CA2967956A1 - Compact electronics system for portable mri - Google Patents

Abstract

A compact software defined radio system is presented that has been designed for portable MRI applications. While traditional MRI systems deliver excellent image quality, they tend to be heavy and hungry for electrical energy. Our previous efforts have made great strides in the design of the actual magnet, to such an extent that the associated electronics system (including transmit power amplifiers, analog to digital converter, digital to analog converter e.t.c.) outweigh the actual magnet. We therefore decided to work on the electronics system, presenting an updated electronics system, which compactly bolts onto the current magnet, adding minimal weight to the MRI system.

Description

Compact electronics system for portable MRI -',.
_...
....
õ It 1 .
, , ' v - ::::::: ,-, 5 ,., , * I 4.=
1141*\
4 4....¨

",.., #
== , f , 'I 0 'tI * i ', 4 'i1,t i = 4I.
4 . ,O = t.' t tl. 4, t 4 ' . : 4/fr, \ ..., = oi.

Logi Vidarsson LT Imaging Inc.

Field of the invention This invention relates to software defined radio (SDR) systems, specifically as they apply to magnetic resonance imaging (MRI), often referred to a "spectrometer" or a "digital-spectrometer" in the context of MRI.
Abstract A compact software defined radio system is presented that has been designed for portable MRI
applications. While traditional MRI systems deliver excellent image quality, they tend to be heavy and hungry for electrical energy. Our previous efforts have made great strides in the design of the actual magnet, to such an extent that the associated electronics system (including transmit power amplifiers, analog to digital converter, digital to analog converter e.t.c.) outweigh the actual magnet. We therefore decided to work on the electronics system, presenting an updated electronics system, which compactly bolts onto the current magnet, adding minimal weight to the MRI system.
Background of the invention The role of electronics in MRI is to both to generate a time-varying magnetic field of sufficient strength that acceptable magnetic resonance can be observed, as well as receive and digitize the electrical signal from said magnetic resonance. Simply put, our electronics system needs to transmit at a sufficient power to generate magnetic resonance, as well as being sensitive enough to detect the small electrical signal from the sample. Previous experience with a traditional system shown in Figure 1, showed a need for approximately lOW of transmit power per channel, and ability to measure a small -100 dBm signal.
The system had four transmit channels, the required amplifiers were relatively large, and once housed inside a tower PC case, that PC case became both the largest and the heaviest component of the overall MRI system. While our system is compact enough to be "cart-based", in order to realize the full potential of portable MRI, it is desirable to engineer a system that can be carried by hand. We therefore directed our attention to the electronics system with the aim of further reducing its size and power consumption.
-p ckage Magnet WI' =
4!4. = . 4.4 too Figure I. Previous MRI system showing the three main components, a laptop for image reconstruction and signal analysis, electronics kit that converts from digital to analog, and the magnet (including tuned four transmit and receive coils).
Shown in Figure 2 are the sub-components of our traditional system and how they connect with the rest of the MRI system. Of the components inside the electronics case, the four transmit amplifiers are easily the largest component. Not only are the amplifiers physically large, but they must be mounted on a large heat-sink and provided with abundant air-flow. Furthermore, it is the largest consumer of electrical power within the electronics kit, consuming about 240 W, out of a total of 250 W. Therefore, we elected to begin our efforts on the transmit system.
USB
4ch SDR 2.5MHz LP filter (4x) TX amp (4x) TX connector(4x) TX _f TX(4x) RX(4x) _________________________ !Quad LNA RX connector (4x) RX coil w LNA
(4x) ===
= ¨ - ¨ - ¨ ¨ ¨ = ¨ = ¨ ¨ ¨ ¨ = ¨ = ¨ - ¨ =
¨ =
-Figure 2. Block diagram of our traditional electronics kit (within dotted border) and how it connects to the other major components magnet and PC.
As it turns out, the power requirements on the transmit side arise from the desire to generate a small time-varying magnetic field on the order of a few micro-tesla. In the context of MRI this transmit magnetic field is often referred to as the B1 field. The current through our transmit coils generates this field, and a lOW power requirement on a 50 Ohm transmit coil, translates into roughly 500 mA of current requirement through the transmit coils.
We present a direct to digital transmit system, where we take advantage of a known issue with pulse-width modulation to modulate our signal to RF frequencies. The resulting transmit side consume far less electrical power (approximately 2.5 W/ch for a total of 10 W for our four channels), generates similar 81 field, and most importantly eliminates the need for a transmit amplifier.
The resulting electronics kit is compact enough to mount directly onto the magnet, eliminating the electronics case altogether.
Description A block diagram of our new electronics system is shown in Figure 3. The digital transmit coil greatly simplifies the electronics system, in fact a direct digital connection is made straight from the FPGA into the transmit coil, this connection is configured to operate at speeds of up to 300 MBits/second. To further improve operational bandwidth we also elected to replace the older Teensy 3.2 microcontroller with a high speed USB controller, and lastly by using the AD9271 designed for medical ultrasound, we not only gain a very capable analog to digital converter (ADC), but it also comes with up to 40 dB gain, something that allows us to eliminate the quad LNA shown in the older system.

PC ____ USB
= 1 04.!
. High speed I 'f t4.=
FPGA __________________________ , I USB controller I
Octal ADC + AFE DC Bias (8x) ' , Figure 3. Block diagram of the new electronics kit. The direct to digital transmit setup results in a substantial reduction in both size and power consumption of the transmit system. Furthermore we have upgraded the ADC which now includes an analog front end originally intended for ultrasound applications but is as it turns out very well suited for our MRI applications. A DC bias circuit is provided to supply 5V power to the low noise amplifier present on the receive coil.
We begin by discussing the design of the new software defined radio (SDR) board, then move on to the digital transmit coil, with a discussion on how we use the FPGA to modulate signals at the desired frequency and phase. Lastly we show experimental results from a spin echo experiment, illustrating both our ability to generate a waveform with desired transmit characteristics, as well as generating and detecting a spin echo signal.
Software defined radio (SDR) Our software defined radio serves as a bridge between the analog world of our RE coils and the digital world of the PC. We seek flexibility to carry out a variety of experiments requiring multiple transmit and receive channels. Power is provided to the receive coils via a bias-T
circuitry, this allows for a modest 30 dB gain onboard our receive coils, and greatly reduces the influence of "cable-noise" on the overall SNR
performance of the system.
USB, from PC ¨ Cypress FX2 High speed 44: 4109 USB controller , = . :*; 4 TX 1 I == . ,, , iotit drt- flr."
¨= ' ortp.4) , Irper4,-4111"6õ.
Platform FLASH __ Spartan 6 FPGA
= $ ' . .
. .
' +, 144eiligoa 1 4 Clock ______ I AD 9271 TX 8 t renoto"ii- ;
3 Y:tii4' r =
= 440:2 Figure 4. A diagram of our SDR board (left) along with a photograph (right).
At this point in time we did not provide a turn-off control signal which traditional systems often employ.
While we could certainly re-route accessible signals on the FPGA to accommodate a request for such a turn-off signal, at this stage we wish to listen to the transmit signal as it is very helpful. In fact it provides much needed feed-back when we tune the four receive channels to work together as synchronously. To prevent damage, we do add a cross-diode circuit to the receive coils, often referred to as "passive-blocking" protection in the context of MRI. There are many excellent references available that discuss receive coil design in detail'.
Digital transmit coil , =
= ,=-.1 =,¨, ,, , ... ..
==.=., .,,,,, : .=
t 4 i ' "
'.;.1. ;,=:;..:.:;:;...,:,,..,..'''.....:, 0 =',. . iµe.
'..'-',:::....,... -.. =,....,...#= ,õ, ,..
., ,,,,fri.:õ=,,,...,õ = =::#:, -= , ...4.
..
i w_ - ...,=,,,..=,.1 it . , -.4.,.. e ..,õ.....õ...õ,....õ,. , , .
., ....
.,....,, ......, ic .,,....: ,, =
, -Figure 5. Photograph of a 4 transmit coil array for RF encoding experiments.
The four square loop coils are oriented on the front/back and left/right sides of the magnet. On the right we see a close-up of the circuit board that not only serves as our digital to analog converter, but also the on-coil transmit amplifier. We refer to this board as the feedboard.
Figure 5 shows a four channel transmit coil, one embodiment of the invention.
We have tested the coil for our portable MR1 system and as we will see later, it does provide sufficient B1 transmit field for our magnetic resonance experiments. A schematic of the transmit coil is shown in Figure 6, at the moment we use a small current limiter resistor, about 1 Ohm to limit the DC current flowing into the coil, as we observed a very poor reliability of our transistors without it. Even as our system matures, we may elect to keep the resistor, as it provides a near-fail-proof way to limit the current in the coil, a useful property to ensure safe operation of the system under many conditions, ensuring our transmit system stays within regulatory limits on RF emission. .
I ------------------- -1 I-- Coil winding L -- C011 winding , I Choke R - current limiter I

DC In CT C_DC
9- Coil resst iance R - Coll resistance RF digital 7)( RF dIgital TX
I
, ------------------Feed board Figure 6. A schematic of one of our transmit coils (left) and an AC
equivalence circuit (right). Currently we use a large capacitor for C DC (470 uF) and a large inductor for L_Choke (33 uH), this results in a series RLC circuit formed by the coil and the tuning 1 See for example Roemer et al. MRM 1990 capacitor CT. Like traditional RF coils, it is necessary to carefully choose the tuning capacitor so that resonance is achieved, in this case minimum impedance, to maximize the AC current, resulting in maximum B1 transmit field.
We operate the transmit coil at 5V DC, in order to generate a 500 mA current, we must have overall resistance that is less than 10 Ohms. Looking at the schematic in Figure 6, we have three resistances in addition to the transistor. Our current limiter (1 Ohm), the internal resistance of our Choke (0.03 Ohms), and the coil resistance (approx. 0.2 Ohms). The CMOS transistor itself has approximately 0.5 Ohm resistance when fully on. This adds up to 1.73 Ohms. Thus, if everything were fully-on, we would have

2.9 A of current flowing in our transmit coil. As we shall see later, to modulate the transmit signal to the desired baseband frequency, we can at most have the transistor on half the time, hence our AC-current at 2.9/2 = 1.45 A easily meets the 0.5 mA requirement.
One important difference compared to traditional transmit coil is the AC
equivalence circuit impedance.
Traditionally owning to availability of off-the-shelf RF-amplifiers, impedance of transmit coils is set to 50 Ohms. In our case, we do not use off-the-shelf amplifiers, and are thus free to choose our own impedance. As expected, minimum impedance results in maximum current and hence maximum B1 field. This is achieved when we choose CT such that 2n-f = 1/V CT L ¨ coil winding , where f is the operating frequency of the system. There may be another benefit as the electric field generated might be less with smaller impedances and hence smaller voltages. While this will require further study, it should be said that Synthetic absorption rate (SAR) is often a problematic limiter when it comes to MRI
system operation. A low [-field transmit system, would naturally have less SAR, since SAR goes like the E-field squared.
As we shall see, harmonics prove to be fundamental to our system. However, by carefully directing these harmonics, we can take advantage of them, and in fact we use harmonics to modulate our signal to the desired baseband operating frequency. The transmit coil is either on or off, there is no intermediate step. To synthesize waveforms with values other than full on or full off, we resort to a scheme known as pulse-width modulation (PWM). There are many excellent references available on PWM2, but the principle is to generate values between full on and full off by quickly switching between the two, analog filtering is then applied to filter out unwanted harmonics. As we will see later, we use this to not only generate arbitrary waveform amplitudes but also modulate the signal to the desired base-band frequency.
Transmit modulation scheme We employ a modified PWM scheme to both synthesize and modulate our waveforms.
A typical PWM
scheme used in a microcontroller3 is shown in Figure 7. Large values represent more proportion of the time spent in the full on state, while smaller values are mostly off.
Typically, microcontrollers implement PWM via counter logic. The user specifies the length of the cycle, we then synthesize values between full off and full on (e.g. 0 and 1), by leaving the device on for the appropriate amount of time. For example, to represent 0.1, we would leave the device on for 10 % of the time for a given cycle. In the 2 One example out of many is Wikipedia (https://en.wikipedia.org/wiki/Pulse-width modulation), retrieved May 23, 2017 http://www.nxp.com/assets/documents/data/en/reference-manuals/K20P64M72SF1RM.pdf, retrieved May 23, microcontroller case, we usually "front-fill" e.g. the device is on for the first 10% of the time in each cycle.
0.1 0.2 0.3 0.4 0.5 0.6 1 _______________________________ 2500 = =

0.5 1500 = 0.2 500 o ____ ¨ ¨

Figure 7. PWM example (left) along with the corresponding spectra (right). In this example, we picked the "counter-period" to be 1, and we then increment the desired output from zero (far left) linearly to 60 % of full output. As expected this generates several harmonics in the spectra (right). Traditionally low-pass filtering would now be applied in order to isolate the low frequency components. We however are more interested in the first harmonic, as if we choose the appropriate PWM cycle, we can place that harmonic right at our MRI operating frequency, in the case of this example I (red arrow). Our filtering then seeks to remove higher harmonics that may cause unwanted RF emissions.
Shown in Figure 7 is a Fourier transform of the PWM waveform, note the "harmonic" seen at the cycle frequency (in our case 1 frequency unit). As we vary the length of the cycle, the frequency of this harmonic moves, ideally to coincide with the center frequency of our MRI
system. Typically, the user is most interested in the low-frequency content, in which case a low-pass filter is used, often just a simple capacitor from the output to ground. We however, are much more interested in the first harmonic at the cycle frequency, hence our coil needs to filter out other harmonics. As we recall from Figure 6, on the AC equivalence circuit, the tuning capacitor and the inherent inductor in the transmit coil, do just that, providing both D/A conversion and modulation in one compact circuit.
It should be mentioned that our transmit coil circuit shown in Figure 6 does not filter out the DC
component of the PWM waveform. At first this seems concerning, as traditional MRI systems employ a highly homogeneous magnetic field, specifications calling for a very challenging 10 parts per million (ppm) magnetic field homogeneity are commonplace. Hence even a modest 10 ltT
field generated incidentally from our transmit coil could place a strain on a 10 ppm homogeneity specification for a 1.5 T
system. However, as described elsewhere' our systems have much less stringent homogeneity requirements, in fact we employ a gradient magnetic field having approximately 490 G field at the bottom of the bore, and rising to about 500G at the top. Since our potential 10 pT DC incidental field is small in comparison to this gradient, or approximately 0.1 G, compared to a 10 G gradient, its effect on the main field is very limited. Furthermore, it is oriented perpendicular to the main magnetic field and its impact is thus further reduced via geometry. It should though be noted that this could result in mild imaging artifacts owning to a location depended frequency shift. We anticipate being able to counter See Vidarsson et al. US Patent: 14/644,725 this effect with careful post-processing of the resulting images, akin to phase correction methods used in fast spin echo sequences on traditional MRI systems.
Phase control of modulation It is not sufficient to provide just modulation of the signal, we must also control the phase of the modulation. Therefore, we must improve the PWM scheme of Figure 7. Fourier analysis offers a hint, the shift theorem states that by shifting the signal in time, we impart phase in the frequency domain.
This principle was used to design our phase control scheme. While modulation can be thought of in "polar coordinates" with frequency (e.g. magnitude) and phase control, we found it easier to view it in terms of complex modulation, where our waveform has both real and imaginary values corresponding to both the value to be modulated, as well as the desired phase of the modulation.
Positive Re Positive prt Negative Re Negative trn 1 ===¨= _____ %.
cos e =
-sin os / X /
X /
/
/ /

/ /
/
/
-cos 8 X
sine / X====

Figure 8. Dividing each period into four zones, we can affect phase control of the modulation by directing our energy appropriately. E.g. instead of simply "front-filling" the entire cycle as in Figure 7, we divide the cycle into four zones corresponding to positive real, positive imaginary, negative real and negative imaginary.
As shown in Figure 8, we can divide one cycle of the PWM scheme into four zones: Positive real, positive imaginary, negative real, and negative imaginary. By imparting energy (e.g.
switch to full on) we imprint the proper amount of real and imaginary modulation. Some examples of our scheme are shown in Figure 9.
=

CA 2967956 2017¨.05-24 - _________________________ 0.6 =
0.4 0 ___ Figure 9. Modulation examples, the modulation of the four numbers is shown above with 1, -1, 21, -31 modulated onto a frequency of period 1.
It should be noted that our modulation scheme limits us to sending current down the transmit coil half the time, hence the maximum AC current will be roughly half of what we could expect from DC analysis of our transmit coil. This limit arises from the fact that both the real and imaginary values will be either positive or negative, not both, hence we will either impart energy in the first half of the cycle or the latter half.
Experimental verification We begin by using our receive system to observe the transmitted waveform, to verify that it we can both digitize and modulate the waveform with our system. Next we perform a spin echo experiment, known as Carr-Purcell-Meiboom-Gill (CPMG) experiment in the context of MRI, to demonstrate that we observe magnetic resonance in a mineral oil phantom.
Figure 10 shows both the transmitted and the received baseband waveform on our SDR, along with their spectra. There is very reaosonable resemblance between the transmitted (or desired) waveform and the one we receive. We do observe a slight dent at high real/imaginary values, resulting in a small harmonic seen in the spectra at 2.095 MHz. We have observed similar behaviour on traditional systems, when high power is required from the transmit system.
.4111 ==== ===÷' 0.2! 4 54.411 44.111 22.411 I
.5.4541 . ..=
Sbol le.11!
, 4.429 04299 03 424235 24.424 26.4245 26.429 264296 207 2.075 226 2.065 2.09 2 095 2.1 Fr06.42157 5950 Figure 10. Transmitted waveform modulated to 2.08 MHz (shown at baseband) --left, along with the received baseband waveform (middle). Also shown is the spectra of the two waveforms (right). On the time domain waveforms, real part is shown in blue, while imaginary is in green. On the spectra, the received waveform is shown in blue, while the transmitted waveform is in green. The similarities are quite strong, suggesting good performance of our transmit system. There is a small harmonic seen in the received waveform (at 2.095 MHz), which corresponds to the slight flattening seen at high real/imaginary values.
The results from the CPMG experiment are shown in Figure 11 showing the potential of our electronics kit for magnetic resonance work. Two things of note: There is a sizable difference between the transmitted signal power (at about 109) and the received signal power (at about 105), and there is a slight phase offset on the four refocusing pulses compared to the excitation pulse at the front. We will discuss these two in series.
le.09 ____ 86+08 =
66+08 46+08 26+08 -26+08 0 0.005 0.01 0.015 0.02 0.012 0.013 0,014 0.015 Figure 11. Experimental results playing a four echo fast spin echo experiment (CPMG style) on our electronics kit. Overview (left) and zoom (right) onto the third echo, showing a very nice spin echo. The frequency was set to 2.11 MHz, corresponding to a proton resonance frequency of 496G.
The large difference between transmitted signal power and received signal power is fundamental to MRI. There are a variety of reasons for this, discussed in many textbooks', but the recorded power difference of 4 orders of magnitude (e.g. 80 dB) is typical. This does place a strain on our SDR system, in fact we must be able digitize vastly different values. While our ADC only digitizes at 12 bits (allowing for approximately 4096 discrete values), it does so at 25 MHz, hence in the presence of measurement noise, the filtering that occurs in the FPGA, greatly increases the 12 bits of "raw"
ADC bit-resolution to more than 50 bits, ample for our needs. This is known as "processing gain" in the context of digital receivers'.
We employ a floating-point representation of the 50 bit+ data, as it is more efficient for data transport to the PC.
There is a slight phase offset in the refocusing pulses compared to the phase in the excitation pulse in Figure 11. Looking at the 5 high-power transmit pulses, the first pulse in the CPMG series is known as the excitation pulse and serves to create the magnetic resonance signal, the following four pulses are known as refocusing pulses and serve to refine and align the signal so that it peak's mid-way between two successive refocusing pulses. Ideally in a CPMG experiment, the refocusing pulses are played with a 90 degree phase offset relative to the excitation pulse, or in our case, purely on the imaginary channel, while the excitation pulse is played on the real channel. We almost achieve that, as our refocusing See for example "Magnetic Resonance Imaging: Physical Principles and Sequence Design" by Haacke et al, 1999 6 See for example "Global positioning system" By Misra and Enge, 2002, as the GPS receiver's rely on processing gain in a fairly extensive manner.

pulses do have a small real component, presumably owning to phase distortion at high transmit power levels.
It should be noted that the results above are obtained using only one transmit channel, while we do have four installed, and the SDR has allowances for up to eight. That said, we anticipate our system to have some interesting properties when it comes to coupling of the different transmit channels. In multi-channel systems coupling needs to be carefully managed. If not, on the receive side, there will be an unnecessary loss of SNR. On the transmit side, the pain is felt via excessive RF power, as opposed to SNR.
When in the off state, our transmit coils present a very high parallel impedance to its neighboring coils.
Hence when off, our coils are invisible to them. When on, our coils present a very low impedance and hence we observe coupling currents in neighboring coils, if they are on at the exact same time. Therein lies the key, if we only transmit with one coil at a time, coupling currents will be much less of an issue.
Under the modulation scheme discussed above in Figure 8, we will, at most, spend half the time in the "on-state" for any given coil. In practice, we have found significant distortion on our system when we push it to transmit close to its maximum rated values, this is very similar to what is seen on traditional RF amplifiers when they are pushed close to their limits. For this reason we rarely operate RF transmit systems close to their operational maximum values. The results shown above in Figure 11, are obtained operating the transmit system at 40 % of its maximum, corresponding to spending only 20 % of the time in the "on-state" for our lone transmit coil. Even at 40%, we note a slight phase distortion, as seen with the refocusing pulses in Figure 11. Hence in the futbre, we may decrease that to 20%, resulting in the coil being "on" only 10 % of the time.
It would therefore be helpful to configure neighboring coils to spend their 20% in the "on-state" at different times, in effect realizing a de-coupled transmit system. Hence by time-sharing, we can achieve a de-coupled transmit system using our electronics system. Naturally, the more transmit channels we have, the less time each will spend in the "on-state", we can somewhat mitigate this by using higher voltages in the transmit system, migrating to say 12V compared to the 5V we use now. This is very different from managing coupling in traditional MRI systems, where neighboring coils are often on at the same time, and hence coupling needs to be managed differently, often with complex pre-amplifier circuits, simpler "geometric-decoupling" e.t.c.
That said, transmit coupling can be both beneficial and harmful. If by chance the waveforms we seek to transmit on each channel are such that they result in coupling currents aligning with those we need to generate. The coupling is beneficial in a sense that the resulting coupling current reduces the current we need to generate. Similarly, coupling can be harmful, if the coupling current opposes the current we need to generate, more current needs to be generated to "overpower" this coupling current. Hence, if left unchecked in a traditional system, much higher levels of RF power may be required. For example, consider the power requirements for two adjacent current loops, often called "Maxwell-coils" or "Helmholtz-coils". If we drive the current to generate a homogeneous magnetic field, the coupling between the two coils becomes harmful, as we must overpower the negative coupling current, similarly if we seek to create a "zero-field" in the middle, much less power is required as the coupling is helpful.
As we use our electronics system for multiple transmit channels, it may be necessary to have some channels couple to accommodate all of the systems requirements. If so, we need to manage the coupling, ideally arranging things such that when two adjacent transmit coils need to be on at the same time, the coupling that occurs is beneficial. We can achieve this by carefully orienting our coils. Note that while we can only send current one way in our transmit coil, we can flip the coil to control which way it will go.
Extensions and various embodiments It should be said that the relatively low frequency of our MRI system is a contributor to the success of our experiments so far. We operate at 2.08 MHz compared to 64 MHz for a traditional 1.51 MRI system.
The transmit system operates at a 300 MHz frequency, hence we have about 144, 300 MHz cycles during one 2.08 MHz cycle. Our modulation scheme translates that into approximately 36 possible "lengths" of the PWM waveform for each quadrant, now since we use different quadrants for positive and negative values, overall, we can accommodate 2 x 36 = 72 states for a real transmit sample. Contrast that to a 64 MHz operating frequency where we have just over 1 cycle per quadrant, and hence only two states can be accommodated for the same real sample. That said, it ought to be possible to apply the same PWM
scheme for each baseband frequency sample. By the same token, the maximum transmit sampling frequency matches that of the operating frequency, thus a similar PWM scheme can be used to synthesize more states. In our system at 2.08 MHz, if we group 8 cycles together, that would result in a transmit sampling frequency of approximately 260 kHz, comparable to the maximum bandwidth used in many clinical MRI systems. Since we can play any of our 72 states during each sub-cycle in our 8-cycle group, that results in 576 states for our real sample, or a little over 9 bits. To operate our system at 64 MHz, we would require a similar meta-cycle arrangement to increase the number of available states from 2 to 512. If the need arises to operate at such frequencies, we fully expect to implement some of the ideas discussed below to increase the transmit,frequency from 300 MHz to well over 1 GHz providing many more states to work from.
In our prototype embodiment discussed above, we utilize a Spartan 6 FPGA. The Spartan 6 FPGA was chosen primarily as it is familiar to us. It has a maximum operating frequency of about 300 MHz. In order to increase the transmit system frequency, we leverage the Spartan 6 output-serializers (OSER2) module, which allows us to operate our logic at slower speeds, (about 75 MHz in our case) via parallelization. For simplicity, the logic data is carried from the Spartan 6 to the transmit coil, via an output pad, using a 3.3V CMOS output standard. If we change to a differential output standard such as low voltage differential signaling (LVDS), we can increase the transmit frequency to about 500 MHz. In differential mode, the Spartan 6 OSER2 module supports up to an 8 bit shift register (compared to 4 bits for CMOS), allowing us to achieve 500 MHz, while slowing down our transmit logic to 62 MHz. It should be said, that using LVDS would require us to add a small buffer circuit to the transmit coil in order to translate from LVDS to CMOS which is what the transistor expects. Such circuits are available commercially, thus we expect to use LVDS in future revisions of our system.
The Spartan 6 FPGA is starting to show its age, in fact Xilinx Inc. recently released a Spartan 7, which is capable of faster logic speeds. There are also other FPGA's we can use from different vendors. We are currently studying using a Zynq 7 hybrid FPGA to access higher transmit frequencies. Furthermore, the Zynq 7 can operate a small Linux operating system which many others have used for embedded applications. This may with time come to replace the PC for even more compact portable MRI.
Currently, we employ a simple "left-fill" of the transmit waveform in Figure 9. This need not be the only way to perform the filling. For example, a "right-fill" or a "centric-fill"
where we turn the device on towards the end of the quadrant, or in the middle of the quadrant instead of the beginning might be helpful. In fact, a "centric-fill" might perform better with less phase distortion. There is also an opportunity to de-couple our coils very nicely by having channels that otherwise would couple strongly perform different filling schemes. That way, if we say keep to 50% of max-power, there is no possible way that two channels, one left-filling, the other right-filling are on at the same time, hence minimal cross coupling would be expected.
While we have confined ourselves to magnetic resonance imaging, there is no reason why our electronics system can not be useful for other fields of interest. RADAR, communications, and even ultrasound are just few of the fields that may benefit from this work.
Claims 1. An electronics system for magnetic resonance imaging substantially as shown and described.
2. A method of magnetic resonance substantially as shown and described.

Claims (2)

1. An electronics system for magnetic resonance imaging substantially as shown and described.

2. A method of magnetic resonance substantially as shown and described.