Dimitri Welting1, Edwin Versteeg1, Ingmar Voogt2, Joost van Straalen3, Martijn Heintges3, Marco Rietveld3, Jeroen C.W. Siero1,4, and Dennis W.J. Klomp1
1Radiology, University Medical Center Utrecht, Utrecht, Netherlands, 2Wavetronica, Utrecht, Netherlands, 3Prodive Technologies, Son, Netherlands, 4Spinoza Centre for Neuroimaging Amsterdam, Amsterdam, Netherlands
Synopsis
Here we propose a setup to boost spatiotemporal encoding for fMRI in the human brain
using high dense receiver arrays (72ch) and fast gradients (6000T/m/s). Moreover, we demonstrate that the operation frequency of a
high-end gradient amplifier can be increased to ultrasonic frequencies so to
avoid unpleasant acoustic noise. By using this amplifier with 8 transceivers,
64 receivers and a 2-axis cooled gradient coil, a light weight setup is
constructed for operation in a 7 Tesla MRI system for fMRI
experiments in the human brain.
Introduction
Layer
and columnar fMRI is within reach at ultra-high fields and can be used to record
brain function at the mesoscopic (<1mm) spatial scale1,2. Such high-resolution requires substantial spatial
encoding power (<0.5mm isotropic) at subsecond sampling times while
maximizing (temporal) SNR. Here we propose a new hardware setup for 7T that
maximizes both RF encoding (72 channels) as well as gradient encoding (next to
the X,Y,Z body gradients, 2 extra gradient axis to drive at 5000T/m/s as
spirals without causing PNS) to boost spatiotemporal encoding. Moreover, the
setup is equipped with an 8 channel dipole transmit coil to mitigate flip angle
non-uniformities. To facilitate the high-power gradient drive, we adapted the firmware
and end-filter of a high power gradient amplifier to drive the gradients at 21kHz
and avoid unpleasant sound perception that generally coincides with increased
gradient encoding. Pending IRB approval and reconstruction framework, we show
raw MRI data obtained from a phantom at 7T using the entire hardware setup.Methods
The gradient amplifier used is an NG500 1.1 (Prodrive Technologies, Son,
Netherlands) as shown in Figure 1a. The FPGA firmware of the gradient amplifier
has been modified to allow for a higher frequency range (bandwidth) output. The
new firmware modifies the output driver to allow the output current to be a
harmonic of the base frequency, increasing the frequency of the gradient
amplifier output current up to 32 kHz. This enables the output to be in a forward
drive mode, i.e. without any feedback loop. The duty cycle of the pulse-width
modulation is controlled by the voltage level at an analog auxiliary input. The
internal end-stage filter was removed and brought to the gradient coil as a
resonant circuit. The high power gradient waveform is compared to an 18kW audio
amplifier we reported on previously3 (Powersoft k20, Italy).
The homemade insert gradient has been made with hollow copper windings for
a Y- (or X) and a Z-direction magnetic gradient field, creating a 2-axis
gradient setup (Fig 2b). Minimal use of epoxy for mechanical support was needed
due to the high driving frequency of the insert at 21kHz in combination with the
relatively high inertia of the coil. Therefore, the entire setup weighs less than
10kg. Capacitors have been added to the windings to filter unwanted amplifier
current noise while enhancing gradient field efficiency by enabling a resonant
mode at 21kHz. To allow for high duty cycles and long scan sessions, cooling
has been included. A water solution can be pumped through the windings to allow
for effective heat dissipation.
To minimize eddy currents, no RF shield was used between the insert
gradient and the RF coils. Eight 7T 1H dipoles transceivers, matched
to 50 ohm at 298MHz, are integrated into the design and tuned in the presence
of the gradient (Fig 1c). Furthermore, a close-fitting 64 channel 7T 1H
head coil array (Wavetronica, the Netherlands) was used as a local receive
array (Fig 1d).
To measure the performance of the ultrasonic gradient setup, a field camera
(Skope, Swiss) has been placed inside the insert gradient.Results
The output voltage and current as monitored inside the gradient amplifier is
shown in Figure 2. At the start and end of the gradient pulse, the
dampening effect of the resonant insert gradient is visible. While the gradient
amplifier can provide more power to the coil than the previously used audio
amplifier, the field cameras show similar temporal behavior and quality when
driven with either setup (Figure 3).
With the gradient amplifier driven at less
than 30% of its maximum, it already provides a 6000T/m/s slew rate at 50mT/m
gradient strength (21kHz, Fig 3).
With a phantom inside and using the 8 channel
transmit coil for spin excitation, signals from the 72 receive channels were
recorded by the MRI system during the 21kHz readout, while using the remaining
three whole-body gradients to control the gradients for additional spatial
encoding. Figure 4 shows a clear 21kHz modulation on top of the echo in k-space.Discussion
While a fixed resonance at 21kHz is pleasant for sound perception and
simplifies cable management, it coincides with limitations in flexibility to
setup the MRI experiment due to the fixed encoding speed. Furthermore, no
standard MR-sequences exist for 5 gradient axis. The 72 simultaneous channels
and high sampling rate needed to capture the 21kHz spirals lead to a relatively
high data rate that can complicate standard MRI data pipelines. Setting up the
reconstruction line, converging the setup in a more user-friendly appearance
and performing safety tests is currently in progress.Conclusion
We successfully increased the operation frequency of a high-end gradient
amplifier by altering the firmware and changing the gradient amplifier output
filter to a resonant circuit. Moreover, a low-weight setup was constructed that
houses 8 transceivers, 64 receivers, 2-axis cooled gradient coil for operation
in a 7T MRI system for high-resolution fMRI experiments in the human brain,
even providing access for visual stimulation. The operation of the setup was confirmed
using field cameras and MR signal detections from 72 receive channels that allow
future phantom and safety studies to explore the potential of the setup in
maximizing spatiotemporal resolution.Acknowledgements
No acknowledgement found.References
1.
Dumoulin, S. O. et al. Neuroimage 168, 345–357 (2018)
2.
Huber, L. et al. Neuron 96, 1253-1263.e7 (2017)
3.
Versteeg, E. et al. in Proceedings of the 27th Annual Meeting of
ISMRM #4586 (2019)