Scalable, In-Bore Array Receiver Platform for MRI
Jonas Reber1, Josip Marjanovic1, David Otto Brunner1, Andreas Port1, and Klaas Paul Pruessmann1

1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

Synopsis

With the number of RF receive channels cable routing and data handling becomes and increasing problem in particular for demanding applications requiring high acquisition duty cycles and bandwidths. To overcome this we present an MR acquisition platform that is capable of acquiring MR signal in-bore and scales its data handling ability with the number of channel. Furthermore the system provides ample, configurable real-time computational power for advanced in-line data processing and low-latency applications.

Introduction

The number of parallel receive channels of the newest generation of spectrometers and simultaneously of receive array coils has drastically increased for the benefit of better SNR yield, higher parallel imaging acceleration and leaner clinical scanning workflow. Furthermore, additional sensors beside RF detectors, such as magnetic field probes, motion tracking systems etc. are increasingly integrated into the bore and coils and require signal acquisition and data transmission out of the magnetic field. On the downside of these trends cable routing has become an increasingly challenging task for coil engineering and handling. In-field acquisition systems [1,2] offer here the advantage to multiplex large numbers of receive channels onto fiber optic connections which are lightweight and RF safe. However, digitization of RF signals in-bore with very high dynamic range and highly different signal levels are found to be difficult especially for ultra-high fields. On top of that the data flow is ever increasing by sequences requiring very high read-out bandwidth and duty cycles. Therefore high channel counts are often inflicted with data flow and handling problems. Hence spectrometer platforms targeted for very high channel counts must scale similarly the provided data handling and real-time computing power. For pushing the limits of MR acquisition speed and SNR there is therefore a need for data acquisition and real-time high-speed processing systems with scalable channel counts and corresponding data handling capabilities operating in-bore but delivering state-of-the art signal quality also at high fields. In this work we present a novel scalable and modular spectrometer hardware platform with configurable, high-dynamic range in-bore receivers offering vast, programmable real-time computing power.

Methods

The basic building block (Fig. 1) of the host unit consist of a National Instruments (Austin, Texas) FlexRIO® Kintex® 7 PXIe platform used for software control, mass storage and commodity IO tasks. These units are as such parallelizable via custom PXIe links such that the whole platform of the spectrometer can be simply stacked. To each FlexRIO® module (of up to 8 can be hosted in a chassis) a custom built data collection and processing unit based on a Zynq®-7000 SoC (Xilinx, San Jose, USA) is connected. This unit connects via 4x7Gbs + 4x1Gbs run-time configurable and re-routable high-speed optical communication lanes and precision clock reference signals to programmable, low-power FPGA units which are MR compatible. Up to 4 of the latter units can be connected to a single Zynq unit and operated fully in parallel. The in-bore units can be equipped with up to 16 broadband in-field receive channels [1]. The 14 bit/125MSps ADC channels can be equipped with high-dynamic range RF frontends on miniature mezzanine board modules. . For initial demonstration MR signals have been acquired from a standard 8-channel head coil (Philips, Best, Netherlands) using clinical sequences. The receiver was operated in-bore and synchronized to the clinical MR console by a TTL trigger and by parsing the RF excitation pulses for phase alignment of the acquired profiles.

Results

The novel RF front-ends of the in-bore units can be modularly equipped for 1.5-11.7T. They provide a variable gain from -15 to 45 dB. The ADC provides a SNR dynamic range of 94.6 dB at 1 MHz bandwidth, an SFDR of 96.9 dB and can acquire a bandwidth of 1 MHz with 100% duty cycle. The total clock jitter of the digitizer was 540 fs. 1 W per channel are required in full operation. Images shown in Figs. 3 and 4 show no visible artefacts attributed to signal digitization or spurious signals form the in-field circuitry.

Discussion

The presented RF receiver offers an analogue signal performance that are on par with out-field digitizer units in compact, low-power, MR compatible modules. The broadband direct-undersampling architecture in conjunction with the modular analogue front-end allows for fast adaptation to multiple frequencies and signal levels dependent on the application. In conjunction with the freely programmable digital system the modules form a MR compatible high-performance platform for various applications. The data routing and collection scheme offers vast, configurable real-time computing power and can in principle handle, process and store data from arbitrary amount of channels at full rate, dynamic range and duty cycle. These real-time computing capabilities will be essential for enabling very large channel counts for keeping data flow, storage and reconstruction tractable by customized, in-line data preprocessing.

Acknowledgements

NanoTera initiative, Wearable MRI project.

References

1) Reber J et al. In-Bore Broadband Array Receivers with Optical Transmission, Proc ISMRM 619 2015. 2) Tang W et al, A HOME-BUILT DIGITAL OPTICAL MRI CONSOLE USING HIGH-SPEED SERIAL LINKS,Magn Reson Med 2015

Figures

Figure 1: System architecture: Parallelizable PXIe chassis are connected via high speed busses to up to 8 custom data collection and processing modules. Each of these modules can host up to 4 in-field FPGA modules with a high-speed optical connections (3-12GBs). Again each of these in-field boards can carry two 8 channel ADC boards.

Figure 2: 8 channel ADC Board (14 bits, 125 MSps). The analogue input stages are hosted on mezzanine boards and can be application specifically equipped covering thereby a large range of applications. The shown examples are RF gain stages for 3T comprising amplification (up to 49 dB), programmable gain scaling and anti-alias filtering. Two of such ADC boards can be connected to an infield FPGA board.

Figure 3: Phantom images acquired with a standard clinical 3T 8 channel coil connected to one receiver module (left). The log plot of the same image shows a clean signal background and low artifact levels despite the comparably strong signal power provided by this scan.

Figure 4: Standard MPRAGE image (1 mm inplane, 1.5 mm slice thickness) acquired with an 8 channel standard head coil at 3T connected to the new receiver system.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2170