Optically Controlled On-Coil Amplifier with RF Monitoring Feedback

Abstract

Purpose

To develop a new optically controlled on-coil amplifier that facilitates safe use of multi-channel RF transmission in MRI by real-time monitoring of signal phase and amplitude.

Methods

Monitoring was performed with a 4-channel prototype system by sensing, down sampling, digitizing and optically transmitting the RF transmit signal to a remote PC to control the amplifiers. Performance was evaluated with benchtop and 7T MRI experiments.

Results

Monitored amplitude and phase were stable across repetitions and had standard deviations of 0.061 μT and 0.0073 rad respectively. The feedback system allowed inter-channel phase and B₁ amplitude to be adjusted within two iterations. MRI experiments demonstrated the feasibility of this approach to perform safe and accurate multi-channel RF transmission and monitoring at high field.

Conclusion

We demonstrated a 4-channel transceiver system based on optically controlled on-coil amplifiers with RF signal monitoring and feedback control. The approach allows the safe and precise control of RF transmission fields, required to achieve uniform excitation at high field.

Introduction

Multi-channel RF transmit systems are increasingly being explored for their flexibility in controlling excitation profiles and mitigating transmit field (B₁) inhomogeneity associated with high field (1–3). However, clinical implementation of such parallel transmission (pTx) systems has been slow due to tissue heating concerns (indicated by Specific Absorption Rate, SAR), which at high field may become highly inhomogeneous and locally severe (4,5). B₁-related tissue heating patterns are difficult to predict and dependent on geometry and location of the object in the B₁ transmit array and the particular phase and amplitude combination of the various channels (6,7). Finally, SAR can also increase beyond a safe threshold due to hardware failure (8). Therefore, monitoring and accurate control of the transmit signal is necessary to ensure operation within SAR safety limits.

With conventional multi-channel transmission (TX) based on remote voltage-mode RF amplifiers, power monitoring and global SAR supervision can be performed by measuring the forward and reflected complex RF voltages at the individual 50 Ω-matched coil terminals through directional couplers (9–11). Local SAR supervision requires direct amplitude and phase measurement of the current in the individual coils. This information can be obtained through a coil current sensor (12,13) or pick up coil (PUC) coupled to the transmit loop (6,14). Real-time RF monitoring for local SAR supervision as well as for coil coupling and loading compensation has been performed successfully in voltage-mode 50 Ω multi-channel transmit systems (6,14–16). However, due to the load dependence of such systems, this approach can inadvertently result in very high voltages at the coil terminal, leading to poor transmit efficiency or even scan interruptions triggered by the safety system.

In contrast, a TX chain built with on-coil current-source switch-mode amplifiers allows direct (optical) control of B₁ with minimal power loss (17–19). The system has negligible load sensitivity due to RF current-mode operation combined with a current envelope feedback integrated in the amplifier’s printed circuit board (PCB) (18–20). In a previous prototype, this envelope could be optically transmitted to a remote location for monitoring (19). To monitor amplitude and phase of the RF signal, we present an amplifier prototype that can transmit a digitized down-converted version of the coil current to a remote controller while delivering higher power than a previous prototype (20). A 4-channel transceiver array for 7T imaging was built by combining each amplifier with a miniaturized transmit/receive (TX/RX) switch and LNA module. Transmit performance and RF monitoring and feedback capability were successfully demonstrated through benchtop and MRI experiments. This on-coil TX and monitoring setup, together with electromagnetic field predictions, will allow accurate SAR supervision.

Methods

System Design and Construction

On-Coil Amplifier with RF Monitoring

Switch-mode RF power amplifiers have been proposed for MRI to achieve high power efficiency at different field strengths (18,19,21–24). High power efficiency is essential for locating the amplifier on or near the coil (18,22). As in our previous 300 MHz prototype (19), the design of the amplifier was based on Current-Mode Class-D (CMCD) configuration (25). A simplified diagram is shown in Fig. 1. Briefly, it consisted of a push-pull CMCD amplifier (which amplifies the carrier, a ~300 MHz signal) connected to a switch-mode envelope amplifier (a Pulse-Width- Modulation Buck converter switching at 1.7 MHz) with envelope feedback (output current amplitude) (26). This RF envelope feedback was designed to stabilize the amplifier output current for current source operation. For an overview on more traditional feedback methods for RF amplifiers, the reader is referred to (27). To generate shaped RF pulses at the coil, the drain voltage of the CMCD stage was modulated by the output of the envelope amplifier (a signal up to ~ 200 kHz) as similarly performed in the envelope elimination and restoration method presented by Kahn (28). A review of this method can be found in (29). Both amplifiers received optical digital inputs, decoded through a low power digital-analog converter (DAC) in the envelope amplifier. A PUC, integrated in an inner (PCB) layer, was used to sense the current supplied to the transmit coil (19).

Figure 1.

Simplified diagram and photographs of the new on-coil amplifier with RF monitoring electronics connected to the coil through miniaturized TX/RX switch and LNA board.

Here, the carrier amplifier was redesigned to increase output power. The CMCD consisted of two eGaN FETs (EPC8010, EPC Corporation, Inc. USA) in parallel, replacing the single FET used previously. The drain current (and thus B₁ amplitude) is thereby doubled at the expense of doubling the port capacitances. The latter necessitated a redesign of the preamplifier stage. To improve component layout and decrease thermal resistance and parasitic impedances, the new design was implemented using a 8 × 5.3 cm² 6-layer PCB. Amplifier output impedance was tuned to maximize decoupling between element pairs, as described for a 1.5 T prototype (18).

The RF down-conversion and digitization electronics presented in Fig. 1 (blue dash box) is further detailed in Fig. 2. The sensed signal from the pickup loop was fed to a lattice balun and split to two pathways using an LC Wilkinson power splitter. Through 50 Ω attenuators, one signal was fed back to the envelope detector (ADL5511) for current source operation, and the other sent to an active mixer (AD8342) . The mixer intermediate frequency (IF) output signal was down-converted for monitoring. The mixer had a local oscillator (LO) input that can be set tens of kHz apart from the scanner’s nominal resonance frequency (~297.2 MHz). No filter was added to reject noise at the image frequency (f_LO-f_IF). The mixer balanced output (+IF, −IF) was connected to the unbalanced input of the an analog-digital converter (ADC, AD7277) using a differential to single-ended operational amplification stage (op-amp AD8605), which with its limited operational bandwidth (~4 MHz) provided an effective low pass filter (LPF). This single-ended signal was converted to a 16-bit serial digital signal (10-bit resolution data with leading and trailing zeros) using the low power ADC (with serial clock f_sclk=25 MHz), and then optically transmitted to a remote pTx control computer (19,30). An optical receiver was connected to a board (TigerBoard, Nano River Technologies, Denmark) that converted the serial peripheral interface (SPI) signal to universal serial bus protocol (USB2). The chip select (CS) signal for SPI was generated from a clock division on the interface side, resulting in data sampling rate of ~781 kHz (=f_sclk/32) during CS-low.

Figure 2.

Simplified circuit diagram that shows the main components for down-conversion and digitization of the sensed RF signal. The sensed signal is connected to the RF mixer (AD8605) through a lattice balun (L_bC_b network) followed by a Wilkinson power splitter (L_sC_s network) and a resistive tee attenuator (R_A1i network). The power splitter fed part of the signal back to the envelope detector (ADL5511) for current source operation. The differential mixer output is connected to a single-ended input of the ADC (AD7277) through an active balun that consisted of an op-amp stage (AD8605), which was also an effective low-pass filter for the sum frequency (594.2 MHz). The digitized down converted RF was sent to the fiber optical transmitter (AFBR-1624Z).

TX/RX Switch and LNA Module for Transceiver Implementation

The on-coil amplifier was connected to a 6-cm diameter loop through a 4.2 × 4.6 cm² 5-layer double sided PCB that contained a miniaturized TX/RX switch switch and an in-house low-noise amplifier (LNA) (31,32) (Fig. 1). The switch provided a balanced connection from the amplifier output to the coil through a pair of low resistance PIN diodes (MA4P1250, MACOM, USA), and a balanced-to-unbalanced (balun) connection to the LNA through a 50 Ω lattice balun. Additional isolation was achieved by a PIN diode in series with the LNA. During transmission, a PIN diode shorts the balun output to ground and high impedance is seen from the coil into the receiver due to the double parallel resonance formed by the C_c-L_b pair. During signal reception, the balun impedance transformation results in optimum impedance seen from the LNA into the coil (around 50 Ω) and element decoupling is provided by the LNA low input impedance (< 2 Ω).

4-Channel Transceiver Array with RF Monitoring for Head MRI

Four “transceiver” modules (amplifier, TX/RX switch-LNA module, and coil) were assembled onto a half-cylindrical poly-(methyl methacrylate) former with 25.4 cm inner diameter, 0.6 cm thickness and 37 cm length. The total distance from the coil to the cylinder inside was 1.6 cm. All electronics were powered by a +5 V DC line and a higher-voltage V_DD (up to +35 V) DC line through a shielded multi-conductor cable. Signals were transmitted through optical fibers connecting the instrument-room electronics to the in-bore electronics (Supporting Fig. S1). Carrier and envelope signals were optical inputs to each amplifier board. The ADC/DAC serial clock (SCLK), CS, RX unblank (to power off +5 V electronics during reception) and LO signal, synchronous with the scanner clock (10 MHz) and common to all amplifiers, were optically supplied to 1:4 fanout buffers (low output skew, ICS854104, IDT, Inc., USA) before galvanic distribution to each amplifier. Signals were generated by a modular optical multi-channel in-house pTx interface that performs vector modulation taking the small RF and TTL unblank signals from the scanner as inputs; each modulator also received individual in-phase (I) and quadrature (Q) control signals from a PCI board controlled by the pTx control computer (19,30). This hardware was located next to the scanner control in the instrument room. A third optical connection transmitted the monitored RF signal from the amplifier to the controller.

Bench and MRI Measurements

New Amplifier with RF Monitoring

We measured amplifier output power with a calibrated B₁ probe coupled to a tuned (297.2 MHz) 6-cm diameter loop located on a gelled phantom with conductivity and permittivity similar to human brain at 297.2 MHz (33), which loaded the amplifier with R_L=9 Ω. Coil current was calculated from these B₁ measurements and peak power was calculated as I²R_L. Input current and voltage were measured to estimate efficiency. B₁ field was measured with the calibrated probe, while the down converted, sensed signal was simultaneously measured with an oscilloscope probe at the input of the ADC and monitored on the pTx controller. LO was set to 297.0 MHz. Phase and amplitude SD were calculated over one-thousand repetition cycles (> 8 min operation) for all channels for a 10 μT, 1 ms duration sine pulse. Images of an oil phantom were acquired with a multi-slice gradient-echo (GRE) acquisition in a human 7T scanner (Siemens, Erlangen, Germany). Sequence parameters were: echo time/repetition time (TR)=2.2/100 ms, 256 × 176 mm² field-of-view, 2.7 bandwidth-time product, 1 ms windowed (Hamming) sinc RF pulse, 5 mm thick slices, with 10 mm center-to-center slice spacing.

Safety Feedback

In safety monitoring mode, sensed transmit fields can be compared to their targets. The system is automatically switched off when a deviation beyond a set threshold is detected. A safety-monitoring test was performed on the benchtop with the 4-channel transceiver by transiently adding a perturbation to channel 1 through actively tuning a neighboring external loop (Fig. 3). In the detuning circuit, the PIN diode signal (detune) was in sync with the RF unblank and had a pulse width of a fraction of the RF duration. When a change in RF pulse integral exceeded a tolerance, a current-shutdown of all coils was triggered, effectuated by setting IQ control signals of all vector modulators for maximum attenuation (-35 dB). RF pulses used in this test had sinusoidal shape and a 1 ms duration, and were repeated every 100 ms. The B₁ amplitude and phase were set to 18 μT (at the coil center) and 0 degrees respectively, for all channels. For this test, LO signal was set to 50 kHz below resonance frequency.

Figure 3.

Optical RF monitoring setup for the safety test. All channels are optically controlled and monitored while a field perturbation is added through the synchronous tuning of an external loop located in proximity to channel 1.

Performance Feedback

The user interface of the pTx controller also allowed a feedback mode in which amplitude and phase of each transmit channel could be automatically adjusted based on the monitored values. This can be useful to calibrate the IQ control, compensate hardware deviations such as minor phase and gain non-linearity, and compensate for residual coupling among channels. For example, in high frequency CMCD amplifiers, there is a phase modulation with drain voltage due to the drain-source capacitance of the amplifier power FETs, generally corrected by one-time phase pre-compensation (18). As an initial, 2-channel test of automatic phase correction, we did not correct this modulation and used the real-time feedback instead. First, the amplitude channel 4 was changed from 4 to 12 μT in 2 μT increments every 10 seconds, while channel 1 (reference) amplitude was kept at 8 μT. Target phase between channels was set to 0 rad, phase error tolerance was set to 0.035 rad and the system was run without and with the monitoring feedback.

A second test was performed to prove automatic simultaneous amplitude and phase correction for residual channel coupling during transmission. The target amplitude of two channels (2 and 4) was set to 15 μT while channel 4 phase was incremented in π/4 rad steps every 10 seconds. Amplitude and phase tolerance were set to 0.1 μT and 0.035 rad respectively.

Target amplitude and phase values per channel for current iteration i (typed in the user interface) were compared to the amplitude and phase values recovered from the monitored signal (IF). The resulting amplitude and phase errors were compared with pre-established tolerances (also set in the user interface). If a threshold was exceeded, IQ control signals for that channel were updated for the subsequent iteration (i+1) using the equations of the vector modulators (VMs) located on the in-house pTx interface:

$$I_{i+1}=G_{i+1}{V}_0\cos \left({\phi }_{i+1}\right)+0.5V$$

$$Q_{i+1}=G_{i+1}{V}_0\sin \left({\phi }_{i+1}\right)+0.5V$$

where V₀ is given by the VM manufacturer, ${\phi }_{i+1}={\phi }_i-d·\mathrm{\Delta }\phi$ and $G_{\mathrm{i}+1}=G_i-d·\mathrm{\Delta }G$, in which $G_i=\frac{{\mathrm{B}}_{1,\mathrm{i}}}{{\mathit{KV}}_{\text{in}}}$, K is a gain constant, and V_in is the input voltage to the interface. The factor d reduces the applied correction to avoid oscillations. Here, only 80% (d = 0.8) of the measured phase and/or amplitude difference was accounted for when updating the I and Q values.

On-Coil TX-RX and Monitoring

Isolation provided by the TX/RX switch and LNA gain were measured with a network analyzer (NA E5061B 100kHz-3GHz, Keysight, USA). Noise figure of the LNA, placed in a RF shield box and connected to +10V, was measured with a NF analyzer (N8973A, Agilent, USA). Performance of the TR switch and LNA was evaluated during MRI using a single transceiver element. SNR maps were acquired from a GRE acquisition with rectangular RF pulse, T_RF= 4 ms, TR=2 s, TE=3.18 ms, FOV=144 × 108 mm², resolution= 1.1 × 1.1 × 2 mm³. Maps were calculated from noise images using the method detailed in (34). To measure noise caused by the new monitoring electronics, noise images (RF amplitude=0) were acquired with and without the amplifier inside the bore.

4-Channel Transceiver Array and Monitoring

Coil coupling during transmit and receive was measured on the bench with the system loaded with a 9.5 L cylindrical oil phantom. Since the transmitter is not 50 Ω matched, we used the monitoring setup to assess transmit decoupling by measuring the induced field in neighboring elements with one channel transmitting at the time (14). To evaluate receive coil coupling, each receive channel was connected to port 2 of the network analyzer through a bias tee while signal was transmitted from port 1 through a probe coupled to a neighbor loop. S21 values were obtained for all channel combinations. GRE MRI of the oil phantom was acquired with TR/TE=250/5.44 ms, FOV=300 × 300 mm², matrix size 128 × 128, slice thickness 5 mm and 600 Hz bandwidth-per-pixel. Imaging and monitoring were performed for different combinations of B₁ amplitude and phase. To assess the contribution of noise introduced by the amplifier and monitoring electronics in the 4-channel setup, noise correlation matrices were calculated from data obtained within the scanner using a saline load (154 mM NaCl). A first noise measurement was performed with the amplifiers disconnected; and a second measurement with the amplifiers connected to the loops and with DC power turned on.

Results

RF Monitoring and Feedback

The new amplifier prototype connected to the 6-cm diameter loop loaded with the gel-phantom delivered up to 100 W +/− 2.3 W (~3.32 A/52 μT at the coil center, 9 Ω load resistance), twice the power delivered by our previous prototype (19) (Fig. S2) with ~74% power efficiency measured at full power at 297.2 MHz (Fig. S3).

Figure 4 shows the results of the bench demonstration of the monitoring electronics. The down-converted signal (green) accurately tracked the B₁ field measured with the probe (yellow) and was successfully recovered on the remote computer (blue). SNR of the down-converted signal before digitization was better than 40 dB. The modulation on the recovered amplitude for the 200 kHz IF signal (top right) stems from the implemented sampling rate (~781 kHz). The dynamic range and bandwidth of the monitoring system were ~32 dB (limited by selected mixer) and ~250 kHz respectively. This dynamic range is reasonable for the current implementation, but would need improvement for higher power modules. The bandwidth is sufficient for tracking the transmit signal at its maximum frequency offset during slice select. Figure 5 shows 3 of 5 recovered RF spectra for 12 repetitions (superimposed) from the multi-slice GRE acquisition and their corresponding images. The nominal down-converted center frequency was 204 kHz, with 5.4 kHz/slice offset for these 10 mm spaced slices (vertical lines indicate the expected frequency values). The frequency peaks were consistently recovered across repetitions (relative SD below 0.46%) and no image artifact or image degradation was present with the new hardware. For all channels, amplitude and phase SD values were below 0.061 μT and 0.007 rad respectively (Table 1). This indicates stability of the monitored signal for long MRI scans, and is comparable to other RF monitoring setups (14). Figure 6 shows monitored RF pulses for all channels for consecutive TRs before and after the disturbance created by synchronous tuning of the external loop. The RF amplitude of all channels was automatically set to zero for the next TR cycle after perturbation. Data collection, processing and updating amplitude and phase values for one channel took 14.7 ms (SD=2.8 ms over 256 repetitions).

Figure 4.

(a) RF pulse measured at the coil (yellow), down converted pulse at the ADC input located on the amplifier (green) and Fourier transform of this down-converted pulse. (b) Down-converted pulse and its Fourier transform recovered on the pTx controller side after digital conversion, optical transmission and conversion to USB2 protocol.

Figure 5.

Recovered RF spectrum (12 repetitions were superimposed) (top) and corresponding images in the multi-slice GRE acquisition (bottom). Vertical lines show the scanner’s down-converted transmit frequencies. Data for 3 out of 5 acquired slices are shown (the two outer and the center slice). A liquid-filled tube was placed diagonally on top of the coil, located above the box phantom, to show slice position along the scanner z-axis in these axial images.

Table 1.

Monitored amplitude and phase standard deviation (SD) for one-thousand repetitions

	SD (μT)	SD (rad)
Channel 1	0.044	0.0059
Channel 2	0.038	0.0057
Channel 3	0.045	0.0073
Channel 4	0.061	0.0058

Figure 6.

Demonstration of emergency system shutoff based on feedback signal. Displayed are monitored RF transmit signal of each channel at three consecutive repetition cycles (TR=100 ms) before and after a perturbation was induced in channel 1 by synchronous tuning of the external loop.

Results of the system performance feedback experiment are shown in Fig. 7. The phase difference between channels 1 and 4 (Fig. 7a) and the amplitude and phase of channel 2 (Fig. 7b) converged to the target values within two iterations. Without the feedback and phase compensation, phase error was maximum around 1.8 radians for the lower B₁ amplitude (lower drain voltage) and decreased as B₁ amplitude increased. This is in agreement with the phase modulation measured in a previous prototype (18). The maximum amplitude and phase error due to residual coupling, before compensation, were approximately 3% and 0.1 rad respectively.

Figure 7.

Demonstration of real-time phase and amplitude correction based on feedback signals in response to changes in remote channels. (a) Monitored inter-channel phase difference while amplitude of channel 4 was changed in increments of 2 μT every 10 seconds and the reference channel (channel 1) amplitude was kept at 8 μT. (b) Monitored amplitude and phase on channel 2 while the phase of channel 4 was changed in increments of π/4 every 10 seconds.

Transceiver Implementation

Gain and noise figure of the in-house LNAs were 32 dB and 0.72 +/− 0.04 dB at 297.2 MHz respectively. With the oil phantom, SNR exceeded 400 for a 80-degree nominal flip angle. Less than 1% increase in the noise SD was measured and no image artifact was detected with the new amplifier situated inside the bore.

4-channel Transceiver Array

Transmit and receive coil coupling was below 20% and −18 dB respectively for all channels (Fig. 8). The TX/RX switch provided better than −48 dB isolation for all channels. The maximum change in noise correlation values obtained with the amplifiers connected and disconnected from the loops was 0.02 (maximum correlation between neighboring channels was 0.26). This is in agreement with the single channel data, and consistent with the fact that there is no current circulating through the TX electronics when the FETs are OFF (V_DD with I_DD=0, +5 V line is triggered OFF during acquisition) as is in the case of zero carrier signal during a noise scan. An image of the oil phantom was successfully acquired when transmitting and receiving the signal with the 4-channel array in a circularly polarized configuration (Fig. 9). The down-sampled RF transmit signal was recovered successfully for most channels. Spikes in the monitored signal in channel 3 were caused by a one-bit shift error for some samples that was attributed to a timing error on one of the SPI-USB adapter boards.

Figure 8.

TX and RX decoupling. (a) Monitored induced B₁ field in each element relative to channel 1 B₁ field (in logarithmic scale) with one channel transmitting at a time. (b) Normalized S21 measured with receive channel j connected to port 2 of the NA through a bias tee while signal was transmitted from port 1 through a probe coupled to loop i. Worst decoupling values were indicated in both maps.

Figure 9.

MRI of the large cylindrical oil phantom and corresponding RF monitored signals obtained in the 7T scanner. The spurious spikes in channel 3 are due to 1-bit shift in the digitized monitored signal, most likely due to a timing error between clock and data signals.

Discussion

We presented a new on-coil amplifier for safe implementation of pTx at high field. The amplifier delivered twice the power (~100 W) of our previous prototype (19) and allowed accurate monitoring and adjustment of the RF field. Its operation is based on down conversion and digital optical transmission of the RF transmit signal sensed on the amplifier PCB. This was accomplished by redesign of the switch-mode amplifier to include an active mixer, operational amplifier, ADC and optical transmitter within the constraints of small footprint, low power dissipation and cost (< $500 per amplifier for a production of 18 amplifiers). This allows for a compact and direct monitoring approach without addition of dedicated hardware which would otherwise be necessary in multiple stages of the TX chain in a pTx system built with remote amplifiers.

We demonstrated stability of the optical monitored signal through benchtop and MRI experiments. Amplitude and phase information carried by the monitored signal were used for safety monitoring and system corrections. A gain in SNR could be achieved by more stringent low-pass filtering of the IF signal at expense of adding more components to the on-coil amplifier. Rapid system shutdown was demonstrated in the case of unexpected monitoring signals. The current 10–20 ms response time of the monitoring feedback is limited by data acquisition and processing time, which can be improved in future implementations. The system furthermore allows flexible implementation of additional safety features, e.g. introducing a controlled switch or current limiting device in the DC power supply.

Current feedback was previously introduced for remote 50 Ω voltage-mode amplification for active decoupling and load compensation adjustments (6,12,15). Due to on-coil current-source amplification, our setup has minimal sensitivity to load changes (~1.65 % B₁ variation across a 90% change in load) and element coupling. This minimizes the burden on the feedback system to effectuate adjustments, facilitating its use for correction of calibration errors, temperature drifts and residual coupling. These corrections are expected to be small enough to not shift the system from a high efficiency operation as it would be in the case of load dependent and poorly decoupled setups. Large deviations in amplitude and phase values will indicate a system malfunction and will trigger a safety shutdown instead of compensation.

A human brain array built with a row of eight 13-cm diameter overlapped loops driven by this new 100 W amplifier will be able to produce a B₁ amplitude of about 13 μT (~550 Hz) throughout the head, about half of what is currently available with commercial 7 T systems. Therefore, many MRI applications would be feasible with the current design, while further improvements are expected with continued development of high-power FET technology.

In the presented experiments, RF monitoring was performed with amplifiers operating up to half of their maximum power. RF monitoring at full power requires higher attenuation of the sensed signal to protect the RF mixer. A preliminary higher power test required shielding the amplifier to avoid degradation of the monitored signal as a result of high power RF coupling from the transmit elements (coils) to the monitoring electronics of same channel as well as coupling to an adjacent channel. We are currently improving shielding to address this issue, within the constraint of maintaining adequate heat management of the electronics.

We successfully demonstrated that an on-coil receiver setup can be combined with the on-coil transmit amplifier without image noise degradation, yielding SNR and preamplifier decoupling comparable to other receiver arrays (35,36). However, transmit efficiency was (~10%) lower due to pin diode losses in the TX path (data not shown). Beside simplification of the setup for hardware safety evaluation, the on-coil transceiver approach can be considered as a permanent solution for systems with limited bore space. In either case, further design iterations, which may incorporate other semiconductor technologies, will be required to improve efficiency and integration of the transceiver.

In summary, we demonstrated a 4-channel transceiver array built with optically-controlled 100 W on-coil amplifiers with integrated digital optical transmission of the down converted RF TX signal for safety monitoring and feedback. Imaging and RF monitoring was successfully performed in a 7T scanner. While further evaluation of safety and performance is necessary, we consider this new hardware to be important for the implementation of safe, efficient, compact and low-cost multi-channel transmit systems for high field MRI.

Supplementary Material

Supp info

Supporting Figure S1: Optical pTx interface and RF monitoring signal interface (left) and four-channel TX_RX array with RF monitoring (right).

Supporting Figure S2: B₁ field, current and peak output power for the new 297.2 MHz on-coil amplifier compared to a previous prototype (Gudino et al. Magn Reson Med 2016). Peak output power was doubled without compromising efficiency, as indicated by the similar slope in the B₁ and I vs V_DD plots.

Supporting Figure S3: Power efficiency (η= 100% P_OUT/P_DD) vs P_OUT. The input current was sensed with a shunt resistor while output current was measured in the tuned loop with the B₁ calibrated probe.