Any-nucleus distributed active programmable transmit MRI coil

Abstract

A magnetic resonance imaging transmit coil includes an input to receive one or more control signals, one or more voltage or current sources to produce one or more source signals, at least two coil segments, each coil segment to receive the one or more control signals and the one or more source signals, each coil segment having one or more switches to convert the one or more source signals to one or more arbitrary frequency currents, the one or more currents causing a magnetic field.

Description

ANY-NUCLEUS DISTRIBUTED ACTIVE PROG RAM VIABLE TRANSMIT MRI COILTECHNICAL FIELD

[0001] This disclosure relates to magnetic resonance imaging (MRI) coils, more particularly to an MRI coil that can be used with arbitrary nuclei.BACKGROUND

[0002] There are 118 known elements. Nearly all of them have NMR (nuclear magnetic resonance) active isotopes and at least 39 different nuclei have been shown to have biological relevance. Despite this, most of today’s MRI is based on only one nucleus -1H.

[0003] 1H has the second highest gyromagnetic ratio, only after H, and it has the greatest abundance in the human body, so it is natural that imaging tools have historically prioritized it when SNR (signal-to-noise ratio) was deficient. Over the decades, however, MRI technology has made immense gains in SNR with hypcrpolarization, high and ultra-high field magnets, anatomy-fitting receiver coils, improved reconstraction, and other techniques. With these SNR gains, the imaging of nuclei other than1H, or X-nuclei, has become more clinically feasible and a variety of studies have emerged making use of the essentially perfect nuclear specificity of NMR / MRI to gain information not possible with1H alone. Ultimately though, the proliferation of these studies is still held back by the low availability of the tools able to perform them.

[0004] In terms of ease of running experiments, nuclear specificity presents a double-edged sword. Studying X-nuclei needs heavy investment to obtain additional expensive MRI hardware, such as RF amplifiers, coils, and receiver chains, to enable imaging the specific nuclei of interest. This high entry barrier severely limits the study of many nuclei which have less general applicability due to low natural abundances in the human body but may have more direct biological relevance. Examples, without limitation, include23Na,31P,13C, and2H.2H has very low natural abundance, but has been recently regaining interest as a metabolic probe when ingested or injected. Other examples include3He or129Xe, which are used for imaging lung function when hyperpolarized and inhaled, and Li, which is used as a treatment for both bipolar disorder and bone fractures. Besides these and other elements notlisted, there are many more of the biologically relevant 39 different nuclei from 33 elements having untapped potential.

[0005] Coils for X-nuclei do not have wide availability. Even if one could acquire one, further complications exist. If one uses1H to obtain structural images and / or co-registers multiple images from different nuclei, the changing of coils introduces complexity and registration issues into the MRI workflow.

[0006] The use of a single coil to efficiently excite arbitrary nuclei at high field has until now been an unsolved problem in macro-scale MRIs. The usual way to do multinuclear studies with a single coil involves multi-tuning the coil. However, complexity rises very quickly with the number of nuclei and quality decreases with more tunings due to increased losses. This results in usually limiting multi-tuned coils to only two or three nuclei. An impressive quintuple-tuned coil has recently been described to image ‘H,19F, '1P,23Na and13C, but it would be challenging to continue to add more nuclei. Some other multinuclear coil attempts used variable circuit elements to tune coils to different nuclei on-the-fly, but they all have limited tuning ranges and may require mechanical changes for better performance. NMR has small length scales, and some reported approaches used untuned micro-spirals and transmission line coils with resistive termination, but these would not scale to macro-scale MRI.BRIEF DESCRIPTION OF THE DRAWINGS

[0007] FIG. 1 shows a comparison of current methods of connecting a coil and methods of connecting to a coil in accordance with the embodiments.

[0008] FIG. 2 shows a schematic of a simulation for an any-nucleus coil.

[0009] FIG. 3 shows a schematic of an embodiment of an any-nucleus coil.

[0010] FIGs. 4A-F show simulated graphs of current at the fundamental frequency versus switching frequency for different coil inductances, resistances, and control signal timings.

[0011] FIGs. 5A-C show graphs of simulation results for different input voltages.

[0012] FIGs. 6A-D show photographs of an embodiment of an any-nucleus coil and a graph of results.

[0013] FIG. 7 shows MR images for1H,23Na,2H, andk'C using an embodiment of an any- nucleus coil.

[0014] FIG. 8 shows images of ex vivo MRI using an embodiment of an any-nucleus coil.

[0015] FIGs. 9A-B show images of an ‘H any-nucleus coil Bi maps.

[0016] FIGs. 10A-D show examples of a saddle-shaped and a birdcage-shaped coil.

[0017] FIG. 11 shows examples of control waveforms used to generate current in a coil.

[0018] FIG. 12 shows control and current waveforms, and current frequencies in a coil.DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The embodiments address issues of X-nuclei coil availability and complications byworking towards an MRI system capable of imaging any nucleus with a single set of hardware. The embodiments focus on the RF transmission and present a completely new kind of MRI transmit coil capable of exciting the resonance of any nucleus in a whole-body MRI scanner on a macroscopic scale.

[0020] The embodiments here depart altogether from the concept of tuned macro-scale coils and present an untuned, yet scalable, 9 cm diameter transmit surface coil capable of efficiently transmitting at arbitrary frequencies using a unique architecture composed of many high-frequency semiconductor power switches integrated directly into the coil structure. By doing so, the coil and RF amplifier are merged into a single device that directly converts DC power into RF fields at any relevant frequency. On the receive side, one can use several single-tuned receiver coils to demonstrate the capabilities of the transmit coil.

[0021] Inductance resists changes in current which results in magnetic fields changes. Inductance makes it difficult to produce fast-oscillating RF magnetic fields. As an example, in the context of human-scale MRI, a 9 cm diameter, single-turn loop coil may have an inductance of approximately 300 nH (nanoHenry). To produce a modest 5 pT magnetic field at the center of the loop, approximately 360 mA of current is needed according to Biot- Savart’s Law. Biot-Savart’s law computes magnetic field B at a position, r, generated by an electric current.

[0022] The formula for inductance is vwhere v is voltage, L is inductance, and is the derivative of current with respect to time. To produce a peak current of 360 mA in the 300 nH coil at the ’H 3T Larmor frequency of 127.7 MHz, it needs a peak voltage of about 87 V at 127.7 MHz. To achieve similar excitations for nuclei other than1H, generally, the frequencies decrease but the required magnetic field and current increase proportionally due to the lower gyromagnetic ratios. The voltages required are about the same for!H and X-nuclei to produce equivalent flip angles in equal time durations. To achieve these large voltages, resonant circuits are usually employed to transform the lower voltage from the output of a RF power amplifier into that required to produce the current across the inductor. Unfortunately, though, such resonant circuits have bandwidths that are inherently limited by the Bode-Fano limit. In a typical MRI context, these bandwidths are on the order of hundreds of kHz, only able to cover the resonance of one nucleus at a time.

[0023] Because resonant circuits are severely limited in their bandwidth, to achieve multinuclear capability, one can imagine that it may be possible to simply directly connect a very high voltage amplifier to an untuned coil to produce the 87 or so Volts. At RF frequencies, though, this becomes difficult because an RF amplifier capable of 87 Volts would usually be quite large. Further, if a transmission line separates the RF amplifier and the untuned coil, this causes large power reflections from the coil, potentially damaging the amplifier and leading to safety issues.

[0024] Alternatively, one may move away from the conventional 50-ohm output RF amplifier and use transistors to switch voltages directly at the coil. Indeed, this is essentially the approach taken by Mandal et al. and Hopper et al. in their ultra-broadband low-frequency NMR system covering up to 3 MHz. (Mandal S, Utsuzawa S, Cory DG, Hiirlimann M, Poitzsch M, Song YQ. An ultra-broadband low-frequency magnetic resonance system.Journal of Magnetic Resonance. 2014;242: 113-125. doi:10.1016 / j.jmr.2014.02.019; Hopper T, Mandal S, Cory D, Hiirlimann M, Song YQ. Low-frequency NMR with a non-resonant circuit. Journal of Magnetic Resonance. 201 l ;210(l):69-74. doi: 10. 1016 / j.jmr.2011.02.014; and Mandal S, Utsuzawa S, Song YQ. An extremely broadband low-frequency MR system.Microporous and Mesoporous Materials. 2013; 178:53 5.oi:10. 1016 / j.micromeso. 2013.03.040.)

[0025] In this approach, an H-bridge switches up to 200 Volts from one side of the coil to the other in an alternating fashion at the RF frequency, producing an RF magnetic field. FIG. 1A shows a circuit architecture having a coil 10 and a DC voltage source 20. The voltage source connects to the coil through a set of switches, 12, 14, 16, and 18. This represents a traditional H-bridge. The switches connect the DC voltage source across the coil in alternating directions depending on which set of switches are closed. FIG IB shows a first configuration in 'which switches 12 and 18 close to connect the DC source in one direction. When switches 12 and 18 open, and switches 14 and 16 close, an AC current is produced in the coil. Current flowing in the coil generates the magnetic field.

[0026] Ultimately though, this approach is limited by the capabilities of the semiconductor devices used. Devices capable of handling higher voltages tend to be slower, and as of the time of writing, the inventors are unaware of any device capable of directly interfacing to the coil without a resonant circuit at frequencies up to 127.7 MHz (;H 3T Larmor frequency) in a reasonable way. As Mandal et al. note for their approach, “it is possible to extend the technique to higher frequencies if smaller coils are used.” Although this works and is like the approach taken by ultra-broadband micro-spiral coils for multinuciear NMR, too many coils would be needed to scale to human-scale high-field MRI.

[0027] In an embodiment shown in FIG. ID, the simplest configuration of an Any-nucleus Distributed Active Programmable Transmit (ADAPT) coil splits the coil in half, one half 22 and one half 24. Two switches 26 and 28 connect the DC voltage 20 source across each half in an alternating manner. FIG. IE shows switch 26 closed causing current flow in coil half 24. FIG. IF shows switch 28 closed causing current to flow in coil half 22. Changing the connections produces AC current in the coil similarly to the H-bridge.

[0028] The advantage of the ADAPT coil configuration lies in the ability to continue to split the coil in a scalable manner by adding more switches and connections to the DC voltage source as shown in FIGs. 1G-I. By splitting the coil into smaller segments, the same voltage from the voltage source can drive more current through the coil segment inductances without adding more voltage stress to the switches. FIG. 1G shows an embodiment where the coil has three coil segment pairs 30, 32, and 34. Each segment pair has a pair of switches. In FIG. 1H, one switch in each pair is closed, switches 36, 38, and 40, the current flows in the direction of the current arrow’s. FIG. II show's the opposite switches closed in each pair, switch 42, 44, and 46. This causes the current to flow in the opposite direction of the flow' of FIG. 1H.

[0029] The embodiments of FIGs. 1G-I provide a solution to limited semiconductor device voltages. Instead of breaking the coil into two halves, one can break it up into smaller pieces, each with a set of switches connected to the negative terminal of the DC voltage source. Now the same voltage source drives several smaller inductances, coil segments, in parallel and can produce larger RF currents. These RF currents together then produce larger magnetic fields than what one set of switches could produce alone. Breaking the loop into segments scales approximately linearly with the radius of the loop, or proportionally to the circumference. In contrast, using many micro-spiral coils in parallel to form an effectively bigger loop, w'ould scale with the area of the loop, or proportionally to the radius squared.

[0030] The embodiments herein enable the direct application of voltage across a coil without being limited by the semiconductor voltage capabilities. The embodiments draw on approaches like the distributed active transformer, which addresses a similar problem of generating higher output voltages from low-voltage CMOS transistors by placing several transistors in a loop. FIGs. 1D-1F show the most basic case and allows for comparison with the H-bridge in FIGs. 1A-1C.

[0031] Like the H-bridge, switches alternating at the RF frequency change the current path of a DC voltage source. Instead of using four switches like in the H-bridge, only two switches are used at the expense of only driving half of the coil at a time. Importantly though, the two switches not connected to the negative terminal of the DC voltage source are no longer used. This removes the complexity of driving a ‘"high-side” power switch whose reference voltage would shift with the large voltage on the coil. Only using “low-side” switches make faster switches available for use. One should note that this embodiment uses one switch per coil segment, but more switches could be used to implement various tradeoffs. For example, switches in parallel could increase the current handing capability at the expense of increased parasitic capacitance or switches in series could increase voltage handing at the expense of increased on-resistance. “High-side” switches could also still be added if ones that are fast enough become available.

[0032] Driving only half of the coil at a time leads to some DC currents in each coil segment. This may cause some static magnetic field inhomogeneities during RF transmission. Increasing the number of coil segments reduces the static magnetic field inhomogeneities as adjacent coil segments with opposite DC currents become smaller and closer together. This also holds true for the DC currents running through the center of the coil from the DC voltage source. If the static magnetic field homogeneity during RF transmission has more stringent requirements though, adding a second identical coil on top of the first can potentially remove these effects. The second coil, rotated by one coil segment, causes the RF currents to run in both directions in the whole loop. Magnetic fields from DC currents would then be opposite in direction and cancelled out.

[0033] Another issue to address lies in the characteristics of the switches, specifically on- resistance and parasitic parallel capacitance. These non-idealities degrade ADAPT coil performance by increasing the coil losses. On-resistance increases loss while current flows, and capacitance increases loss when charge on the capacitor is dumped to ground when the switch is closed. Designing the ADAPT coil should take these loss mechanisms into account.

[0034] Directly switching a DC voltage across inductors does not produce a perfect sine wave. This introduces harmonics of the fundamental switching frequency, and the amplitudes depend on the exact switching pattern and circuit inductances and capacitances, especially the parasitics. These harmonics have no effect on the MRI signal and should be much smaller than the fundamental frequency. However, they may increase SAR (specific absorption rate), and one should consider them for safety. Several harmonic cancellation techniques used in power electronics inverters can remove harmonics without resonant circuits or traditional frequency filters. More ADAPT coils could be overlaid on top of each other or more complicated control signals could be used to implement these techniques.

[0035] In a simulation, one embodiment used a gallium nitride (GaN) power switch, chosen for its very low on-resistance of 0.04 ohms and fast switching speeds of over 100 MHz for realistic simulations involving switch imperfections. The simulation included an extra capacitor and series resistor placed in parallel with the GaN power switch to represent a transient voltage suppression diode to protect the power switch from damaging excess voltage. The equivalent circuit of FIG. ID was simulated with the inductance between the DC voltage source and the point at which the coil segments split set to 100 nH, referred to here as the source inductance.

[0036] FIG. 2 shows the circuit schematic made in Advanced Design System (ADS). SRC1 52 represents the DC voltage source that provides power for the resulting RF magnetic fields, SRC2 54 represents a power supply for the power switch integrated circuit, or chip, XI 56 and X2 58 represent the power switches, SRC3 60 and SRC4 62 represent digital control signals for the switches, L I 64 represents the inductance from the DC source to the coil segments (source inductance), L2 66 and L3 68 represent two complementary coil segments, and R1 70 with C2 72 and R2 74 with Cl 76 represent parasitics from a transient voltage suppression diode. One should note that while the embodiments use digital control signals, other embodiments may use analog control signals. In the case of analog control signals, the switch may also act as a more continuously varying variable resistor. One should note that in the case of an analog control signal, the switch behaves like a standard amplifier and can be implemented as one. The term “switch” as used here includes amplifiers, diodes, relays, transistors, and variable resistors. The probes for electric current were positioned at I Probel 78 and I_Probe2 79.

[0037] The coil segment inductance was swept for a constant DC voltage supply to show the effect of increasing current with decreasing coil segment inductance. To show the effect ofthe source inductance, the source inductance was also swept for a constant segment inductance and DC voltage supply. Source and segment resistances were also swept, and skews between two complementary control signals and their rise / fall times were also swept. While the example given uses two control signals, other embodiments may use only one control signal from which other control signals are derived, such as by connecting switches of opposite polarity. Finally, the DC voltage was also swept for constant coil segment and source inductances for a few frequencies of interest. Due to the symmetry of the ADAPT coil, only one pair of coil segments needs to be simulated and a deceasing simulated segment inductance represents splitting the coil into more segments. Harmonic balance simulations in ADS were used for faster simulation speeds than transient simulations, and simulation data from ADS was imported to MATLAB (The MathWorks, Natick, USA) for plotting.

[0038] FIG. 4 shows the results of the ADS simulations for operating the ADAPT coil between 10 MHz and 130 MHz for a few different coil segments and source inductance values. As the inductance of each coil segment decreases, shown in FIG. 4A, the current flowing through each coil increases, leading to higher magnetic fields and showing one of the benefits of splitting the coil into more segments in the ADAPT structure. Interestingly, there are dips in the plot for various frequency ranges. These occur when the circuit and switching parameters happen to convert more of the DC voltage into DC fields or frequency harmonics instead of the desired fundamental frequency. FIG. 4B shows the current for various source inductances, FIG. 4C show the current for various segment resistances, FIG. 4D shows the current for various source resistances, FIG. 4E shows the current for various control skews, and FIG. 4F shows the current for various control rise / fall times. These show the effects of the non-idealities in the circuit and non-ideal control signals.

[0039] As shown in FIGs. 5A-C though, dips in the FIG. 4 plots do not imply lower magnetic field strengths because when the DC source voltage increases, the ratio between fundamental frequency current and harmonic current becomes better and the dips can be overcome. This effect at higher voltages appears because when the drain voltage on the switch is negative enough, reverse conduction through the switch occurs and stops parasitic oscillations which contribute to the harmonics. As the source inductance changes, shown in FIG. 4B, there is much less of an effect on the current as the source inductance mostly carries DC current. This assumption becomes better with higher source inductances, but the effect of changing source inductance mostly changes the locations of the dips in the plot.

[0040] FIGs. 5A-C shows the results of the ADS simulations expressed in the time and frequency domain for a few nuclei of interest with the DC source voltage swept to change the RF amplitudes and the segment inductances and the source inductance held constant at 10 nH and 100 nH, respectively. Simulated plots are shown for operating the ADAPT coil at the 3T Larmor frequencies of ’H,I3C, and15N, chosen as examples that are well spaced in frequency. Note that the13C frequency lands in one of the dips in the plots from FIG. 4 with 10 nH segment inductance and 100 nH source inductance.

[0041] As the operation frequency decreases, the time domain current waveforms, and the resulting magnetic fields, become less sinusoidal. This is mirrored in the frequency domain plots. The higher frequencies are more sinusoidal because of filtering by the inductances and capacitances. The switch drain voltage limits the maximum RF output, because if the voltage exceeds a threshold, such as 40 V for the GaN switch used here, the switch may break. As can be seen in the plots, lower operation frequencies also tend to hit this limit at lower DC source voltages. Ultimately though, in this embodiment the maximum current and RF magnetic fields achievable are similar for the different frequencies. The efficiencies, calculated as the current in the fundamental frequency divided by the square root of power from the DC voltage source, are also similar for different frequencies at higher field strengths. This current efficiency is for a particular coil segment with these circuit parameters, but magnetic field efficiency can be found by assuming a coil geometry and knowing how many coil segments are needed to create it. If eight coil segment pairs of 10 nH can create a 9 cm diameter coil, the efficiencies shown in FIGs. 5A-C can be converted from A / ^ / W to qT / y[W at the center of a 9 cm diameter coil by multiplying the A / s / W value by about 14 / \ / 8”qT / A, where 14 comes from Biot-Savart’s law to convert current in a 9 cm loop to magnetic field at the center and V8" comes from eight times the power for eight coil segment pairs, but with a square-root to express it in terms of current efficiency. With more complicated control signals, efficiency is expected to increase, especially for the lower RF frequencies.

[0042] Decreased segment inductances also decrease switch voltage stress for any particular current, and thus magnetic field, because v =

[0043] The inventors implemented the simulated architecture using commercial parts assembled on a commercially fabricated four-layer FR4 printed circuit board (PCB) from JLCPCB (Shenzhen, China) shown in FIG. 6. FIG. 3 shows a block diagram of the chips usedand their connections, as well as the auxiliary equipment used to operate the ADAPT coil. The diagram shows one pair of segments on a coil, which would be replicated for any number, N, of pairs of coil segments 108. Coils could further be replicated.

[0044] The MRI Scanner 82 and waveform generator 80 provide inputs to two low voltage differential signaling (LVDS) digital converters 84 and 86, which were used to independently control the two sets of out-of-phase high-power switches 100 and 102. Having two independent inputs allowed for the ability to switch both sets of switches off at the same time when not transmitting and producing no current flow. These two inputs were then replicated to each relevant switch using the LVDS repeater 88. The embodiment used the GaN switches. Relevant chips for digital logic conversions such as LVDS to CMOS translator 96 and voltage and power regulation using regulators 90, 92, 94 and all associated decoupling capacitors were also added. Transient voltage suppression (TVS) diodes 104 and 106 w'ere placed across each switch to protect them in case of abnormally strong transients.

[0045] To operate the coil, the pulse sequence from the MRI scanner 82 triggered externally generated LVDS signals which set the desired RF frequency. The RF amplitude was set by a DC-coupled audio amplifier output voltage 92, which plays the role of the main DC voltage source that gets converted to RF. Only hard RF pulses were used in this demonstration, but one should note that the DC voltage and RF frequencies can be time-varying. One should also note that the source here comprises a voltage source, but it could be a current source as well. The output of the source, whether a voltage source or a current source will be referred to as a source signal. One should note that while one source is used here, multiple source signals instead of one could also be used and routed to different coil segments.

[0046] In other embodiments, RF amplitude modulation could be done by making the audio amplifier output time-varying. One should note that the amplifier here is an audio amplifier, but an amplifier or source with frequencies outside of the audio frequency range could be used as w'ell. The amplifier bandwidth could affect the transient characteristics of the coil.

[0047] Because the coil is untuned and its self-resonant frequency is above any operational frequency, designed by making the inductance of the coil segment low;enough such that the resonance between it and the switch parasitic capacitance is high, coupling between the transmit coil and the receive coil is negligible and no extra transmit coil detuning is required during receive.

[0048] A benchtop DC voltage sweep wTas conducted at six operation frequencies to empirically characterize the relationship between the DC supply voltage (400 mV to 6000mV in 400 mV increments) from the DC-coupled audio amplifier and the magnitude of the RF magnetic field output by the fabricated ADAPT Coil. The frequencies were chosen to be 12.94, 19.60, 32.11, 33.78, 51.69, and 127.7 MHz, corresponding to15N,2H,i3C,23Na,31P, and ’H nuclei at 3T, respectively. To measure the magnitude of the RF magnetic field, a calibrated electromagnetic compatibility probe (model 100B from Beehive Electronics, Sebastopol, USA) positioned 19.6 mm from the center of the coil was used in conjunction with the real-time spectrum analyzer mode of a FieldFox N9918A (Keysight Technologies, Santa Rosa, USA). A SDG6022X function generator (Siglent Technologies, Shenzhen, China) was used to send 25 ms RF pulses every' 500 ms to the inputs of the sinewave-to- LVDS converters.

[0049] FIG. 6A-C show pictures of the fabricated coils in this example and D shows a graph of the results of the benchtop voltage sweep. FIG. 6A shows the top view of coil 109, with FIG. 6B showing an up-close view of connector 110. FIG. 6C shows the back of coil 109. The coil elements follow the block diagram in FIG. 3 and the layout is a version of FIG. 1G with eight coil segment pairs. As mentioned above, FIG. 3 shows the schematic for one coil segment pair. While this embodiment uses 8, no limitation to that number is intended nor should it be implied.

[0050] FIG. 6D presents the outcomes of the benchtop voltage sweep, showcasing the relationship between the probed RF output and the DC voltage from the audio amplifier. Within the swept voltage range, a positive and roughly linear correlation is evident between the magnitude of the RF output and the DC voltage. Despite aligning with the trends depicted in FIG. 4, the experimental results reflect some nuance not captured by the simulations discussed above. Notably, one can observe in FIG. 6D that the empirical magnitude of the RF field begins to saturate at higher DC voltages along the sweep range, especially for 51.69 MHz and 127.7 MHz.

[0051] At the top of FIG. 7, from left to right are images ofJH,2H,23Na, andIJC from small vials of samples. The samples consist of distilled water, salt water of three different salt concentrations, deuterium oxide, and urea (13C). The lower left of FIG. 7 shows a combined image of the four nuclei. The lower center of FIG. 7 shows an optical image of the samples, the lower right of FIG. 7 shows an illustration of the samples.

[0052] FIG. 8 shows ex vivo images of a bone-in ham steak, and a pig knee, for1H and23Na. This demonstrates the imaging results with actual tissue. The Experiment below provides more information on these images.

[0053] Experiment

[0054] Without changing out the ADAPT Coil, ’H,2H,i3C, and23Na images were taken of a phantom by changing the control signals for transmission. The phantom was composed of five 10 mL bottles arranged in a circle and a 1 mL tube with13C enriched urea (8 M concentration) plus gadolinium (0. 1 mM concentration) placed in the center as shown at the center bottom of FIG. 7. The five 10 mL bottles were composed of distilled water, deuterium oxide (Sigma-Aldrich, St. Louis, USA), and salt water of three different NaCl concentrations (10 g / L, 20 g / L, 40 g / L). Because the scanner used does not have multinuclear capabilities, a custom receiver chain was built for single-channel receive-only coils. This receiver chain upconverts X-nuclei signal to the1H frequency to use the existing 'H receiver chain for data acquisition. Separate single-tuned surface receive coils were used for each nuclei imaged.

[0055] Pulse sequences were written using the Spin bench / RTHawk platform (Vista.ai, Los Altos, USA) and run on a GE 3T MR750w scanner (GE Healthcare, Waukesha, USA). Gradient echo (GRE) sequences were used with the following parameters: coronal imaging, 3.2 ms non-selective Hard RF pulse (except for1H with a 2 mm slice selection), 300 ms TR, 3.7 ms TE, 6 V DC voltage transmit coil input (except for 'H with 200 mV) for all nuclei; 1.25x1.25 mm2resolution, 256x256 matrix, and 125 kHz readout bandwidth for2H; 10x10 mm2resolution, 16x16 matrix, 25 kHz readout bandwidth, and 100 averages for2H; 5x5 mm2resolution, 32x32 matrix, and 50 kHz readout bandwidth for13C and23Na with 100 and 25 averages respectively. Flip angles at the center of the phantoms were estimated based on benchtop Bi calibrations using the same equipment and cables as the imaging experiments to be 3.5°, 10°, 53°, and 63° for1H,2H,13C, and23Na, respectively.

[0056] Ex vivo ’H and23Na images of a bone- in ham steak and a pig knee slice were also taken as shown in FIG. 8. The deviations in parameters from the phantom imaging were: 1) non-selectiverH excitation, and 2) for23Na imaging, 100 and 500 averages for the ham steak and pig knee images respectively with 100 ms TR.

[0057] For display,1H images were cropped to 128x128, and X-nuclei images were interpolated to 128x128 with zero-filling and a Hamming window.

[0058] To map!H Bi+, the experiment used a 3-mm thin slice phantom (30 mL solution of I g / L CuSCL and 50 g / L NaCl, -260 ms Ti) placed 27 mm away from the ADAPT Coil. Sequence parameters were: 1 s TR, 3.7 ms TE, 2.5x2.5 mm2resolution, 128x128 matrix, non- selective hard pulse excitation, 3 averages. All images were received with the body coil while excitation was achieved either with the body coil or the ADAPT Coil without moving anyhardware. The ADAPT Coil was in position and powered on (but not transmitting) when the body coil was used to transmit. Two images Sx(r) and S2(r) using body coil excitation were taken with a and 2a flip angles. Let C (r) be the magnetization weighted by body coil receiving sensitivity, and assume full relaxation of magnetization with the long TR,5t(r) = C(r) sin(a(r)), [1]S2(r) = C(r) sin(2a(r)) = 2C(r) sin(a(r)) cos(a(r)), [2]and

[0059] Then, given any image SA(r) acquired with the same setup, but using the ADAPT Coil for excitation,SA(r) = C(r) sin(|3(r)) [5] and p 6(zr x)H

[0060] Eq. [6] was used to compute the Br map of the ADAPT Coil for a range of voltages (up to 10V) shown in FIG. 9. One cannot double the DC voltage to double the flip angle due to the nonlinearity of the ADAPT Coil. The experiment did not have reference transmit coils for other nuclei.

[0061] FIG. 9A shows the1H Bi:maps for different voltages, offering insight into the cause of the RF field saturation discussed above. In the current coil implementation, higher Bi fields seem to disrupt the phase relationship between switches at different positions in the coil, leading to some destructive interference. This interference may result from feedback between the Bi field and the control lines to the switches, influenced by the asymmetric routing of the control lines and ethemet connector. Future coil implementations will pay closer attention to coil layout symmetry and shielding of the control lines. FIG. 9B shows one-dimensional plots of flip angle vs. DC voltage for the 4 regions of interest shown on the left side of FIG. 9B.

[0062] In this manner, one can build arbitrary nuclei transmit coils for macro-scale MRI. To cover more anatomy, a transmit coil array can straightforwardly be made using the surface coil presented here. Because no coil tuning and matching is involved, array coils using the ADAPT coil are expected to be much easier to build than in the usual case where coil couplings affect their tuning. Because there is no resonance involved, the effects of patient loading should also be negligible relative to the inductance of the coil.

[0063] The same concept for scalability can be applied to other shapes of coils rather than an array of surface coils, including solenoids and gradient coils. One can construct a birdcage coil by dividing each RF current-carrying segment of the birdcage into smaller inductances like what has been done for the surface coil presented here. FIGs. 10A-10D show different configurations for a “saddle” and “birdcage” shaped coil. FIG. 10A show’s a single saddle- shaped coil 112, with connectors such as 118 for each segment, which may be the same or different type of connector as connector 110 from FIG. 6B. FIG. IOC shows a saddle-shaped coil 116 having more vertical segments, and FIG. 10B show’s a birdcage- shaped coil 114 that has horizontal segments that form a circle. FIG 10D shows an embodiment of the coil having w’ires connected. One should note that while this discussion mentions specific shapes of coils, any shape of coil may be used.

[0064] Higher output currents and frequencies should be possible by making each coil segment have even smaller inductances. Higher frequencies can also be achieved by choosing a switch with less parasitic capacitance, but this is often at the expense of higher on- resistance.

[0065] Implementing amplitude and frequency modulated pulses can be done by changing the DC voltage source into a time-varying one for amplitude modulation and changing the switching frequency over time for frequency modulation. Due to the nonlinearity of the coil, careful characterization of the coil and pre-compensation are likely needed to improve waveform accuracy. FIG. 11 shows an example of this in simulation using the data from FIG. 5 and the circuit model from FIG. 2.

[0066] The input voltage waveform and control pulses w’ere optimized to produce a hamming-windowed sync pulse with a time-bandw’idth product of 4, given the circuit model from FIG. 2. The top row’ is the ideal pulse, along with its ideal frequency profile in the last column. FIG. 11 shows, from left to right, the input voltage, the control phase, the output current, and output frequency, and a zoomed-in view of the output frequency for each pulse. The second row’ shows the result of inputting the ideal pulse as an input to the circuit in FIG.2, which results in a severely distorted frequency profile. The third row shows corrections for the input voltage only, and the fourth row shows corrections for both the input voltage and control phases. Control phase corrections are needed because as the input voltage changes, there are slight changes in the output current’s phase due to the nonlinearity of the realistic switch model. The RF amplifier can shape the pulse based upon non-linearity of the coil.

[0067] FIG. 12 shows the average control waveforms for13C at 32.1 MHz, and]H at 127.7 MHz sent directly to the switches in the circuit model of FIG. 2. FIG. 12 also shows the current waveform in the second row, and the currents in the third row, which are related to the magnetic fields. The simulation also produced many intermodulation frequences. The DC voltage for the simulation was set to 0.5 V. This example shows that currents can be produced at multiple desired frequencies simultaneously.

[0068] Benchtop tests show that the ADAPT Coil can also function as an arbitrary-nucleus receiver, but in its current implementation, a large SNR penalty makes it suboptimal. Mainly, the on-resistance of the switches, harmonic conversion, and switching losses contribute more to coil noise than the usual case of just copper and passives, especially when the signal on the coil is low, as seen by the low efficiency at low currents as in FIG. 5. Potential other paths to broadband receive coils include using resonant coils with broadband high-impedance amplifiers or having an amplifier with such low noise that a resonant amplification stage is not needed, perhaps with cryogenic amplifiers.

[0069] Even without being paired with an optimal arbitrary nuclei receive coil, the ADAPT coil provides many benefits by itself. For standard ‘H studies, the ADAPT coil can potentially drastically reduce MRI transmit coil production costs by not requiring any specialized human labor to assemble. This may be particularly appealing to emerging low- cost MRI scanner initiatives. All the components are commercially available, and the coil can potentially be ordered fully assembled from a standard PCB facility. When parts are purchased in bulk, unit costs are reduced, further reducing MRI system costs. Furthermore, using an audio amplifier instead of a much more expensive and usually less power-efficient RF amplifier to provide the RF power reduces the costs.

[0070] Unlike standard MRI coils that only work at a single static field strength for a single nucleus, the ADAPT coil can work at any static field strength for either the same nucleus or any other nucleus. It can be reused between MRI scanners of different static field strengths or used at different static field strengths in an MRI scanner with a variable static field strength with variability in space or time.

[0071] Besides lower costs, the ADAPT coil could also provide easier transmit Bi calibration due to negligible loading effects, and because the coil in the current stops almost instantly when the switches are open, it could also provide much lower deadtime between the transmit and receive time in an MRI system for ZTE / UTE studies. In one embodiment, this reduction in deadtime occurs because the transmit coil can operate at shorter transmit pulses without resonant circuits.

[0072] Other variations include programming or otherwise configuring the control signals to cause a voltage or current source to produce source signals in the form of currents having a desired waveform and for a desired duration. The one or more control signals could be configured to cause a voltage or current source to produce currents of multiple desired frequencies to excite multiple nuclei simultaneously.

[0073] The control signals could be configured to cause a voltage or current source to produce currents of low frequencies in the range from DC to kHz for multiphoton excitation, spatial encoding, or field shimming. The control signals could also be configured to cause the voltage or current source to produce currents of low frequencies in the range from DC to kHz and high frequencies in the MHz range for multiple purposes simultaneously or sequentially. For example, field shimming using low frequencies could be done while exciting nuclei using high frequencies. kHz and MHz magnetic fields resulting from the currents could also be used together to produce multiphoton excitation on one or more nuclei.

[0074] The transmit coil could be combined with a receive coil to produce X-nuclei images of any nucleus simultaneously or sequentially, using the various control signals as discussed regarding FIG. 12. The transmit coil could be used at multiple MRI scanners of different static magnetic fields for the same or different nuclei.

[0075] N MR signal excited by the transmit coil can also be used for NMR spectroscopy, detection of specific molecules via NMR spectroscopy, for spatially localized NMR spectroscopy (MRS), and for magnetic resonance spectroscopy imaging (MRSI).

[0076] Additionally, this written description makes reference to particular features. It is to be understood that the disclosure in this specification includes all possible combinations of those particular features. For example, where a particular feature is disclosed in the context of a particular aspect, that feature can also be used, to the extent possible, in the context of other aspects.

[0077] Also, when reference is made in this application to a method having two or more defined steps or operations, the defined steps or operations can be carried out in any order or simultaneously, unless the context excludes those possibilities.

[0078] All features disclosed in the specification, including the claims, abstract, and drawings, and all the steps in any method or process disclosed, may be combined in any combination, except combinations where at least some of such features and / or steps are mutually exclusive. Each feature disclosed in the specification, including the claims, abstract, and drawings, can be replaced by alternative features serving the same, equivalent, or similar purpose, unless expressly stated otherwise.

[0079] It will be appreciated that variants of the above-disclosed and other features and functions, or alternatives thereof, may be combined into many other different systems or applications. Various presently unforeseen or unanticipated alternatives, modifications, variations, or improvements therein may be subsequently made by those skilled in the art which are also intended to be encompassed by the embodiments.

Claims

WHAT IS CLAIMED IS:

1. A magnetic resonance imaging transmit coil, comprising: an input to receive one or more control signals; one or more voltage or current sources to produce one or more source signals; and at least two coil segments, each coil segment to receive the one or more control signals and the one or more source signals, each coil segment having one or more switches to convert the one or more source signals to one or more arbitrary frequency currents, the one or more currents causing a magnetic field.

2. The transmit coil of claim 1 , wherein the coil comprises any one of the group consisting of: a surface coil, an array of coils, a saddle-shaped coil, a birdcage coil, a solenoid, or a gradient coil.

3. The transmit coil of claim 1, wherein the one or more switches comprise one of amplifiers, transistors, integrated circuits, diodes, relays, or variable resistors.

4. The transmit coil of claim 1 , wherein the coil segment converts the one or more source signals to one or more arbitrary frequencies that excite the resonance of any nucleus or nuclei.

5. A method of operating a magnetic resonance imaging transmit coil, comprising generating one or more control signals; generating one or more source signals using one or more voltage or current sources; receiving the one or more source signals and the one or more control signals at two or more coil segments; activating switches according to the one or more control signals at two or more coil segments to convert the one or more source signals to one or more arbitrary frequency currents to cause a magnetic field.

6. The method of claim 5, wherein the one or more control signals are configured to cause the one or more voltage or current sources to produce a current having a desired waveform and for a desired duration.

7. The method of claim 5, wherein the one or more control signals are configured to cause the one or more voltage or current sources to produce currents of one or more desired frequencies to excite one or more nuclei simultaneously or sequentially.

8. The method of claim 5, wherein the one or more control signals are configured to cause the one or more voltage or current source to produce currents for multiphoton excitation, spatial encoding, or field shimming.

9. The method of claim 5, wherein the one or more control signals are configured to cause the one or more voltage or current sources to produce currents of multiple desired frequencies for multiple purposes simultaneously or sequentially.

10. The method of claim 5, further comprising combining at least two coil segments with a receive coil to produce images of any nucleus simultaneously or sequentially.

11. The method of claim 5, further comprising combining at least two coil segments with a receive coil to produce NMR spectrums of any nucleus simultaneously or sequentially.

12. The method of claim 5, further comprising combining at least two coil segments with a receive coil to produce images or spatial localization of NMR spectrums or specific molecules found via NMR spectroscopy.

13. The method of claim 5, further comprising using the transmit coil with one or either an MRI scanner with a variable static magnetic field, or multiple MRI scanners of different static magnetic fields for the same or different nuclei.

14. The method of claim 5, further comprising using the transmit coil to reduce deadtime between transmission and reception in an MRI system by providing shorter transmit pulses without resonant circuits.

15. A magnetic resonance imaging (MRI) receive coil, comprising: two or more coil segments, each coil segments having: an input to receive one or more control signals for the coil segment; switches connected to the coil segment to change available current paths of currents induced by received magnetic fields; and one or more amplifiers to receive power and convert the power to radio frequency(RF) input voltage signals at one or more desired frequencies.

Images