1-μT Bm, we could alter the detection
frequency arbitrarily by changing the Bm; we could choose to step up
the Bm to 100 μT, to give a signal of about 4.2 kHz. The other advantage
is related to a concomitant gradient field. Maxwell equations indicate
that gradient fields are always related with another gradient field component,
the concomitant gradient field (Myers et al., 2005; Norris and
Hutchison, 1990). If the strength of Bm is comparable with that of the
gradient fields when the spins are mainly aligned to the direction of
the weak Bm, then the relatively high concomitant field influences the
spin motion. Due to this effect, the spin begins rotating along the axis
of the vector sum of the Bm and the concomitant field connected with
the frequency encoding gradient (Gx), and makes an echo signal. For example,
when we were trying to detect a 43 Hz gamma wave, the Bm was
about 1 μT. For the Gx of 0.13 μT/cm and the dimensions of the bottle,
the maximum strength of the concomitant field was about 0.47 μT.
The comparable concomitant field strength messed up the expected
spin dynamics. Therefore, the measurement frequency had to be
stepped up to a much higher frequency to ignore the concomitant
field effect. In our experiment, we used K-step to increase the measurement
frequency to about 1.45 kHz, corresponding to a Bm of 34 μT. The
maximum strength of the concomitant field was only 1.4% of the Bm. A
similar trial for the purpose of reducing the effects of the concomitant
field by changing Bm during phase encoding has been introduced
(Myers et al., 2005). Moreover, there are some other advantages from
using the K-step. The low frequency of the BMR signal and the relatively
wide bandwidth of the proton resonance peak, make it difficult to use a
conventional image sequence for the micro-Tesla MRI. Several steps in
the gradient field strength, due to such a weak external Bm, will touch
zero frequency. This problem can be solved by stepping up the Bm up
to several tenths or hundreds of μT. Also we might arbitrarily choose a
low-noise band as a detection band. Usually, we can detect 1/f noise
in a low frequency range because of flicker-current noise from the coil
system in the micro-Tesla NMR system. Besides, current sources for gradient
fields could be severely contaminated by the power line noise and
its harmonics. We could be free from those particular noise peaks.
During the BMR experiment, the frequency of the applied AC current
at the dipole was 43.33 Hz (gamma brainwave) which corresponds to a
Bm strength of 1.02 μT.