A magnetic resonance system is disclosed. The system includes a transceiver having a multichannel receiver and a multichannel transmitter, where each channel of the transmitter is configured for independent selection of frequency, phase, time, space, and magnitude, and each channel of the receiver is configured for independent selection of space, time, frequency, phase and gain. The system also includes a magnetic resonance coil having a plurality of current elements, with each element coupled in one to one relation with a channel of the receiver and a channel of the transmitter. The system further includes a processor coupled to the transceiver, such that the processor is configured to execute instructions to control a current in each element and to perform a non-linear algorithm to shim the coil.
|
0. 11. A system comprising:
a radio frequency (RF) transmit coil having a plurality of current elements;
a multichannel transmitter having a plurality of transmit channels, each of the plurality of transmit channels coupling to at least one of the plurality of current elements in the RF transmit coil and being configured for independent selection of at least one of frequency, phase, magnitude, and time; and
a processor coupled to the multichannel transmitter, wherein the processor is configured to execute instructions to control a current in each element and to perform a non-linear algorithm that implements a non-linear relationship between flip angle and magnetic resonance signal intensity to shim the RF transmit coil, and wherein the non-linear algorithm uses a penalty function that returns a value describing at least a phase and a magnitude for each current element in the RF transmit coil.
7. A method comprising:
receiving a user selected criteria for optimization;
executing instructions using a processor to configure a multichannel transmitter and a multichannel receiver based on the criteria, the transmitter and the receiver each coupled to a multichannel coil wherein each channel of the coil includes a current element and each channel of the coil is coupled in one to one relation with a channel of the transmitter and a channel of the receiver;
executing instructions using a processor to perform a non-linear algorithm that implements a non-linear relationship between flip angle and magnetic resonance signal intensity to shim the coil, and wherein the non-linear algorithm uses a penalty function that returns a value describing at least a phase and a magnitude each current element of the multichannel coil; and
acquiring magnetic resonance data based on a signal received from the coil.
1. A system comprising:
a transceiver having a multichannel receiver and a multichannel transmitter wherein each channel of the transmitter is configured for independent selection of at least one of frequency, phase, time, space, and magnitude and each channel of the receiver is configured for independent selection of at least one of space, time, frequency, phase and gain;
a magnetic resonance coil having a plurality of current elements, each element coupled in one to one relation with a channel of the receiver and a channel of the transmitter; and
a processor coupled to the transceiver, wherein the processor is configured to execute instructions to control a current in each element and to perform a non-linear algorithm that implements a non-linear relationship between flip angle and magnetic resonance signal intensity to shim the coil, and wherein the non-linear algorithm uses a penalty function that returns a value describing at least a phase and a magnitude for each current element of the magnetic resonance coil.
2. The system of
3. The system of
5. The system of
6. The system of
9. The method of
10. The method of
0. 12. The system as recited in claim 11 further comprising:
a receiver having at least one receive channel configured for independent selection of at least one of frequency, phase, magnitude, and time; and
an RF receive coil having at least one current element, the at least one current element being coupled to at least one of the receive channels.
0. 13. The system as recited in claim 12 in which the RF receive coil includes at least two current elements, each of the at least two current elements being coupled to at least one of the receive channels.
0. 14. The system as recited in claim 12 in which the at least one receive channel comprises at least two receive channels.
0. 15. The system as recited in claim 11 in which one of the plurality of transmit channels is coupled to more than one of the plurality of current elements in the RF transmit coil.
0. 16. The system as recited in claim 11 in which one of the plurality of current elements in the RF transmit coil is coupled to more than one of the plurality of transmit channels.
0. 17. The system as recited in claim 11 in which the multichannel transmitter is further configured for dynamic spatial adjustment of the plurality of current elements.
0. 18. The system as recited in claim 11 in which the RF transmit coil includes a transverse electromagnetic coil element.
0. 19. The system as recited in claim 11 in which the processor executes instructions to perform dynamic RF field control of the RF transmit coil.
0. 20. The system as recited in claim 11 in which the plurality of current elements in the RF transmit coil are configured such that neighboring current elements are decoupled relative to each other.
0. 21. The system as recited in claim 20 in which the plurality of current elements are geometrically decoupled by orienting dipoles of the current elements for phase cancellation.
0. 22. The system as recited in claim 20 in which the plurality of current elements are geometrically decoupled by increasing a distance between current elements until mutual inductance is reduced below a threshold value.
0. 23. The system as recited in claim 20 in which the plurality of current elements comprises at least one of a plurality of transverse electromagnetic (TEM) coil elements and a plurality of loop coil elements.
0. 24. The system as recited in claim 20 in which neighboring current elements are decoupled using at least one of reactive decoupling and capacitive decoupling.
0. 25. The system as recited in claim 11 further comprising a tune and match network that includes at least one of a phase shifter and a phased line with diode protection.
0. 26. The system as recited in claim 12 wherein the receiver is a digital receiver.
|
By Ohm's Law the current density Jc=σE, and by Euler's Law the electric field displacement ∂D/∂t=∂∈E/∂t=jω∈E so that equation [1] can be rewritten as:
∇XB1/μ=(σ+jω∈)E [2]
The complex value of E can be written in terms of the magnetic vector potential A, and the electric scalar potential φ, such that:
∇XB1/μ(Je+Jd)=(σ+jω∈)(−jωA−∇φ) [3]
Influencing the B1 distribution and loss in human tissues loading the coil are the B1 field induced eddy current density Je,
Je=σE=−jωσA [4]
and the accompanying electric field displacement current density Jd for tissue specific values of σ and ∈.
Jd=−jω∈E=ω2∈A [5]
The sum and distribution of RF eddy current losses to the tissue conductor, and RF displacement current losses to the tissue dielectric, are determined by the total power loss density Pt loss,
Pt loss=0.5Iv(J·J*/σ)dv for J=Je+Jd [6]
Note that the displacement current density Jd increases with ω2 compared to the eddy current density's Je linear proportionality with frequency. This is consistent with the observations of interfering wave phenomena (commonly though inaccurately referred to as dielectric resonance) for head images at high B0 fields. In contrast to initial concerns that Je would severely limit B1 penetration and lead to excessive power absorption, Jd related inhomogeneities and losses may dominate at fields of 3 T and higher in the human head and body, leading to B1, SNR, loss and heating patterns not predicted by Biot Savart, quasi-static, or otherwise over simplified models.
RF Heating
RF loss induced temperature contours in the human anatomy can be calculated by equating the total losses Pt loss from the coil's RF field Equation [6] to the bio-heat (perfusion) equation below. Table values can be used to supply tissue specific thermal parameters for perfusion rate R, tissue and blood densities ρ and ρb, blood temperature Tb, blood specific heat cb, and tissue thermal conductivity K.
Pt loss=Rρρbcb[Tb−T]−∇K∇T [7]
With the increased RF losses Pt loss expected in higher field studies, the potential for excessive RF heating increases. A non-invasive means of monitoring heating in tissue is needed. Temperature contours in human anatomy at higher fields have not been reliably predicted or measured by NMR methods. Indirect method of assuring RF safety by correlating average RF coil input power with heating data (FDA guidelines) determined a priori, are used. Practical aspects of high frequency field propagation and losses in MRI head and body applications follow.
RF Problems
Inhomogeneity
“High field” for human imaging currently ranges from 3 T clinical imaging to 9.4 T imaging research. The Larmor band for proton imaging over this span of field strengths ranges from 30 cm tissue wavelengths at 3 T to 9 cm wavelengths at 9.4 T. Shortened wavelengths due to high tissue dielectric constants at high frequencies result in wave interference patterns with consequential RF field gradients and resultant image inhomogeneities over a field of view of human anatomic dimensions. Using the Finite Element Method (FEM),
SAR and SNR
SAR and SNR RF field dependent SAR and SNR also present problems at higher B0 fields. As shown in
RF Solutions
A variety of solutions can be used for addressing the problems associated with MR imaging of human anatomy at high magnetic field strengths. Some solutions are as follow.
Image Processing
To achieve uniform appearance of an image, apply image signal intensity correction or other post signal acquisition processing algorithms. See
Nested Transmit Coils and Receive Coils
Local transmit coils of conventional, circularly polarized birdcage or TEM design generate uniform, transverse B1 fields. However, a high field human head image from such a coil used for both transmission and reception results in an inhomogeneous image with signal bias to the center of the head as described in
Results
Strengths of high field human head and body imaging are outlined and described in this section.
Multi-Channel Transmit and Receive Coils
Alternatively, transmit and receive coils can be configured such that the B1 field of a coil can be optimized for image homogeneity or other criteria by controlling the RF currents on multiple, independent coil elements. One multi-channel example particularly well suited for high field use is the multi-channel TEM coil composed of transmission line elements which can be independently driven, controlled, and received efficiently at high frequencies. See
Lines Vs. Loops
One choice to be made in multi-channel coil design is whether to use line elements or loop elements in the coil construction. While both element choices can be made more efficient by transmission line design, the line element is the shortest and therefore the most efficient. Optimization schemes are simplified with the uni-directional currents on the line elements. Line element or “runged” coils such and the birdcage and TEM resonators are inherently more homogeneous than loop arrays as well. See
Design Execution
Specific execution of any coil design can be used to determine performance criteria. For the multi-channel TEM volume coil, as with other designs, a closer fitting coil can improve the efficiency of transmission and reception. Interference patterns leading to image inhomogeneity are more extreme when elements are spaced closely to each other and to the anatomy. See
RF (B1) Shimming
To control the phase, magnitude, timing and frequency of B1 field generating currents on independent coil elements of multi-channel coils, multiple spectrometer transmit and receive channels are used. RF signals on each channel are independently modulated to effect the desired field control. This modulation is in turn controlled from the console by user interaction, programmed algorithms, or automated, feedback driven optimization protocols. See
Parallel Transceiver
In one example, a parallel transceiver was configured to accomplish this level of RF control. The parallel transceiver schematized in
Applications
Head Imaging
Results from the application of some of the RF solutions presented, follow.
Multi-Nuclear Imaging
Signal-to-noise and spectral resolution gains greatly benefit multi-nuclear NMR at high field strengths. By tuning alternating elements of a TEM coil to two frequencies, and transmitting and receiving at these two frequencies, multi-nuclear image acquisition can be simultaneously accomplished. See
Body Imaging
4 T
The methods described herein for head imaging at high fields apply to full body imaging as well. Because human trunk dimensions are larger than in the head, significant RF artifacts become problematic at proportionately lower field strengths. The TEM body coil as shown in
7 T
For an initial mapping of the RF “landscape” in the body at 300 MHz (7 T), coronal, saggital, and transaxial scout images were acquired with a TEM body coil as in
Localized Imaging and Spectroscopy
Whereas whole body imaging at the highest fields still presents some challenges, high quality images and spectra from localized regions of interest in the head, extremities, and superficial anatomy such as breasts can be easily measured with little complication.
Conclusions
RF artifacts due to extremely short brain and muscle tissue wavelengths of 12 cm at 7 T and 9 cm at 9.4 T can become the increasingly powerful RF shims and gradients used to localize ROIs and to optimize selected criteria therein by new families of RF protocols and feedback driven optimization algorithms. Shorter wavelengths bring the new ability to “steer” RF fields to targeted anatomies and acquisition mechanisms. New RF shimming, localization and optimization techniques can solve many of the RF problems encountered at high field strengths, but will further amplify the SNR benefit already gained.
In one example, the objective is to develop human brain imaging at 9.4 T and body imaging at 7 T. In so doing, groundwork is laid for determining the feasibility of head and body imaging to 11.74 T and 9.4 T respectively. To that end, the physical and technical problems noted previously are solved.
Strategy 1. Head Imaging at 9.4 T and Beyond.
Overview
Advancing human head imaging at 9.4 T and higher fields entails modeling, instrumentation, B1 optimization algorithm development, and validation of these technology and methodology advancements through applications testing.
Modeling
Initial modeling for the purpose of RF coil design and B1 optimization methods development uses the Remcom Inc, Finite Difference Time Domain method to numerically predict RF field magnitude and loss contours in the NLM Visual Human digital atlas of the adult male head.
Coil Design
Modeling is used for optimal coil design. B1 transmit (B1+) uniformity is a concern at the highest magnetic fields. Destructive B1+ interference between complex B1+ vectors from coil elements result in “bright center, dark periphery” patterns in axial images of the brain with various head coils at 7 T. FDTD modeling with the Remcom segmented and meshed human head model (as in
B1 Field
Modeling is used for B1 optimization algorithm simulation. The finite-difference time-domain method is used to model heads and bodies with 5 mm resolution in different locations within head coils at 400 MHz and 500 MHz, and in body coils at 300 MHz and 400 MHz. This is accomplished by calculating the field produced by each tuned element using the FDTD method. Then the results are loaded into a Matlab routine where the phase and magnitude of the driving voltage of each element is varied until a configuration yielding an optimal (defined in various ways) image intensity distribution in a region of interest. Image intensity is calculated as for parallel acquisition, but with full k-space sampling. Image intensity at one location is then calculated as |sin(γτΣNBN+)|ΣN|BN−| where BN+ and BN− are the circularly-polarized components of the RF magnetic field created by element N that rotate with and counter to nuclear precession respectively, γ is the gyromagnetic ratio, τ is the pulse duration, ΣN indicates a summation over the N elements of the coil, and pairs of vertical bars indicate that a modulus is to be taken. For an example of this approach as applied to optimization of homogeneity in the head at 7 T, a 3D model of a head in an 8-element coil of the type given in
For B1 Field “Autoshim”
High speed modeling is developed for B1 optimization automation. The Finite Difference Time Domain (FDTD) method can be used to solve Maxwell's equations for the case of the head loaded TEM coil, and has yielded excellent results. The FDTD method is one method of modeling the B1 fields in the human head and body. However, the computational resources for high resolution, high detail modeling are large, and run times even with a code modified to run on a multiprocessor machine are still very long (hours). Computational speed is especially a concern when the addition of image feedback criteria are contemplated for automated B1 field optimization routines. Consider the process of modeling high fidelity, anatomy loaded coil topologies, especially those driven by dynamic excitation protocols by the FDTD method. One example of a modeling method for high frequency RF solutions for high field MR is presented. One modeling scheme provides a fast solution of Maxwell's equations for the coil-tissue model. This facilitates numerous changes in coil configuration, excitation, or other parameters to rapidly find optimized results. The FDTD technique is iterative and cycles until convergence occurs, defined by the residual errors. This explains the long run times even in multiprocessor machines. For large problems, such as radar cross section calculations which are scattering problems, the FDTD method can be superseded by integral equation methods.
The integral equation method (IEM) can be used for electromagnetic problems. These formulations result in the Electric Field Integral Equation (EFIE), or the Magnetic Field Integral Equation (MFIE), or the Combined Field Integral Equations (CFIE) which is the weighted sum of the first two equations. The formulations are in terms of the Green's function, which is the free space version of the form e−jk
In this case the formulation considers that the human tissue model be subdivided into regions which are piecewise homogeneous. The effect of loss in each region considers that the homogeneous region Green's function be modified. These equations are then solved by the method of moments, leading to a complex matrix equation and subsequent solution. In contrast to the FDTD and FEM methods in this section, the integral equation method considers that only the fields/sources at the interfaces be stored, and fields at interior points are not stored. Thus, computer data storage requirements are much reduced, but matrices are full. Two dimensional problems with this formulation are extended to the present three dimensional coil and anatomy models. A Green's function that accounts for the loss in homogeneous regions of tissue models is generated. The significant increase in solution speed expected with implementation of the integral equation method for MR modeling facilitates the number and complexity of the time dependent, interactive, RF models considered.
Instrumentation
A 9.4 T, 65 cm bore Magnex superconducting magnet with a 40 cm i.d., asymmetric gradient head coil and shim set is used for human head imaging and for porcine models proposed. This magnet system is interfaced to a Varian Inova console, and to a custom RF front end including a new parallel transceiver to drive a number of multi-channel TEM coils as pictured in
RF Subsystem
The parallel transceiver includes a multichannel receiver, and a multichannel transmitter including programmable attenuators, phase shifters, power amplifiers and T/R switches. Demonstration of a transceiver is presented elsewhere in this document (
TABLE A1
Parallel Transceiver Specifications
RF Amp:
Power
500 W/channel
Channels
16
Total Power
8 kW
Freq. Range
30-405 MHz
Phase:
Range
0-358.6°
Resolution
1.4°
Magn:
Range
0-63.75 dB
Resolution
0.25 dB
Max. Table size
~600 steps
Reconfiguration time
66 μs
Digital Receiver speed
64 MHz
Intermediate Freq.
20 MHz
ADC depth
14 bits
Parallel Transceiver
A high speed parallel transceiver is developed and implemented for automated B1 shimming transmit SENSE, and other “on the fly” applications at 9.4 T. The 66 μs command cycle of this first generation transceiver is sufficient for controlling phase and gain settings for interactive RF shimming. But this is too slow for other applications. The present subject matter reduces this time delay to a sub-microsecond control cycle to facilitate Transmit SENSE, automated shimming and other high speed applications using on-the-fly waveform modulation and feedback control. The broadband power amplifiers and other system components also allow for simultaneous or interleaved multi-nuclear applications, swept frequency methods, and other frequency dependent protocols.
One element of an example of the parallel transceiver is a transmitter module including a single chip modulator, and a single Field Effect Transistor (FET) power amp. The compactness of this unit (approximately 1 cm×2 cm×3 cc) allows integration of this unit with its dedicated coil element, which also serves as the heat sink for the FET. Together with a similarly compact PIN diode protected preamp, this parallel transceiver is integrated at the probe head (coil) for improved RF speed, performance and flexibility.
The present subject matter device, includes an adjustable integrated circuit designed to allow for frequency shifting over 100 kHz of bandwidth, phase shifting over 360 degrees, and amplitude adjustment over 30 dB. The circuit architecture includes the blocks shown in the figure. The incoming signal is amplified to a known output value through an automatic gain control (AGC) circuit to maintain optimal operating conditions for the mixer. This signal is mixed with a low frequency signal to adjust the input signal around the center frequency. The low frequency signal is produced by a voltage-controlled oscillator (VCO) controlled through the frequency adjustment line. The output of the mixer is then supplied to the phase adjustment circuitry. The phase of the signal is modified according to the phase control signal from zero degrees to 360 degrees through capacitors that are switched into the system. In addition, the signals' amplitude is adjusted by the system, taking into account the fixed input amplitude supplied via the AGC.
In one example, a bipolar junction transistor (BJT) process such as IBM's 0.35 um SiGe is used because of the linear, high voltage signals. One example of a design is verified in simulation against the system specifications. From the specifications, a system architecture is selected and blocked out. Each circuit block is designed and simulated to the specifications of the architecture and then combined into the full architecture for system simulation. From here, the circuits are laid out, checked, and extracted. Using the extracted decks, the system are re-simulated to ensure that parasitics of the chip are not affecting the system performance. These simulations can be performed with Analog Artist, Spectre, and Spectre RF; layout can be performed with Vituoso XL; and checking can be done with the appropriate checking decks for the process.
RF Coils
The parallel transceiver diagramed in
B1 Optimization
B1 field dependent phenomena associated with high field MRI and the parallel transceiver and multi-channel coil technology are addressed with B1 optimization algorithms. By controlling the currents in individual RF coil elements, in phase, gain, frequency and time, the RF field can be manipulated to optimize signal from a targeted region of interest for SNR, SAR, CNR, homogeneity or other criteria. The RF (B1) shimming is automated, similar to B0 magnetic field shimming. However, unlike magnetic field shimming, RF coil elements in a transmit array are typically not orthogonal; thus constraining the solution set as compared to B0 shimming options. A two step approach is taken to B1 shimming. The first entails mapping the B1+ profile for every coil element within an array; the second step uses the B1+ map to optimize an RF shimming routine for a specific criteria (i.e. SNR, SAR, CNR, homogeneity, etc).
B1+ Mapping Routine
In an array the total B1+ at any given pixel is the summation of all the RF coil's B1+ at that pixel.
Methods for mapping the individual B1+ fields for individual coil elements within an array are too slow for routine use. The results from the routine described below, have shown B1+ mapping can be achieved in approximately one minute. If an array is transmitting the same waveform, but the phase and magnitude for individual coils can be varied, then the individual coil's B1+ can be decomposed into two components. The first component, m, is the complex-valued scalar representing the phase and magnitude of the RF waveform. The second component, B, is a map that describes the relative distribution of the complex valued B1+ at each pixel. FIG. AX may help to visualize the variable B. When the flip angle generated by the B1+ is small, the net B1+ at any point is approximately proportional to the received image pixel, P. Now equation [8] can be restated as
or, in matrix notation
P=mB [10]
To map the relative distribution of the B1+ fields for each coil, this equation must be solved for B. A least squares solution is possible by collecting several different types of images; these images must have different, non-degenerate sets of m. In this example, m is varied by adding 180° to the transmit phase to one channel at a time. Using linear algebra reductions, m for a n-element RF coil using this pattern is most simply expressed as
Correspondingly, P is enlarged to reflect the data from the additional images, and B is solved for in this system of equations by a linear least squares formulation.
Mapping is evaluated by varying m and collecting a global B1+ map with a two flip-angle method. This map is then compared with a simulated map produced by |m×B|. The two maps should be proportional to one another, within ±5%. These trial FLASH data sets were acquired with a resolution of 256×256, but tests have shown that resolutions as low as 32×32 are sufficient for mapping.
B1 Shimming Routine
The B1 shimming algorithm described below is designed to optimize RF homogeneity. The algorithm is general however, and can be used to optimize other distributions by changing the criteria or targets. A linear solution may not provide an optimum result. If using equation X+2, with B known, then P becomes the target distribution and is set to unity. This creates a situation where the magnitude is set to unity and the phase is implicitly set to zero (since the first grayscale image is used as a reference, its phase is defined as zero everywhere). Therefore, the linear solution finds a compromise between two opposing goals: find a homogeneous magnitude, and keep the same global phase distribution as the original image.
In one example, a non-linear algorithm is used. In this algorithm, a penalty function is created that returns a single value describing the appropriateness of the result for each m. One solution will be the values of m that return the lowest penalty value. Because each element has two variables, phase and magnitude, there are 2*N degrees of freedom. To ensure the solution is the global and not a local minimum, the optimization routine is restarted along several values of m. For each starting point the routine follows the path of steepest decent to the local minimum. The lowest local minimum is the global minimum. Although the search space is large, the efficient global search routine finds the global solution in about one minute for an eight channel coil as described in section C.
where An and φn represent the nth element's current magnitude and phase respectively, ω is the Larmor Frequency, and the term jbnx anxe describes the remainder of the steady state geometrical phase shift and magnitude relation between the nth drive and the x direction B1 magnitude at a given point in the phantom. A similar equation can be written for the y component of the B1 field. The final x and y directional B field may then be calculated as the sum over n of B1xn. The overall field is found from these two quantities. Using the finite element simulations, the quantities jbnx anxe and jby anye are found for each point on a grid in the phantom placed on the selected field of view and for each current element on the coil. With this information, field distributions were calculated from arbitrary elemental current magnitudes and phase angles. The final step was to determine each An and φn to obtain a distribution approximating a desired distribution. To accomplish this a cost function was defined for a fit to the desired distribution and then an iterative optimization algorithm was applied. Such an approach was implemented and results obtained by using the simulated annealing approach. Simulation results in the figure show a convergence to an arbitrarily chosen, off axis B1 contour reached through current element phase and magnitude control.
Verification of this RF localization method is in two forms. With drive parameters determined by the B1 field distribution tailoring technique described above, verification by means of additional finite element simulation is performed. Drive amplitude and phase values is fed back to verify the expected distributions. A coil and a phantom are used in one example, Target distributions can be selected and proximity of match is reported at 9.4 T.
The simulated annealing optimization technique yields relatively slow convergence and a suboptimal solution to the problem of tailoring an RF field distribution to a desired shape. To understand the difficulty in optimizing a field distribution within the load object, begin by imagining a three dimensional mountainous terrain distribution where the height represents the field intensity within the slice of interest caused by each drive element. In terms of the above problem this field is represented mathematically as
Anej(ωt+φ
where Zn(x) represents a complex function. Changing the amplitude of the nth drive element varies the amplitude of the distribution and varying the phase morphs the distribution into other shapes. The final field intensity is then the magnitude of the sum of the distributions caused by each elemental drive,
The goal of the optimization is to form a distribution into a desired shape by morphing the distributions into shapes that sum to a localized distribution of interest. This is stated in an optimization program as
where D(x) is the desired distribution. In general this type of optimization is difficult.
To solve this problem, various solutions are considered, including the following two. These approaches fall within the field of convex optimization. The first is to define an upper bounding function that is convex. That is to say, define a function that is at all places less optimal than the actual distribution for all phases and amplitudes but itself has no local minima. Because there are no local minima, efficient gradient decent techniques can be used to find the most optimal solution in the space of distributions. Using the optimum of the bounding function as a starting point, traditional optimization techniques can then be used to improve on the actual solution.
Another example is to define an approximating function that roughly follows the shape of the original distribution but also has no local minima in the 32 (16 elements*2 degrees of freedom) dimensional space. By guaranteeing no local minima, algorithms can be employed to optimize the solution. This approximate solution can then be used as a starting point for efficient minimization of the original problem. These techniques can yield solutions to these optimization problems in real time, with benefits for MRI and MRS applications.
Head (and Body) Imaging
Human studies are considered for testing, validating, characterizing, improving, and determining the practical limits for head and body imaging at the highest fields currently available. The RF dependent parameters (SNR, SAR, B1, temperature, etc.) are measured and mapped for head and body imaging initially at 7 T and 9.4 T respectively.
Strategy 2. Body Imaging at 7 T to 9.4 T
Overview
The methods for exploring 7 T and higher field body imaging require high frequency modeling, hardware development, and data acquisition similar to those for head imaging at still higher frequencies. Because some of the methods and technologies considered and discussed for the head are used for highest field body imaging as well, only body specific components of the plan are discussed below.
Modeling
Initial modeling of highest field human body applications used the FDTD method to numerically predict RF field equipotentials in the National Library of Medicine's Visual Human digital atlas with segmented anatomy adjusted to conductivity and permittivity of 300 MHz. To lay the groundwork for higher field imaging, 400 MHz models are performed as well.
Instrumentation
As with 9.4 T head imaging, the equipment and sequences specific to 7 T human body imaging are not commercially available. Therefore the hardware, controls and pulse protocols required for this study are developed. The MRI system used is a Magnex 7 T, 90 cm bore magnet with Magnex whole body gradients and shims (i.d.=63 cm). A Varian (Palo Alto, Calif.) Inova console is used, together with an 8 kW solid state RF power amplifier from Communications Power Corporation (Hauppauge, N.Y.). An actively detuned, 300 MHz RF body coil of the model dimensions was built together with its PIN detuning circuits and system control interface as shown in
Body Coil
A 7 T body coil of TEM design is driven at every element by a parallel transceiver as for head imaging at 9.4 T. The multi-channel 32-element TEM body coil has the same control functions and flexibility as for the multi-channel TEM head coils. Before making NMR measurements body coil performance is first verified by bench measurements. Unloaded and loaded conditions are verified to 1) tune to the correct Larmor frequency, 2) match to 50 ohms at each element, 3) meet free space (unloaded) homogeneity specifications above, 4) generate an efficient B1 gain over the FOV, 5) withstand a peak power rating of 16 kW (32×500 W) for body coils, (8 kW(16×500 W) for head coils at 5% duty), 6) withstand an average power of 600 W for body coil, and 200 W for head coil, 7) sustain 10,000 WVDC isolation between coil conductive elements and human tissue, 8) actively switch between tuned and detuned states and vice-versa within 10 μsec., 9) decouple from local receiver coils by >40 dB, 10) have >25 dB isolation between quadrature modes, 11) meet all electrical power, control, RF, and mechanical interface specifications for host MRI system, and 12) include fully functional RF power monitoring, fault detection, and failsafe feedback circuits.
Receiver Coils
Specialty receiver coils as pictured in
Body Imaging
The models predicted significant RF artifacts with at least one sharp line running longitudinally through the body, primarily due to destructive interference of the short (12 cm) wavelengths in the high water content tissue dielectrics at 300 MHz. Imaging proved these predictions generally correct. See
Observed artifacts and RF power requirements do not appear to preclude the possibility of useful whole body imaging for at least some applications at 7 T. Improvement in both signal and homogeneity appear to be gained by using the body coil for excitation together with separate receiver coils for higher sensitivity, local image signal reception. Breast imaging and spectroscopy appears to be an early, obvious application of 7 T body imaging. See
B1 Shimming
As with head imaging at 9.4 T, B1 shimming with the proposed 32-channel, multi-nuclear body coil can be useful in correcting for RF related artifacts and optimized criteria selection in targeted regions. An example of the potential utility of B1 shimming is revisited in
Strategy 3. RF Safety.
Temperature Calculation, Measurement
RF heating can be calculated and measured in porcine models of adult human weight to identify the RF power level limit that can be safely transmitted from the head and body transmit coils tested. The FDA has specified guidelines for RF energy dosimetry to the human body in two ways. 1) Average and local specific absorption rates (SAR) are discussed above and 2) RF power deposition must be insufficient to cause a core temperature increase in excess of 1° C. and localized heating to greater than 38° C. in the head, 39° C. in the torso and 40° C. in the extremities. Equating the power loss density [Eq. 6 of this section] to temperature through the bioheat equation [Eq. 7 of this section], temperature contours are calculated for 3D head and body models. Contours are calculated for the NLM human head and body for each RF head and body transmit coil used, for each frequency. The result can be xy, xz, and yz temperature maps for each coil configuration, for each frequency 300, 400, 500 MHz.
Because the thermal criteria most directly target the safety issue (localized thermal tissue damage), because local temperature can be directly measured, and because temperate contours are more extreme due to shorter wavelength losses in tissue at higher frequencies, it is prudent to validate temperature models by direct temperature measurement. Temperature can be measured on the xy, xz, and yz planes modeled, and also adjacent to receiver coils placed on the porcine model within the head and body coils, and rectally. In this way any power coupled to receiver loops is monitored for local heating, in addition to core heating measured.
To secure a minimum of N=3 measurements per study, six, 120 lb. farm pigs can be used to test one head or body coil with and without separate receiver coils per year to assure RF safety with all coils, coil combinations and field strengths planned. A total of thirty pigs are required for the five year study (with a half dozen spare). To most accurately simulate the RF portion of the NMR experiment, anesthetized pigs are placed on a patient table bridge, inside the head or body coil. The receiver coil array is placed about the pig's head or chest depending on the coil. The receiver coil can be detuned. RF power is set to the SAR guideline limit for humans. Temperature vs. time is measured for multiple locations inside the pig model until the temperature reaches steady state, or until the FDA temperature limit is reached locally or globally. The head or torso is instrumented with seven temperature probes from the eight channel Luxtron thermometer to be used. The eighth channel probe remains in a thermometer referenced control cell (water bath) outside the coil. For transmit power input from the coil, temperature is measured in the anesthetized pig's head or chest.
The present subject matter seeks to identify the location and magnitude of “hot spots”, then to correlate the RF coil input power with hard to detect, RF coil specific heating patterns and limits in tissues.
A more detailed explanation of the methods is provided below. Fluoroptic probes from a Luxtron 3000 unit can be inserted via an 18 gage hypodermic needle, into the drilled head or chest wall muscle. A linear probe with transducers 1-4 spaced at 1 cm intervals can be positioned in the brain, or subcutaneous fat and intercostal muscle, centered beneath the receiver coil, to a depth of 0 to 4 cm. Two more single transducer probes can be inserted into the scalp, or chest wall immediately beneath the receiver coil elements to measure local heating adjacent to the coil conductors. A seventh transducer can be inserted rectally to monitor the core temperature of the animal, outside the body coil. The eighth remaining probe is dedicated to a water calibration bath apart from the animal measurements. The heart rate, h.r., can be monitored. B1 for each power setting and frequency can be measured at the head or chest wall with a field probe, after the animal is sacrificed. The coil input power P can be determined by MRI system means provided using a four port directional coupler and a power meter. The line losses to the coil can be calibrated, as can any reflected power. Past experience with surface coils indicates that each heating study from resting state to steady state heating, and back to resting state, takes about 4 hours. Two studies per pig can require a minimum of 8 hours each.
The predicted sources of error from the experiments above can be attributed to the anesthetically suppressed thermoregulatory reflexes of the pigs. Resulting errors can be conservative (safe), predicting slightly more heating than will occur in non-anaesthetized, healthy humans. The body of data can tablulate a priori information correlating RF power input to maximum local temperature contours for the different coil combinations, applications, and frequencies studies. Also in the data are the heating rates for these parametric variations.
9.4 T Human Imaging: Preliminary Results
Objective: To investigate the feasibility of human head imaging at 9.4 T.
Introduction: Brain imaging has been increasingly used in research applications at 7 T and 8 T since 1999. This study examined whether the human brain can be imaged safely and successfully at still higher field strengths.
Methods: The methods consisted of modeling, hardware development and data acquisition.
Modeling: Initial modeling used the Remcom Finite Difference Time Domain method to numerically predict RF (B1) field magnitude and loss contours in the Visual Human digital atlas of the adult male head.
Hardware Development: A 9.4 T, 65 cm bore Magnex super-conducting magnet (
Measurement: For initial results, an eight element coil was used (
Results: With IDE and IRB approval, and in compliance with FDA guidelines, NMR data and interviews were obtained from an initial 10 healthy normal volunteers; 14 males, and 4 females. Of the 9.4 T imaging experience, 10 subjects reported “sleepiness”, 8 reported “light headedness or dizziness”, 4 reported metallic taste, two got cold, one got warm, and one reported “body numbness”. There were no requests to conclude studies early with some lasting two hours, and no other remarkable incidents or long-term effects reported.
Discussion and Conclusions: Human MR imaging to field strengths of 9.4 T can be accomplished, according to these results. The Larmor wavelength in the human tissue dielectric at 400 MHz is approximately 9 cm. By conventional methods and thinking, this wavelength would preclude any possibility of achieving safe and successful human scale imaging. RF interference patterns from a conventional, uniform field volume coil would create severe inhomogeneities in the anatomy, as shown in
9.4 T Human MRI: Initial Results
The following discloses human images generated using a high field of 9.4 T. A 65 cm bore magnet was used together with an asymmetric 38 cm i.d. head gradient and shims set. A multi-channel transmission line (TEM) head coil was driven by a programmable RF front-end to control the relative phase and magnitude of each channel independently. The RF field control methods facilitated compensation of RF artifacts attributed to destructive interference patterns in order to achieve homogeneous 9.4 T head images, or to localize anatomic targets. Preliminary RF safety studies were performed on porcine models.
Introduction and Background
Images at 9.4 T can be obtained using a large male cynomolgus macaque. The SNR at 9.4 T more than doubled the SNR at 4 T. RF uniformity and SAR for the monkey study were high and low respectively. The technology and methodology can be used for achieving successful 9.4 T results in larger animals. Based on these preliminary results and the need to double the SNR and spectral resolution in laboratory primates, from the 4.7 T field strength, analysis was conducted using 65 cm bore, 9.4 T system.
The macaque head measures approximately 1/7 the volume of an adult human head. While SNR and spectral resolution predict human results, the uniform RF field contours in the monkey head at 9.4 T do not indicate the B1 field in the human head of significantly greater electrical dimension. Predictions of human images at 400 MHz Larmor frequencies can be extrapolated upward from 7 T data, downward from 11.1 T images of fixed brains ex-vivo, or directly from head coil bench studies and numerical models at 400 MHz. Data from all four sources can be used to predict RF artifacts for images achieved by the conventional means of placing a human head inside a homogeneous, circularly polarized volume coil resonant at 400 MHz. These high frequency RF artifacts can be described as dielectric resonance. The artifacts can relate to regions of signal deficit in a brain image due to regions of destructive interference enhanced and exacerbated by the increasingly shortened wavelength of the incident RF (B1) field in the brain dielectric. Means for correcting these RF artifacts have been considered, including adiabatic excitation, dielectric lensing, and B1 shimming.
Materials and Methods
The 9.4 T magnet is installed in a facility designed to house it. Surrounding the magnet is a rectangular box shield comprised of 350 tons of welded plate steel 8″ thick at the magnet center and tapering to 1″ thick at the end plates. The end plates include a window and a door on the system user end, balancing a brass plate bulk head (patch panel) access on the opposite end. Also enclosing the magnet and incorporating the steel magnet shield is an RF enclosure (Lindgren RF Enclosures, Inc., Glendale Heights, Ill.) to isolate the magnet environment from outside signals by 110 dB at 400 MHz. Within the magnet are an asymmetric head gradient and shim set, also of Magnex design and manufacture. Subject access to the magnet is provided by a motor driven, cantilevered table (High Field NMR Systems, Inc, Birmingham, Ala.). Interfacing the magnet is a custom RF system featuring a multi-channel, parallel transceiver system of in-house design. Sixteen 500 W RF power amplifiers are used for one example of the system. Controlling and processing the system data acquisition functions is a console. Specifications of the magnet, gradients, shims, and RF systems follow.
Magnet Subsystem
The 9.4/650 magnet (Magnex model MRBR 9.4 T/650) shown in
Cryostat
The cryostat includes a central all-welded stainless steel helium vessel which is surrounded by two aluminum gas-cooled radiation shields. The cryostat measures 3.15 m long and 3.48 meters high with a 0.65 meter clear bore. It weighs 30 tons without cryogens. The complete assembly is contained in a stainless-steel outer vacuum vessel with a vertical service turret located centrally on top of the cryostat. The turret provides access to the helium reservoir for the demountable magnet leads, helium level probe, and helium transfer siphon. The system is equipped with two Leybold model 5100 2-stage cryocoolers with CP6000 compressor units. The helium reservoir contains 2500 litres of liquid helium with 1600 liters volume above the refill level. The liquid helium evaporation rate is 0.2 liters/hour, and the refill interval is six months.
Gradients and Shims
The 9.4 T, 65 cm bore magnet is fitted with an asymmetric, torque compensated, self shielded “head” gradient and shim set (Magnex model: SGRAD III 640/400/SP). The dimensions include 160 cm length, 64 cm o.d., and 40 cm i.d. with the isocenter displaced 18 cm from the utility end to give ample access for human head and laboratory primate imaging and spectroscopy. The peak operating voltage is 1 kV and the peak current is 500 amps with 180 A RMS. At 80% peak, the strength and 98% rise-time for X, Y, and Z gradients are 40 mT/m and 150 μs respectively. The water cooled gradients can dissipate 10 kW, ΔT=10K at a flow rate of 15 liters/min. To limit the vibration, the space between inner and outer gradients is filled with a “concrete” of epoxy resin and alumina aggregate. The gradient set includes an actively shielded Z0 shim for B0 correction, a set of high power, second order resistive shims, and a 24 tray passive shim system.
Console
A Varian console is interfaced to the 9.4 T magnet system and used to generate the data presented herein. The console configuration includes a Sun Blade 2500 as a host workstation to control the transmit excite signal and data acquisition through single proton frequency transmit and receive channels respectively. (Sun Microsystems, Inc., Santa Clara, Calif.). The console is configured to transmit and receive on 8 channels or 16 channels. Each of the multiple receiver channels is mixed to an intermediate frequency (IF) signal of 20 MHz and then oversampled at 64 MHz by 14 bit ADCs on Echotek ECDR-814 digital receiver boards for an effective 20 bits after filtering and decimation to a spectral width of 10 kHz. (Echotek, Inc., Huntsville Ala.). The digital receiver boards are housed in a VME64x card cage chassis with a Motorola MVME5500 single-board computer (Motorola Embedded Communications Computing, Tempe, Ariz.) as the bus master that runs control software under the VxWorks operating systems (Wind River, Alameda, Calif.). A second Sun Blade 2500 workstation was added to increase the data handling capacity to process and store the data from the multiple receive channels. Active diode protected, decoupling preamplifiers were used for the multi-channel TEM coil interface.
The console's single proton frequency transmit channel is split into multiple, parallel paths at the 1 mW level. Programmable phase shifters and attenuators are included in each transmit path to provide for console controlled phase and magnitude control for each channel. Each channel is then amplified by 500 W broadband amplifiers and independently switched in time by channel dedicated TR switches. Other details of the RF parallel transceiver system and applications follow.
RF Methods
Modeling
RF related artifacts are noted with human imaging at 7 T. At 9.4 T, XFDTD models of a human head inside a conventional inductively driven, circularly polarized TEM head resonator are constructed. The coil was of clinical standard 27.5 cm i.d. and generates a uniform B1 field in free space. When the NLM Visual Human male head was added to the model, the B1 field uniformity across the head was degraded as shown in the three center planes of the 3D model below. The patterns of non-uniformity are similar to those seen in 7 T, but more severe. These exploratory models and other data indicate that coils of conventional circularly polarized, uniform field generating volume coil designs as used at lower frequencies will be sub-optimal for use at 9.4 T. Other coil designs and applications methods are available.
Parallel Transceiver
Adjusting the relative magnitude and phase of independent current elements in a coil structure can be used to compensate for high field, RF artifacts by means described as B1 shimming. Controlling the phase angle and magnitude of the RF signal on separate coil elements can be accomplished using manual means such as trimming coaxial cable lengths and adding coax attenuators. These mechanical approaches to coil field control are laborious, time intensive, and do not lend themselves to feedback driven algorithms for interactive and automatic image optimization. To solve these problems an electronic, computer controlled parallel transceiver is used for the RF front end of the 9.4 T system.
The parallel transceiver schematized in
RF Coils
Transmission line (TEM) coils can be used for high frequency RF coils. Independent TEM elements can be driven, received, and controlled with the parallel transceiver described above. Both linear and loop transmission line elements can be used with 400 MHz head coil construction. The linear element generates a flux vector typically parallel to the sample surface, and the loop generates a perpendicular vector. The linear element is shorter for a given coil length, and therefore electrically more efficient at high frequencies. The coil comprised of linear elements also generates a more uniform field, especially in the periphery of its volume. The current phase angles on the linear array mimic those of other TEM or birdcage coils, incrementally phased over 360 degrees to generate a theoretically uniform transverse field through the coil volume. In the loop array however, currents in adjacent element legs are in approximate phase opposition, resulting in severe phase gradients in the RF field near the elements of the coil. Accordingly, a TEM coil with linear elements demonstrated certain properties for the 9.4 T head coil design.
While RF non-uniformities can be actively controlled by methods described thus far, there are a number of considerations for the passive coil design. For the multi-channel TEM head coil design parameters include the length, i.d., o.d. and shape of the coil, the number, dimensions, and spacing of the TEM elements, the choice of dielectric material properties and dimension, the relative degree of coupling or decoupling between the elements, and the positioning of the head within the coil. The coil model of
RF Safety
9.4 T is above the 8 T and one consideration is RF heating at 400 MHz. The limit for RF induced temperature elevation vs. RF input power vs. time was analyzed in a porcine model of the human head In vivo. For power settings of 12-20 Watts, temperature vs. time was directly measured by fluoroptic thermometry, and recorded for human adult sized farm pigs (50-100 kg). Anesthetized pigs were instrumented with fluoroptic probes inserted through burr holes over an array of model guided positions in the porcine scalp and brain. The pig was then fitted with a head coil of the design and dimension to be used in human studies. RF power was then applied at continuous wave (CW) power level increments starting at 12 W. The pigs' heads weighed approximately 4-6 kg, similar to human head and necks. Temperature vs. time was then recorded until the FDA guideline limit of 1° C. above core temperature was reached. See
RF Shimming
Data for programmable B1 phase and magnitude control are shown in
RF Localization
As modeled in
The 16 element, multi-channel TEM head coil of
Results and Discussion
B1 and B0 bio effects were monitored in terms of subject safety. For a given RF excitation sequence, increased energy is lost to the tissue conductor and dielectric due to the very short Larmor wavelength at 400 MHz, compared to lower field strengths. A consequence of this energy loss is RF heating, a potential safety concern. While no safety concerns were interpreted from subchronic exposure of rats to 9.4 T, the use of 9.4 T for human studies compels close observation of subject experience from exposure to this field. Results from initial animal models and human studies follow.
RF Heating
The FDA maximum temperature guideline limit of 1 μC. above core temperature is first reached in the brain where calculations have shown SAR and ΔT to be minimum. See
Maximum absolute temperatures are measured in the brain with SAR and ΔT values modeled and measured to be minimum in the brain compared to the scalp. Although maximum temperature change might occur in the scalp, the scalp is well perfused and adjacent to an infinite, 22° C. ambient heat sink. The healthy normal thermal profile of a head, human or porcine, peaks in the brain which is maintained at core body temperature. When energy is added from a large surrounding head coil at a low rate such as in an NMR study, the subject's core temperature and thermal profile across a head is raised as if a fever. In a well perfused subject the absolute temperature first reaches the FDA thermal limit in the brain as opposed to the scalp. This response assumes low power input to the head from a uniform incident field inducing a heat load distributed and dissipated by healthy, normal, thermoregulatory reflexes.
9.4 T Experience
Imaging
Using the parallel transceiver and RF coils of
RF artifacts due to short brain and muscle tissue wavelengths of 7 T (12 cm) and 9.4 T (9 cm) MRI correspond to increasingly powerful RF shims and gradients used to localize ROIs and to optimize selected criteria therein by various RF protocols and feedback driven optimization algorithms. Shorter wavelengths allow steering of RF fields to targeted anatomies and acquisition mechanisms. A TEM coil is independently tuned (as opposed to monolithic birdcage resonators) becomes a solution for B1 field control because every element can be tuned to manipulate the B1 field component of each with high sensitivity. Implementation of phase and magnitude control is served by a multi-channel TEM coil interfaced to the parallel transceiver reported.
Conclusions
Preliminary results from initial 9.4 T studies indicate successful human head imaging at this field strength. A 9.4 T superconducting magnet of human bore size was developed. A parallel transceiver front end was used to give console control to new methods of RF shimming and parallel imaging. This transceiver is employed to the development of feedback driven, interactive and automated ROI targeting and image optimization algorithms. 9.4 T Larmor wavelengths significantly shorter than anatomic dimensions allow image localization and optimization.
A Parallel Transceiver for Human Imaging at 9.4 T
Objective: The present subject matter includes a parallel transceiver for automated B1 shimming and other applications at high field strength, including for example, 9.4 T.
Introduction: In ultra-high field MRI, the incident RF transmit field becomes non-uniform in human anatomy due to short wavelength interference patterns in tissues. The consequential image inhomogeneities can be at least partly compensated by B1 shimming. This is a process by which the RF field is actively controlled by adjusting the relative impedances (magnitude and phase) of independent elements in a coil structure. Control of the phase angle and magnitude of an RF signal can be conducted by RF shimming using manual means such as trimming coaxial cable lengths and adding coax attenuators. These mechanical approaches to coil field control are laborious, time intensive, and do not lend themselves to feedback driven algorithms for interactive and automatic image optimization.
The present subject matter includes an electronic, computer controlled parallel transceiver for an RF front end of a 9.4 T system for human studies and other applications.
Design: The parallel transceiver schematized in
Results and Discussion: The parallel transceiver performance is shown in Table D1. Human head 9.4 T images acquired with this parallel transceiver are presented elsewhere. In one example, the present subject matter includes four degrees of freedom in RF control at the front end of the MRI system: phase, gain, frequency, and time. The 66 microsecond command cycle of this transceiver allows controlling phase and gain settings for interactive RF shimming. In one example, the system reduces time delay to four microseconds to facilitate Transmit SENSE, automated shimming and other high speed applications requiring on-the-fly waveform modulation and feedback control. The broadband power amplifiers and other system components allow for simultaneous or interleaved multi-nuclear applications, swept frequency methods, and other frequency dependent protocols.
Conclusion: The parallel transceiver of the present subject matter allows magnetic resonance operations using shorter Larmor wavelengths of ultra-high field MRI. In one example, with four new degrees of freedom at the RF front end, new algorithm and feedback driven protocols for high field imaging methods are used.
TABLE 1
Parallel Transceiver Specifications
RF Amp:
Power
500 W/channel
Channels
16
Total Power
8 kW
Freq. Range
30-405 MHz
Phase:
Range
0-358.6°
Resolution
1.4°
Magn:
Range
0-63.75 dB
Resolution
0.25 dB
Max. Table size
~600 steps
Reconfiguration time
66 μs
Digital Receiver speed
64 MHz
Intermediate Freq.
20 MHz
ADC depth
14 bits
Ultra High Field MRI: High Frequency Coils
Abstract
The following is directed to RF volume, array, and surface coil modeling, design, construction, control, safety and human In vivo application examples for field strengths from 4 T to 9.4 T. While a variety of coils are included, some portions focus on the transmission line (TEM) technology head, body, surface, and array coils.
Introduction
High frequency coils are used with high field magnets for high signal MRI. High signal clinical MRI can be performed with magnets of 3 T (tesla) field strength. Magnets of 7 T, 8 T, 9.4 T and 11.74 T or greater field strengths can be used for research. Radio frequency (RF) coils are used to inductively couple to human anatomy at Larmor frequencies of 500 MHz (or more or less) to make these magnets useful. RF coil efficiency is inversely proportional to wavelength, both the wavelength of the coil circuit itself and the wavelength(s) of the media (air, plastic, anatomy, etc) through which the RF field propagates. To improve coil efficiency by minimizing its electrical wavelength dimension (λ), the physical dimension of the coil can be reduced, and/or the coil circuit elements can be more highly distributed. Practical considerations such as the image field of view (FOV), field homogeneity, count and control of transmit elements, count and control of receive elements, and RF safety features are applications dependent. Surface coils can be designed for more localized and topical applications such as imaging a superficial organ or tissue. By contrast, volume coils can be used to encompass a head or body for global imaging. Surface and volume coils can be tuned to two or more Larmor frequencies for multi-nuclear, metabolic imaging. Volume coils can be used to transmit a uniform excitation field, in combination with surface coils applied for receiving the excited NMR signal from a local region of interest with high sensitivity. Global FOV coverage and local sensitivity can be acquired by receiving with arrays of surface coils. Parallel reception from surface coil arrays can achieve imaging speed benefit as well. Multi-channel control over the coil transmit functions can be used to interactively or automatically control the nuclear excitation field in magnitude, phase, time, and/or frequency to effect a number of imaging optimization criteria.
RF Coil Losses
Increased signal-to-noise (SNR) and its dependent spatial resolution, temporal resolution and contrast, are among the chief reasons for performing NMR studies at higher magnetic field strengths. In theory the SNR increases to the power of 7/4 with field strength for “ideal” lossless samples. However, for biomedical MRI of lossy tissue, RF dependent losses, defined as energy dissipated or signal reduced, diminish the theoretical SNR ˜B1/√R, where B1 is the strength of the RF magnetic field generated by a unit current in the coil, and R is the effective resistance of the coil and load. In human studies SNR is thus reduced by decreasing the numerator while increasing the denominator in RF coils and in human tissues. Additionally, much of the coil generated RF energy is dissipated in the tissues as thermal energy. The potential safety risk of excessive heating is increasingly present at higher fields. One objective in high field coil design therefore is to reduce the RF tissue and circuit losses.
Increased SNR can be achieved by making NMR measurements at higher magnitude B0 magnetic fields. However, magnetic field strength is only one of several parameters affecting the NMR signal-to-noise ratio. RF coil and tissue losses can limit the potential SNR gains to be realized at higher field strengths. SNR and transmit efficiency will suffer when any one of the parameters in the following proportionality is not optimized.
SNRα(ω2B1)/√(RΩ+Rr+Rtissue) [1]
RΩ, Rr, and Rtissue are the coil's ohmic resistance, radiation resistance, and coupled tissue losses respectively.
RF losses in the NMR experiment can be categorized as coil losses, and as tissue losses Rtissue. Coil losses can be further divided into resistive losses to the coil circuit RΩ, and radiative Rr losses to the coil's field. Tissue losses result from B1 field induced eddy current losses to the tissue conductor, and radiated E field displacement current losses to the tissue dielectric. The physical nature of these RF losses is presented in more detail following.
At higher frequencies the phase change due to the finite propagation velocity of transmit and receive signals on coil conductors is no longer negligible. As coils approach wavelength dimensions, the performance of the conventional lumped element designs succumb to 1) non uniform current distributions resulting in decreased homogeneity, decreased fill factor, and increased radiated field losses, 2) increased ohmic losses in the coil circuit inductors due to decreased 1/e skin depths, 3) increased dielectric losses in the coil circuit capacitors due to decreased conductance efficiency compounded by an increased requirement for series capacitance to achieve high frequency resonance, and 4) self resonance below the desired frequency of operation. An RF coil conserves energy in its circuit and its surrounding field. A typical coil loses some of its energy at the average rate of Ploss, the power dissipated to ohmic resistance R106 of the coil circuit, and to radiation resistance Rr of the coil field. The peak RF current I in the coil loss equation [2] (in this section) below is measured at the coil's terminals. The equation applies reciprocally to transmit and receive conditions for the resonant coil whose reactance is tuned to zero at the desired Larmor operating frequency. The “Q” factor, used as a figure of merit for a resonant coil's efficiency, is defined by the ratio of the average energy conserved over the energy lost per cycle to RΩ+Rr. These two RF coil loss sources together with their impact on the NMR experiment is presented below.
Ploss=(I2/2)(RΩ+Rr) [2]
Coil Resistive Losses
The ohmic resistance RΩ in the coil circuit is the sum of the real resistance of the coil conductor (inductor) RL and the real capacitor loss RC. The RΩ dissipates RF power PΩ loss as heat in the coil circuit. Because RF currents tend to flow near the surface of a conductor, RΩL is dependent on the resistance Rs of a coil conductor's surface area. For an arbitrary coil loop with wire length (, cross sectional perimeter wp, conductivity σ, and permeability μ, the ohmic resistance at the resonant frequency ω, is given by,
PΩ loss =(I2/2)RΩ [3]
RΩ=RL+RC [4]
RL=Rs((/wp) [5]
for, Rs=√(ωμ/2σ)=1/σδ [6]
RC=1/(ωQC)
The skin depth δ is defined as the depth below the conductor surface at which the current density has decreased 1 neper below its surface value. By the relationships, RΩ losses increase with √ω, and ω, for conductor and capacitor losses respectively. The consequences of this loss source for the NMR signal are a coil circuit wavelength (frequency and size) dependent erosion of both the RF transmit efficiency and received signal-to-noise. Ohmic losses of a head or body coil can be reduced by increasing the skin (surface) area and decreasing the inductance of the coil conductors. In a distributed circuit (vs. lumped element circuit), this resistance will contribute a diminishing fraction of the total loss accumulating with increasing field strength.
Coil Radiation Losses
The RF energy loss to the radiated field of the coil is termed “radiation resistance”. This field loss is to the coil environment including loss to the human subject, the magnet bore conductors, and to free space. The radiated power can be broadly defined as the difference between the input power Pin to the coil and the ohmic losses in the coil.
Pr loss=Pin−(I2/2)RΩ [7]
In equation [10] in this section the radiation resistance and associated losses of an NMR coil are proportional to ω4 and to S2. For high field, clinically dimensioned NMR coils, ω4 and S2 respectively can increase the coil's Pr loss. At higher NMR frequencies, clinical sized RF circuits begin to behave less like energy conserving coils, and more like energy radiating antennas. Decreased RF transmit power efficiency, increased RF power absorption in the human subject, and decreased signal-to-noise of highly radiative coils can potentially limit the efficacy of high field human NMR experiments. Problematic RF radiation resistance can potentially be decreased by TEM coil design. Such designs apply Transverse ElectroMagnetic (TEM) theory, also referred to as transmission line theory, to reduce high frequency radiation losses.
Coil-Tissue Losses
RF losses in the tissue loading the coil become even more significant at higher frequencies. Loss to the tissue conductor increases with frequency due to a proportional increase in tissue conductivity and resulting eddy currents. Loss to the tissue dielectric increases with frequency due to a proportional increase in displacement currents. The Larmor wavelength becomes significantly shorter in the tissue dielectric as well. At 128 MHz (3 T) the wavelength in high water content anatomy is on the order of 27 cm. At 300 MHz (7 T) this wavelength decreases to 12 cm. Consequently a homogeneous RF coil can no longer generate a uniform B1 field in the head or body at higher the higher magnet field strengths. The high frequency coil's field propagation and losses in its human tissue load are described by the time-dependent field equations. A time-harmonic magnetic field B1/μ in a lossy, anisotropic, inhomogeneous coil-tissue system can be described by the differential form of the Maxwell-Ampere Law:
∇XB1/μ=Jc+=∂D/∂t [11]
By Ohm's Law the current density Jc=σE, and by Euler's Law the electric field displacement ∂D/∂t=∂∈E/∂t=jω∈E so that equation [9] in this section can be rewritten as:
∇XB1/μ=(σ+jω∈)E [12]
The complex value of E can be written in terms of the magnetic vector potential A, and the electric scalar potential φ, such that:
∇XB1/μ=(Je+Jd)=(σ+jω∈)(−jωA−∇φ) [13]
Influencing the B1 distribution and loss in human tissues loading the coil are the B1 field induced eddy current density Je,
Je=σE=−jωA [14]
and the accompanying electric field displacement current density Jd for tissue specific values of σ and ∈.
Jd=−jω∈E=ω2∈A [15]
The sum and distribution of RF eddy current losses to the tissue conductor, and RF displacement current losses to the tissue dielectric, are determined by the total power loss density Pt loss,
Pt loss=0.5|v(J·J*/σ)dv for J=Je+Jd [16]
Note that the displacement current density Jd increases with ω2 compared to the eddy current density's Je linear proportionality with frequency. This is consistent with the observations of interfering wave phenomena (commonly though inaccurately referred to as dielectric resonance) for head images at high B0 fields. In contrast to early concerns that Je would limit B1 penetration and lead to excessive power absorption, Jd related inhomogeneities and losses may dominate at fields of 3 T and higher in the human head and body, leading to B1, SNR, loss and heating patterns not predicted by Biot Savart, quasi-static, or otherwise over simplified models.
Coil-Tissue Heating
As noted, sub-optimal SNR is not the only consequence of lossy coils. RF loss induced temperature contours in the human anatomy can be calculated by equating the total losses Pt loss from the coil's RF field of equation [16] in this section to the bio-heat (perfusion) equation below. Table values can be used to supply tissue specific thermal parameters for perfusion rate R, tissue and blood densities ρ and ρb, blood temperature Tb, blood specific heat cb, and tissue thermal conductivity K. (28-30)
Pt loss=Rρρbcb[Tb−T]−∇·K∇T [17]
With the increased RF losses Pt loss expected in higher field studies, the potential for excessive RF heating increases.
Coil Design
RF signal losses to the coil circuit and its magnet bore environment can be minimized by coil design. Similarly, RF field propagation and losses in the tissue load of the coil can be controlled by coil design. The RF field can be optimized over a region of interest for homogeneity, SNR, or other criteria. The specific absorption rate (SAR) and consequential RF heating can be reduced by proper RF coil design and excitation. Examples of coil designs and approaches to control RF loss and penetration for human biomedical applications from 3 T to 9.4 T and beyond are presented below.
Surface Coils
One approach toward limiting RF losses and controlling RF fields in high frequency applications is to limit the electrical length of the coil, for example, to less than λ/10. The electrical length of a coil can be limited by reducing its physical dimensions. So doing reduces the coil's ohmic and radiative losses (Eq. 3-10 in this section). The low inductance of a small coil also extends the upper frequency limit at which the coil will resonate. The most common small coil is exemplified by an LC circuit loop (
Surface Coil Tissue Losses
To better visualize RF tissue losses and resultant SNR, B1 penetration, SAR, and heating from simple surface coils, Equations 11-17 are graphically solved for a simple, symmetrical model by the Finite Element Method (FEM). In the example a small, 7 cm diameter surface coil as in
Surface Coil Safety
The total tissue losses from the combined eddy current losses and the displacement current losses in the tissue model are plotted in
Arrays
The small surface coil maximizes SNR by minimizing RF losses. The electrically small coil circuit minimizes resistive and radiative losses. The small tissue volume to which the coil field is coupled limits tissue losses, but also limits the field of view. To extend the FOV while maintaining small coil efficiency, an array of small coils can be composed. In phased arrays (
Mutual Decoupling
The loop elements for the phased array extend directly from
System Integration
Surface coils and arrays can be used as local, receive only coils together with more larger head or body transmit coils. In this arrangement, the small loop surface coil or array improves reception sensitivity (SNR) from a local region of interest (ROI). The larger head or body coil improves excitation uniformity over the same ROI, and beyond. Additional to mutual flux decoupling through reactive, geometric, or material means, most array receivers are temporarily decoupled from larger volume coil transmitters. The local receiver array is decoupled or detuned when a head or body coil is tuned for transmit. The receiver is then tuned, and the body coil detuned during FID signal reception. This switching between tuned and detuned states for receive and transmit coils is accomplished quickly and efficiently with PIN diode circuits imbedded in the coil circuit. (
Transmission Line (TEM) Coil Circuits
High frequency coil circuit design can include improving coil efficiency by reducing coil size. The resonant lumped element (LC) loop can be described as a surface coil and as an array element. This fundamental LC loop can be improved upon however, especially when larger coil dimensions are used such as in head or body coils at 3 T and above. For example, lumped element circuits can be replaced with distributed circuits in the form of transmission lines. Transmission line circuits can be used to limit radiation loss and to preserve current uniformity in circuits exceeding one tenth the wavelength of its carrier signal. When circuit length exceeds 0.1λ, use a transmission line.
Two or more parallel conductors capable of supporting a transverse electromagnetic (TEM) mode of energy propagation define a transmission line. Examples of TEM or transmission line circuits include coaxial lines, strip lines or micro strip lines, wave guides and other conductive cavities for which two or more walls can support the TEM mode. An RF coil for biomedical applications at 3 T and above can benefit from TEM design. In one example, an RF coil, includes a coaxial line cavity resonator such as that used for centimeter band radar applications. A transmission line resonator can be used for NMR. The following describes TEM or transmission line design and high frequency circuits.
The building blocks of an RF coil is shown in the circuits of
Lumped element circuit theory can be used to write equations for the time harmonic series and parallel lumped element circuits (
Zin=R+j(ωL−1/ωC) [18]
Yin=jωC+1/(R+jωL) [19]
The resonance frequency f for each circuit is:
f=ω/2π=1/[2π√(LC)] [20]
Transmission line theory is used to write analogous equations for the TEM circuits. For a voltage V, and current I, varying at angular frequency ω along a general transmission line, the input impedance is:
Zin=V/I=(V1e−γz+V2e+γz)/(I1e−γz+I2e+γz) [21]
for γ=α+jβ=√(R+jωL)(G+jωC) [22]
The parameter G is defined as the conductance, 1/Ω of the transmission line's dielectric layer. The phasors V1 and I1 propagate in the forward +z direction, while V2 and I2 are reflected in the −z direction. The propagation constant γ is the complex sum of the attenuation constant α, nepers/m and the phase constant β=2π/λ, radians/m. In the absence of reflected waves, the characteristic impedance Z0 of a line is defined as:
Z0=V1/I1=(R+jωL)/(α+jβ)=√[(R+jωL)/(G+jωC)]=R0±jX0 [23]
For high frequency, low loss, resonant lines where ωL>>R and ωC>>G, and the reactance X0=0,
Z0≈√(L/C), α≈R/2Z0+GZ0/2, and ν≈ω/β1/√(LC) [24]
In the air dielectric transmission line the phase velocity ν=3.0(108), the free space velocity of light. The lossless coaxial transmission line's characteristic impedance can be specified by its outer conductor i.d.=2b, its center conductor o.d.=2a, and its dielectric material ∈ separating the two conductors.
Z0=√L/C=(η/2π)(ln(b/a)) for η=√(μ/∈) [25]
For the intrinsic TEM mode wave impedance η of the coaxial line, the typical permeability μ closely approximates that of free space, μ0=4π(10−7), H/m. The permittivity ∈, of the free space (or air) dielectric is ∈0=8.854(10−12), F/m. The input impedance for the general transmission line of length l terminated in ZT can be written:
Zin=[ZT+Z0 tan h((α+jβ)l)]/[1+(ZT/Z0)tan h(α+jβ)l)] [26]
When ZT=∞, the general open circuited line input impedance becomes:
Zin=Z0 cot h((α+jβ)l) [27]
The open quarter wave line (l=λ/4) shown in
Zin=−jZ0 cot βl [28]
For ZT=0, the short circuited line input impedance is:
Zin=Z0 tan h((α+jβ)l) [29]
The shorted quarter wave line shown in
Zin=jZ0 tan β1 [30]
Expansion of Eq. [29] by a standard identity gives:
Zin=Z0(sin h2αl+j sin 2βl)/(cos h2αl+cos 2βl) [31]
Assuming Z0 is real, the input impedance of a line segment will be real at all frequencies for which βl=nπ2, where n is an integer. The frequencies for which the shorted (or open) lines will be resonant are
fr=ωr/2π=βν/2π=nν/4l=n/[4√l(LC)], n≡integer [32]
For the resonance frequency of a quarter wavelength (n=1) shorted (or open) transmission line segment, Eq. [32] in this section reduces to the analogue of Eq. [20] in this section where L and C are the distributed reactance coefficients for the resonant line segment. If line losses are attributed entirely to R, (G=0), the resonant Qr value for a section of low loss line terminated in an open or short can be shown to be:
Qr=β/2α=ωrL/R [33]
TEM Surface Coil Examples
Examples of high frequency, transmission line surface coils are seen in
These examples open the question of the relative merits of a loop coil vs. a linear element coil. A FDTD model in
Volume Coils
In one example, a surface coil or array for high field biomedical applications are receive only coils. Large head or body sized volume coils can be used for uniform excitation of the ROI to which the receiver is applied. In one example, a head coil operates in transmit and receive modes without an independent receiver. Head and body coil are relatively large circuits and include lossy electrical loads.
Birdcages
In one example a lumped element LC circuit, such as a high pass cage structure, such as the bird cage resonator are shown in
TEM Volume Coils
An NMR coil, including a coaxial line cavity resonator is an example of a TEM design using transmission line principles. A shielded birdcage design can be viewed as a transmission line coil where its foil strip rungs (strip-lines or center conductors) define a load matched, characteristic impedance with its shield (ground plane or outer conductor). In one example, a TEM coil design can be used at frequencies for human head and body imaging at the high field strengths.
A TEM volume can be designed to resonate at high frequencies and with a large volume. The coil includes multiple, electrically short transmission line elements (
The independent line elements can be directly driven, inductively excited, or a combination of both. The line elements and the field generating currents they carry can be independently controlled as well. Each line current can be used as a transmit element, a receive element, or both. Independent elements can be switched on or off, tuned or detuned. The currents on each element can be independently modulated in magnitude, phase angle, frequency, or time. Points of attachment for these independent drive or control lines and circuits are marked A, B, C, on the drawing in
By reactive decoupling of its elements the TEM coil effectively becomes a line element array. In this form it can be used for parallel imaging applications. Its SENSE performance improves with an increasing element count. A one dimensional reduction factor of approximately of 5.2 has been achieved with a 16 element coil and 6.3 with a 32 element coil TEM “array” for example.
Multi-Channel Coil Interface
The multi-channel TEM coil described above is the transducer of a more complex, multi-channel transceiver as depicted in
TEM Arrays
As with surface coils, transmission line approaches render both linear elements and loop elements. Both can be addressed as independent elements, one difference being that the linear element generates a flux vector parallel to the sample surface, and the loop generates a perpendicular vector. To describe the performance of line and loop elements as high frequency arrays, an eight element head coil is considered. (
To demonstrate, the two coils of
TEM Coil Fields, Losses, and Safety
A Maxwell model of an eight element TEM coil with a three layer head model diagrams a discussion of RF fields and losses affecting the success and safety of human imaging with volume coils at high frequencies. The head model includes a layer for low water content, low conductivity skin, scalp, skull; a layer for high water content, high conductivity brain, and a layer mimicking the aqueous humor in the eyes. All layers were adjusted to frequency dependent electrical parameters and temperature dependent thermal parameters consistent with the conditions and tissue layers modeled at 170 MHz (4 T). The homogeneous coil was linearly excited to generate the non uniform B1 field contour shown in
Image Optimization by Coil Design and Control
Design
Ultra-high field head imaging can produce the signal-to-noise sought. As modeled in
To mitigate the RF non-uniformity issue in high field imaging, “RF field shimming” techniques have been considered. The multi-channel TEM coil of
Control
The artifacts due to short tissue wavelengths of 7 T (12 cm) and 9.4 T (9 cm) MRI will be the increasingly powerful RF shims and gradients used to localize ROIs and optimize selected criteria therein by a new family of RF protocols and feedback driven algorithms only now being imagined. Each element of a TEM coil is independently tuned (as opposed to monolithic birdcage resonators) and becomes a solution for B1 field control because every element can be tuned. Every element in the TEM coil can be independently controlled, driven and received unlike the endring-bound cage, although the homogeneity advantages of the cage structure are preserved. Implementation of phase and gain control can be performed with a TEM coil (
A programmable B1 phase and gain control is shown in
High frequency coil optimization methods employing RF field gradients together with RF field magnitude and phase shimming can be used to produce fairly homogeneous head images at 7 T and 9.4 T. (
Multinuclear Coils
In addition to magnitude and phase angle, the independent TEM coil elements can be tuned to multiple frequencies for multi-nuclear imaging at ultra high fields. One method of tuning this coil for simultaneous operation at two or more frequencies is to tune independent elements to dedicated frequencies in a periodic pattern. For example, a dual frequency coil of 24 total line elements uses 12 even count elements tuned typically to the 1H frequency, alternating with 12 odd count elements tuned to the “X” nucleus Larmor frequency. This design approach enjoys a number of advantages. The double tuned TEM coil behaves essentially as two independent coils of the same form factor and field profile. Decoupling between circularly polarized 31P, 13C or 29Na and 1H Larmor operational modes at 3 T and 4 T for example typically exceeds 20 dB in a double tuned head coil. The dual tuned head coil couples to the same FOV at each frequency facilitating 1H shimming on the same volume from which X nuclei signal is acquired. The double tuned TEM head coil couples to the same FOV with approximately equal transmit efficiency and receive sensitivity at each frequency. The coil efficiency at each frequency approaches 90% that of a single tuned TEM coil at either frequency. Results from double tuned TEM head coils are shown in
Actively Detuned Coils
The TEM coil can be actively detuned for use with local surface coils and receive arrays. In one example, imaging uses a larger volume coil, such as a body coil to uniformly excite a region of interest. A smaller “specialty” receiver coil, such as an array, can be used to maximize sensitivity from a localized region of interest.
Body Coils
The TEM volume coil can be used as an ultra-high field body coil to 7 T and higher. Because the TEM coil can be viewed as an array of independent transmission line elements its achievable frequency and performance are dependent on the line elements and their geometric arrangement about a volume. The resonant frequencies achievable with TEM volume coils are independent of coil diameter because there are no inductive end rings limiting frequency and performance. Full sized TEM body coils can resonate at the Larmor frequencies for 8 T and higher. A number of degrees of freedom in coil control are another advantage of independence from “hard wired” end ring connections. The current elements of the TEM volume coil can be inductively driven, or capacitively driven by one or more signal lines. Because of increasing energy loss to large loads (e.g., body) at high frequencies, driving four or more elements is often required as demonstrated in
Cardiac imaging at 4 T offers one example of how a high field problem is presented, and how it is then corrected by RF shimming. Human body dimensions are measured in multiple wavelengths at ultra-high fields. Accordingly, B1 field penetration is highly non-uniform. RF related image artifacts result as exemplified by the sharp signal drop out in the right atrium shown in
At 7 T, body imaging artifacts become more severe. As the FDTD model in
Conclusion:
RF coil solutions exist for various magnets and field strengths for human imaging. This includes 9.4 T and above. Human head and body dimensioned coils resonant at the highest Larmor frequencies for these magnets can be demonstrated using TEM (transmission line) design approaches and principles. Transmission line volume coils and arrays can be used separately in transmit and receive mode, and together where one coil excites a uniform field and the other receives from a localized area with high sensitivity. High frequency anatomic loading conditions require current control at the coil element level to compensate RF artifacts, minimize RF power requirements and to optimize NMR performance criteria per subject in localized regions of interest. This control is enabled by new programmable transceivers together with interactive feedback. In one example, a particular pulse sequence and automated optimization algorithm is used.
Impact of Coil Diameter and Number of Coil Elements on B1 Destructive Interferences with Stripline Coil Arrays at 7 Tesla.
Introduction
B1 transmit (B1+) inhomogeneity is a major concern at ultra high magnetic field. It has been shown that destructive B1+ interferences (DB1I) between complex B1+ vectors from coil elements result in “bright center, dark periphery” patterns in axial images of the brain with coil arrays at 7 T. The following example (with 1, 8 and 16 stripline coil arrays) illustrates how B1+ profile is impacted by altering a) the distance between the coil elements and the imaged sample and b) the number of coil elements.
Material and Methods
Experimental data were obtained at 7 Tesla (Magnex, Varian) with a 3 liter 90 mM NaCl sphere phantom (90 mM NaCl saline). Three cylindrical transceiver array coils were utilized, using 7 cm wide and 14 cm long stripline elements (StE1). Coil diameter and number of StE1 were respectively 23 cm/8StE1, 23 cm/16StE1, 32 cm/16StE1. RF power was split with an 8-way or 16-way splitter, with incremental phases (45° or 22.5°). Signal was sampled with a multichannel digital receiver and magnitude images from all channels were summed together. We acquired B1+ calibration series consisting in 12 TurboFlash images with a preparation RF pulse (PP) of increasing duration. Additional experiments were performed with a single StEl positioned at 0 cm, 2.5 cm or 7 cm from the phantom.
In all experiments StE1 were used both for transmission and reception. Simulation data were obtained with XFDTD software (Remcom, Pa.). Model designs reproduced the 3 coil arrays used with the phantom (permittivity 80, conductivity=1.17 Siemens/m) as well as an 8 channels, 32 cm Ø coil. For each coil, 2D Complex B1+ profiles were generated for each individual StE1. Those complex profiles were then combined either as Sum of Magnitude (SOM) or as Magnitude of Sum (MOS). In MOS, the sum of complex profiles allows to reproduce the DB1I which occur when pulsing through all StE1s, while in SOM B1+ magnitude profiles are added as positive scalar, which does not reflect the experimental reality. The Ratio MOS/SOM estimates the amount of destructive B1+ interferences: it equals one without DB1I, and falls within the range [0,1] with DB1I.
Results and Discussion
Simulated Data
Overall (
Experimental Data
Experimental data confirm with Flash images obtained with a single StE1 (
Discussion
Our investigation demonstrates, at 7 Tesla, experimentally as well as with simulated data, that larger diameter head coil arrays as well as more numerous coil elements result in much smoother patterns of B1 destructive interferences in the periphery of a phantom (even though those destructive interferences still have large amplitude). The determination of how such geometric properties can be integrated into global RF shimming approaches is valuable, and includes multiple levels of correction: coil design, RF phase/amplitude modulation for each coil elements, multidimensional RF Pulses, Transmit SENSE.
Region of Interest Localization Through B1 Field Tailoring
Nine centimeter wavelengths in the human brain at the 400 MHz Larmor frequency for 9.4 T lead to pronounced B1 field contours and consequential image inhomogeneity. The B1 gradients can be controlled over a volume, and thus, allow steering of a constructively interfering field node to spatially correlate with an anatomic region of interest. The present subject matter includes independently controlling the phase and magnitude of TEM current element in a transmit head coil to target a local region of interest with a desired B1 distribution.
Background
For high field MRI, the degradation of B1 field homogeneity is a problem that affects image quality, among other things. Through control of the field gradients, images of selected regions of interest can be optimized for higher signal to noise ratios, lower power deposition (SAR), and other criteria.
Field control measures can be applied to achieve increased homogeneity and provide localization. The present subject matter describes a specific technique for localization of a region of interest. The present subject matter has been verified with simulation, and will be empirically validated.
Methods
A 16 element 9.4 T head TEM coil used in the laboratory has been drawn and tuned to approximately 400 MHz by the finite element method (FEM). Simulations with a centrally located cylindrical phantom were performed. The coil rung to rung diameter is 28.8 cm, shield diameter 34.5 cm and rung length 17.8 cm. The phantom diameter is 20 cm, length 20 cm, relative permittivity 81 and conductivity 2 S/m. Simulations with each rung driven independently were performed to determine the phantom response to individual drives. The RF field (B1) at a point in the phantom due to nth current element is broken into x and y components where the x component value is denoted B1 xn=Anej(ωt+φ
A similar equation can be written for the y component of the RF field. The final x and y directional B1 field can then be calculated as the sum over n of B1 xn. The overall field is found from these two quantities.
Using the finite element simulations, the quantities anxejb
Results
Simulation results in FIGS. G2.1-G2.8 show a convergence to an arbitrarily chosen, off axis B1 contour reached through current element phase and magnitude control. Successively improving matches as calculated by sum of least squares difference are shown sequentially in steps 1 through 8 shown in FIGS. G2.1-G2.8 respectively, demonstrating the algorithm's convergence to a solution improvement as compared with a uniform field.
Conclusions
The feasibility of automated B1 field localization of anatomic regions of interests has been demonstrated. A new technique for independently controlling the phase and magnitude of TEM current elements in a transmit head coil to target a local region of interest with a desired B1 distribution has been provided.
Some Notes.
An example of the present subject matter uses information about the phase variation of the receive sensitivity profiles of individual coils over a specified region, and the SNR of individual coils over that region, in order to calculate target variations in the phase and amplitude of B1+ to improve the resulting SNR when the receiver channels are properly combined.
After this target variation is calculated, the B1+ transmit maps are used to calculate the phase and magnitude of individual coils required to achieve this phase. In one example, the present subject matter includes varying the waveforms of individual elements with or without concurrent B0 gradients to achieve the target variation of B1+.
An example usage includes Single Voxel Spectroscopy with larger voxels. In SVS, the phase of the spins from the entire volume are averaged together by each coil. Then each coil is phased and combined by weighting the individual signals by their respective SNR. Voxel position determines which coils have better SNR. The goal is then to produce a phase variation across the voxel during transmit that will counter the destructive phase combination that is experienced during receive.
B1 Destructive Interferences and Spatial Phase Patterns at 7 T with a Head Transceiver Array Coil
RF behavior in the human head becomes complex at ultra-high magnetic fields. A bright center and a weak periphery are observed in images obtained with volume coils, while surface coils provide strong signal in the periphery. Intensity patterns reported with volume coils are often loosely referred to as “dielectric resonances,” while modeling studies ascribe them to superposition of traveling waves greatly dampened in lossy brain tissues, raising questions regarding the usage of this term. The present subject matter includes a transceiver coil array used in volume transmit mode, multiple receiver mode, or single transmit surface coil mode. An appropriately conductive sphere phantom shows destructive interferences are responsible for a weak B1 in the periphery, without a significant standing wave pattern. The relative spatial phase of receive and transmit B1 proved remarkably similar for the different coil elements, although with opposite rotational direction. Additional simulation data closely matched the phantom results. In the human brain the phase patterns were more complex but still exhibited similarities between coil elements. The results suggest that measuring spatial B1 phase can help, within an MR session, to perform RF shimming in order to obtain more homogeneous B1 in user-defined areas of the brain.
With the advent of ultrahigh magnetic field (UHF) human scanners (4 T and higher), RF interactions with the human body become a critical challenge as RF wavelength approaches the size of the target object. Unlike at lower fields, human head images acquired at 7 T with a volume coil display a much stronger B1 magnitude and signal intensity in the center than in the periphery. This is not a direct consequence of coil design or geometry because the same volume coil, when empty, produces a uniform magnetic field. Surface coils, on the other hand, provide at 7 T a bright signal in the periphery (close to the coil) that decays toward the center, as is observed at lower magnetic fields, albeit with a different spatial profile. Interferences between dampened RF traveling waves may be responsible for the characteristic image intensity distributions in the brain at UHF. The present subject matter considers the significance of dielectric resonances or standing waves for the characteristic nonuniform signal intensities observed with transmit RF volume coils at ultra-high magnetic field.
Multiple channel transmit array coils can be used to develop “RF shimming” techniques aimed at counterbalancing the typical, strong transmit B1 inhomogeneities at UHF, or multidimensional RF shaping with transmit SENSE. Some of the designs used for building such arrays produce multielement coils that are virtually identical in construction to volume coils most frequently used in UHF human head imaging except that the multiple elements are now highly decoupled. These array coils provide the opportunity to examine the B1 phase pattern of individual coil elements in order to better understand the global B1 profile generated in the corresponding volume coil.
RF penetration in the human head at UHF has been noted with volume coils or with surface coils. Such data can be obtained with a single transmit channel applied to the coil at one, two, or four ports and a single receiver channel (signal from all coil elements being merged hardware-wise experimentally or in simulations). This has been the case even when utilizing coils made out of multiple elements such as TEM and quadrature surface coils. It is feasible to use transceiver array coil at UHF when not considering the issue of coil-specific spatial phase profiles in RF transmission and reception.
In the present subject matter, in a 7 T field, the RF interferences between coil elements in a volume coil are considered, with an emphasis on B1 spatial phase distribution. An eight-channel transceiver coil array was used that allows the use of an identical coil structure as a volume transmit coil, multichannel receive coil, or multiple single transmit surface coils, making possible a direct comparison between transmit and receive B1 profiles. Comparisons are made both in phantoms and in the human head. Experimental results compare favorably for 300 or 64 MHz.
Rationale
Transmit and Receive B1 Fields
The elliptical RF magnetization inside the sample can be written as
{circumflex over (B)}1+=({circumflex over (B)}x+{circumflex over (B)}y)÷2 1
{circumflex over (B)}1−=({circumflex over (B)}x+{circumflex over (B)}y)*÷2, 2
where i=√{square root over (−1)}, {circumflex over (B)}1+, {circumflex over (B)}1−, {circumflex over (B)}x, and {circumflex over (B)}y are complex vectors (as identified with the circumflex accent), and the symbol * indicates the conjugate of a complex quantity. Consider that {circumflex over (B)}1+ is in the direction of the nuclear precession, so that it is the component that determines the excitation. Consider a rectangular, constant amplitude RF pulse of duration τ, the RF flip angle is V|{circumflex over (B)}1+|γτ, where V is a dimensionless scaling factor proportional to the coil driving voltage, and γ is the gyromagnetic ratio. The measured image signal intensity (SI) also depends on the local receive field {circumflex over (B)}1−*, so that
SI∝ρ|{circumflex over (B)}1−*|sin(V|{circumflex over (B)}1+|γτ), 3
where ρ is proton density. SI and {circumflex over (B)}1+ are a function of space, i.e., x and y when considering an axial slice. Image signal has a sinusoidal dependence to B1+ and a linear dependence to B1− (where B1+=|{circumflex over (B)}1+| and B1−=|{circumflex over (B)}1−*|). Differences in spatial distribution between {circumflex over (B)}1+ and {circumflex over (B)}1−* become significant at high magnetic field and result in spatially distinct transmission and reception profiles.
Transmission Versus Reception
During signal reception, the MR signal from one source is seen by multiple detectors and single versus multiple channels data sampling strategies affect signal detection efficiency. Sampling each coil element on a different digital receiver, rather than hardware-wise merging those voltage sources before sampling, avoids B1 cancellation between signal sources exhibiting unbalanced differences in phase.
When using RF transmission through all coil elements, which constitutes a single volume coil transmission, RF from multiple sources (each coil element) are merged at each point in the sample. Thus, there is a unique merged transmit B1 seen by the sample while there are as many distinct receive B1 components as the number of receive coil elements. However, since the principle of superposition dictates the resultant B1 field of multiple element coils, the sum of individual receive B1 patterns of each element must reproduce the effect of using the array coil as a single volume coil receive as well.
For the eight-channel array coil used in this example, to mimic numerically a “single volume coil receive,” a complex addition of the data recorded from the eight receive channels is performed with an appropriate phase shift for each coil element. Alternatively, merging the individual coil elements can be done hardware-wise by using RF combiners with appropriate cable lengths. The result can then be compared with the sum of the eight individual magnitude images obtained with the eight different receivers. This comparison provides information on the spatial patterns of signal addition or cancellation in the single volume coil receive mode due to the reception profile of each element in the coil.
For RF transmission, however, the analogous flexibility does not exist. It is easy to achieve the single volume coil transmit mode by dividing the RF power into eight transmit lines with an RF splitter, with proper cable lengths to adjust the phase of each coil element. However, the transmit B1 fields generated by each coil element cannot be summed as magnitude quantities in a single excitation because transmitting through more than one coil element involves summation of complex transmit B1 vectors. Still, it remains possible to transmit through one coil element at a time in a “single surface coil transmit” mode and acquire the resultant signal through all coil elements in a “multiple receiver coil” mode with the eight digital receivers. This way, pulsing each coil element alone, individual transmit profiles are obtained over the entire sample and their magnitude can then be summed to perform indirectly a comparison with the volume transmit coil mode.
Materials and Methods
Hardware
Experiments were performed with a 90-cm-bore magnet operating at 7 T (Magnex Scientific, Abingdon, UK), driven by a Varian Inova console (Palo Alto, Calif., USA). As for the coil, it is an eight-element linear stripline array of 16 cm length, constructed on a 12-mm-thick cylindrical Teflon substrate. For RF excitation, a single 8-kW RF amplifier was used (CPC, Brentwood, N.Y., USA), and the RF power was split with equal amplitude and phase with an eight-way power splitter (Werlatone, Inc., Brewster, N.Y., USA). Further transmit phase increments of ˜45° between neighboring coil elements were obtained by cable length adjustment in the transmit path. Performance can be demonstrated using an eight-channel digital receiver system that was developed in house, based on an Echotek (Huntsville, Ala., USA) ECDR-814 board oversampling the 20-MHz IF at 64 MHz and 14-bit resolution, with digital band pass filtering.
Phantom Data
Performance was demonstrated using a spherical flask of 3 liter capacity, filled with 90 mM NaCl solution in water. Note that a mixture of NaCl can match the average conductivity but not the average permittivity of the brain at 7 T. There is a tradeoff based on coil behavior: with 90 mM NaCl, coil loading (and thus tuning and matching adjustments) approximates that which is observed with a human head.
In this example, data were obtained in an axial slice crossing the center of the phantom. The center of the spherical flask was positioned close to the isocenter of the gradient coil. At the beginning of the session, shimming was performed for X, Y, Z, and Z2 with the FASTESTMAP technique and the reference resonance frequency was determined at the center of the phantom. Transverse axial images crossing the center of the phantom were obtained with TE/TR=5 ms/1 s, flip angle=15°. Measurement of noise was performed with same imaging parameters, except RF pulsing was blanked.
B1 Mapping
Transmit B1 profiles were mapped using magnetization preparation followed by ultrafast gradient recalled echo imaging. Magnetization preparation was accomplished with a variable-duration rectangular pulse ranging from 0 to 3000 μs (300-μs steps). Imaging parameters were TR/TE=18 ms/5 ms, FOV=25.6×19.2 cm2, matrix=256×64, flip angle=3° (in phantom center). Images were acquired 16 s apart.
The transmit B1 was mapped under two conditions: first, transmitting through the eight coil elements simultaneously, the “volume coil transmit mode,” and second, transmitting through one element of the coil at a time, the “surface coil transmit mode.” In both cases signal was received through eight channels. The surface coil transmit mode was repeated for each of the eight elements, resulting in eight different B1+ maps. When transmitting through one channel only, the power output of the RF amplifier was reduced by 9 dB (˜eightfold reduction) in order to maintain the same final RF power delivered to a single coil element to achieve the same B1+ amplitude per coil element as in the volume coil transmit mode. Also, each transmit side of the other seven transmission/reception (T/R) switches was terminated with a 50-Ω terminator in order to maintain coil characteristics as closely as possible (match, tune, and coupling) between volume coil transmit and surface coil transmit situations.
Once B1+ maps are obtained, receive B1 profiles can be derived from a GE image acquired with a known nominal flip angle. In order to do so, one additional GE image (TR/TE=1000 ms/5 ms) was obtained with the flip angle set to 15° in the center of the phantom.
B0 Mapping
GE images were obtained at two different echo times (6 and 10 ms) in order to map the B0 field.
Human Data
Human data were acquired. For human experiments, the coil was squeezed into an elliptical geometry (20 cm short axis, 24 cm long axis) in order to reduce heterogeneous distances between coil elements and the volunteer's head. Standard axial gradient echo images were obtained, transmitting through all eight elements (volume coil transmit mode), collecting data on the eight separate channels (receiver array).
Simulations
XFDTD software version 6.0 (Remcom, State College, Pa., USA) was used to perform simulation of B1 field distributions in the sphere phantom at 300 MHz. All simulations were run with an Intanium-2 1.6-MHz processor (Intel Corp., CA, USA) hosted at the Minnesota Supercomputing Institute (University of Minnesota). The XFDTD software utilizes a finite difference time domain method to solve Maxwell's equations. The coil model design reproduced the transmission line coil array used in the experiments. The phantom model, a 3-liter sphere, was assigned permittivity (∈r) of 80 and conductivity (σ) of 1.17 Siemens/m. Those numbers were interpolated from the literature for a saline solution of 90 mM at 300 MHz. Software calculations were obtained with a single coil element transmitting RF power, in order to reproduce comparisons performed with the phantom where either magnitude or complex data from coil elements were merged (see data analysis). {circumflex over (B)}1−* and {circumflex over (B)}1+ maps (the lower tilde indicates simulated data) for one coil element were derived from Bx and By values generated by the XFDTD software and then properly combined as in Eqs. [1] of this section and [2] of this section. The resulting maps were then duplicated in order to get eight channel data sets, each of them subjected to in-plane rotation of appropriate increments of 45° in order to generate eight coil elements simulations (as obtained experimentally). Additional calculations were also performed with the frequency set at 64 MHz in order to reproduce 1.5 T conditions. Only steady-state values obtained in an axial slice crossing the center of the sphere were considered for this example.
Data Analysis
Prior to calculation, all data were rescaled to account for variable gain level between channels, derived from the noise SD acquired from each channel while RF power was blanked.
Relative Intensity Profiles
Assign {circumflex over (p)}r,j as the complex value signal obtained with coil element j in spatial coordinate r. In order to simplify the formalism, the subscript r will not appear in most of the mathematical expressions; thus, {circumflex over (p)}j will be used instead.
Relative intensity images (RI) maps were obtained as
Rooted sum of squares (RSS) and normalized rooted sum of squares (NRSS) were computed as
The latter provides pixel values similar to the sum of the magnitude of each channel in the center of the phantom where all channels are expected to contribute for approximately the same amount of image intensity with equal phase.
For example, considering that in the center of the sample each coil j∈[1, 2, . . . , 8] contributes the same signal intensity Sj=1, then sum of magnitude is Σj|Sj|=8, root of sum of the squares is RSS=√{square root over (Σj(|Sj|)2)}=√{square root over (8)}, and the normalized version is NRSS=√{square root over (NΣj(|Sj|)2)}=8.
Merged Versus Nonmerged Data
In order to evaluate the amount of signal cancellation due to phase differences between complex signals of different coil elements for a given pixel, we computed the sum of magnitude (SOM) of each coil complex image,
as well as the magnitude of the sum (MOS),
The ratio between SOM and MOS can be used to compare the two corresponding maps. Such comparison ratios between magnitude and complex sums were also calculated, although indirectly, for receive B1 field maps and for transmit B1 field maps.
Magnitude B1 Mapping
For each pixel, SI modulation as a function of τ, duration of the magnetization preparation pulse, is explained with
SI(τ)∝Mz0 cos(V0|{circumflex over (B)}1−|γτ), 9
where Mz0 is the longitudinal magnetization at equilibrium. |{circumflex over (B)}1+| maps were obtained by fitting SI modulation with a cosinusoidal function. First, calculate |{circumflex over (B)}1+| when transmitting through all eight channels (volume coil transmit mode), and then calculate |{circumflex over (B)}1,k+| maps for each element k when transmitting one channel at a time (surface coil transmit mode). Note that the subscript k∈[1, 2, . . . , M] refers to a particular coil element used for RF transmission while the subscript j∈[1, 2, . . . , N] refers to a particular coil element used for reception (N=M=8). In volume coil mode the transmit B1 field vector {circumflex over (B)}1+ is written without a second subscript.
Once |{circumflex over (B)}1+| map is known, next is the determination of the actual flip angle at each pixel location of a standard GE image. |{circumflex over (B)}1,j−*| maps can be then derived for each channel data from the GE steady-state signal equation. Note that with the saline phantom, ρ and T1 are homogeneous through the sample and TE<<T2. In order to numerically mimic a volume coil receive, a |{circumflex over (B)}1−*| map was also calculated from the MOS image obtained with the same data (see Eq. [8] of this section).
Relative Spatial Phase Patterns for Receive B1 Field
Removing the common phase term is done to investigate coil-specific relative phase patterns. In a complex image obtained after 2D Fourier transformation the phase of the raw data from one channel at image coordinate r, φr,jraw, can be written as
φr,B1 com=φr,B*1+φr,B′*1com 11
where j is the channel (or coil element) number. In Eq. [10] of this section, the “coil specific relative spatial phase” term φr,B^1,j
where {circumflex over (p)}raw is the pixel raw complex value and arg({circumflex over (p)}raw) is its phase. The ΔB0 map can be derived from the phase of the complex ratio between two images obtained at 10 and 6 ms echo time, and φr,res accounts here for a slight offset of k-space center (see Note 1). φ0,j, ΔB0, and φr,res were removed pixel-wise from each channel complex image before any further analysis. Thus, in these calculations, the pixelwise complex values {circumflex over (p)}j were defined by
{circumflex over (p)}j={circumflex over (p)}jraw·e−iφo.j·e−iyTEΔBo·e−jφcom, 13
with the phase of {circumflex over (p)}j corresponding to
{circumflex over (p)}j={circumflex over (p)}jraw·e−iφo.j−e·iyTEBo−e−iφr.res,
The “common phase” term φr,com in Eq. [11] of this section accounts for two coil independent sources of phase: a transmit B1 phase φr,B1+, which reflects the phase of the magnetization vector Mx,y immediately after the RF pulse in the rotating frame (for a given location in the sample, this term is common for all receiver channels) and a common receive B1 phase φr,B1−*com. (The angle of π/2 rd between the transmit B1 vector and the plane of induced rotation of the magnetization vector during RF excitation is not explicitly mentioned). In order to estimate the common phase term, it is hypothesized that given the spherical symmetry of the phantom and the eightfold cylindrical symmetry of the coil structure, all coil elements would have spatial B1−* phase pattern similar in shape, but at different position in space (rotation increments of 45° every adjacent channel), whereas resulting common phase term components would be identical for all coils. An estimated φr,com as
where N is the number of coils. Since the interest is primarily in describing the relative spatial phase pattern of each coil element, the signal intensity weight (in Eq. [15] of this section) was not included. Intensity weight can also be used to obtain phase estimates with a low level of noise.
In order to evaluate the prominence of a reproducible phase pattern, each coil phase pattern of the phantom has also been subsequently rotated by an appropriate number of 45° increments in order to have all coil elements at the same location as channel 3 (arbitrary choice). Those eight rotated complex maps were then summed together; A single value decomposition (SVD) analysis was performed through those eight complex maps.
Although the human brain does not have the spherical symmetry of a sphere phantom, Eqs. [10] and [11] of this section were used to obtain a rough estimation of the relative spatial phase pattern of receive B1 field in human data.
Phase maps were also obtained with simulated data, in which case some of the terms in Eqs. [10] and [11] of this section vanish because either they are absent from the simulation process or they reflect experimental imperfections ((ΔB0, φr,res).
Relative Spatial Phase Patterns for Transmit B1 Field
Each transmit or receive magnitude B1 map, as well as the relative spatial phase pattern of the receive B1 field for each coil element, was calculated from data sets acquired with only one transmit B1+ pattern (either volume coil transmit mode or surface coil transmit mode).
However, measuring the relative spatial phase pattern for each transmit {circumflex over (B)}1,k+ field entails utilizing the eight data sets obtained when pulsing RF through one transmit coil element at a time. The phase in complex data collected with coil element j when transmitting RF through coil element k becomes
φr,B1com=φr,B*1+φr,B′*1com 17
As described for the relative phase patterns of receive B1, first remove φ0,j,k, φr,ΔB0, and φr,res from the raw data, which simplifies the apparent phase expression
φr.j.k=φ3.5+φ1.8+φr.com. 18
When transmitting through one coil element only, image SNR was very low in some areas.
Reducing the resulting phase noise entailed summing, pixelwise, the complex data of each channel j (j=1, 2 . . . , 8) obtained when transmitting through one coil element k. For each coil element k, this sum was used to estimate the phase expression φr,B1,k+rel+φr,com:
Those eight sums were then summed together in order to estimate φr,B1,com, assuming (similarly to receive B1) that the coil-specific relative phase terms φr,B1,k+rel would tend to cancel each other:
Finally, each relative coil transmit phase, φr,B1,k+rel, was derived from the two previous equations:
Additional phase calculations (rotation, sum, and SVD) were performed as described for the relative phase patterns of receive B1 field.
Polar Coordinates Analysis
Additional analyses were performed in polar coordinates in order to evaluate phase components that are periodic about the center axis of the phantom (i.e., Z axis). In order to do so, complex B1 maps were 2D interpolated at regular intervals along the radius (ρ) as well as along the angle (θ) dimensions.
All data analyses were performed in Matlab (The Mathworks, Inc., Natick, Mass., USA).
Results
Phantom Studies
Coil Element Relative Intensity Profiles
Relative intensity profiles, shown in
Signal Loss in MOS Images
Using the same data from which
Magnitude B1 Mapping for Individual Coil Elements
Sum of Magnitude Versus Magnitude of Sum
Signal Reception
Using the present subject matter, one can now directly reconstruct the sensitivity maps that would correspond to a single channel volume coil mode as |Σ{circumflex over (B)}1,j−| (
As in
Transmit B1
Note, however, in
Relative Coil Spatial Phase Patterns
Receive B1
Relative spatial phase φr,B1,j−*rel was calculated as described under Methods. Three main characteristics can be noted about this spatial phase distribution (
Transmit B1
Phase Patterns in Polar Coordinates
This subject matter considers that summing the eight complex coil phase patterns obtained with the sphere phantom, regularly distributed around a cylinder—thus with circular periodicity—yields a common phase map, which reflects two kinds of terms. The terms of the first kind, such as ΔB0, are not linked to the actual B1 field. Although they are constant for a given pixel for all coil elements, they are typically sensitive to rotations about the center. On the other hand, B1 specific common terms are insensitive to rotations by increments of 360/8=45° for a circular sample and an eight-element coil geometry; these terms are easier to consider in polar coordinate: for a given ρ, those terms are either constant along θ or have a periodic distribution along θ at frequencies that are harmonics of the number of coils (eight in our case).
Such a representation is shown in
Verification that the model in Eqs. [10] and [11] of this section could account for most of the measured phase components, entailed comparison of a phase map derived from experimental data (
The same polar analysis was performed with simulated data at 64 MHz. The results, shown in
Human Data
However, despite such visible differences, it is important to note that certain similarities can still be identified between relative phase profiles of the different coil elements as shown in
Analysis of the phase maps obtained with the human head in
Discussion
The benefits of using multiple receive coils sampled individually have been well established, first for increased SNR and later for spatial encoding with parallel imaging techniques. Subsequently, it was demonstrated that parallel imaging performance increases with magnetic field strength beyond ˜4 T, justifying even larger numbers of receive coil elements to be used at high magnetic fields. These multiple coil elements are usually operated in the receive only mode while a larger volume coil is employed for RF transmission. However, it has also been shown recently that there are advantages to using multiple element array coils for RF transmission as well, especially at high field. In this work, the RF characteristics in signal excitation and reception are examined in greater detail at an ultrahigh field for a coil operating in a single volume coil mode or a multielement array coil mode.
With a sphere phantom and a cylindrical multiple transceiver array coil, it was possible to demonstrate a relative spatial phase pattern, both for transmit and for receive phases. These phase maps utilized a phase model that distinguishes between common and coil-specific phase components. Even though such relative components will differ in different coil arrangements, this model proves efficient in identifying reproducible relative phase patterns. The phase patterns measured were highly reproducible between coil elements, and the corresponding transmit and receive average B1 patterns were remarkably similar both in amplitude and in phase, except for their anticipated opposite rotational direction. When viewed in polar coordinates, it was noted that, as a function of the polar angle, the phases varied slowly in the center and rapidly in the periphery of the sample, close to the coil elements. This phase behavior was responsible for large B1 cancellation at the phantom periphery in the single volume coil mode both for the transmit B1 field as well as for the receive B1 field.
The situation becomes much more complex with a human head, which is heterogeneous and has a complicated 3D shape that cannot be approximated with a sphere phantom. Furthermore, each coil element has now a different load and interacts with a differently shaped portion of the head with variable head to coil distances. In experiments with a human brain, it was found that the relative phase patterns for receive B1 fields were significantly more variable between coil elements than was observed with a phantom. However, the data illustrates strong, destructive B1 interferences in the periphery as well as similarities in the spatial phase patterns of the coil elements. Some features observed (less destructive interferences for coil elements with higher sensitivity in larger brain areas; higher phase pattern similarity for coil elements 180° apart) could help improve RF shimming or RF coil design.
Simulation data obtained with the XFDTD software confirmed an experimental finding, reproducing large destructive interferences in the periphery of the sample as well as large B1 phase modulation through the sample within one coil element, even though a slightly steeper gradient of destructive interference was observed in the simulated data.
There is a disparity between signal reception and RF transmission in a MR experiment aimed at volume coverage of the brain. On the receiver side, collecting the signal from each element separately on multiple receiver channels overcomes the deleterious complications of the complex phase patterns. Moreover, in the case of parallel imaging, where coil sensitivity profiles are used for spatial encoding, complex patterns in coil sensitivities are an advantage; hence, parallel imaging performs better at UHF because of the approach to far-field conditions for radiofrequency waves. RF transmission, on the other hand, as it is used so far, becomes much more challenging because the same underlying electrodynamic mechanisms result in a much weaker B1+ in the periphery of the brain when all the elements are utilized to excite the spins in the entire volume of the sample, as in a single volume coil. This can be overcome, in principle, using the transmit side in an analogous fashion to the receiver side, constructing pulses that utilize the spatial phase and amplitude profiles of each element and applying an element specific RF pulse pattern to the coil, as in transmit sense.
Besides the general pattern of peripherally weak B1 in the volume coil transmit mode, the cancellation can also be severe enough to result in very little B1+ in some regions compared to others and hence in dark spots in the image at power levels that attain maximal excitation in other regions. Such inhomogeneities will also be different for the same head but positioned differently with the coil so that the distances between the current carrying elements and the sample are altered. Aside from hardware modification, options are limited with a single volume coil. A hardware modification might entail current carrying elements to compensate for commonly occurring inhomogeneities, without having recourse to adjustments for each individual study. With a multiple element transmit coil, however, control of the phases and amplitudes of each individual element is possible, enabling routine use of RF shimming techniques. In this regard, the present subject matter a method and system for calculating RF adjustments “on the fly,” quickly enough to be compatible with the time frame of an actual patient MR exam.
No observations were made as to significant standing-wave-like pattern that could explain the weak transmit B1 amplitude (i.e., B1+) or the reception sensitivity (i.e., B1−) at the periphery relative to the center of the conductivity adjusted phantom or the brain. In the human brain or equivalent phantoms, relatively high intensity center observed with volume coils previously is not really a consequence of a dielectric resonance per se; such dielectric resonances are observed in a phantom of distilled water but are greatly damped when conductivity is increased. Either with conventional volume coils or in an array operating in the volume coil mode, the center-bright pattern either in reception or in transmission is a consequence of cancellation in the periphery rather than some sort of enhancement in the center. This cancellation is larger at higher magnetic fields (
Conclusion
The results show experimental evidence at 7 T with a head eight-element transceiver coil array that destructive B1 interferences between coil elements are responsible for typical image intensity profiles at ultrahigh field, often reported as “bright center,” without significant dielectric resonance pattern contribution. Simulated data were in agreement with those experimental findings and reproduced similar B1 destructive interferences, larger at 7 than at 1.5 T. The results also demonstrate that reduced SNR gains with increasing field magnitude in the periphery compared to the center of the brain measured with a volume coil are not inherent to the field strength per se but are recoverable using multichannel receiver arrays.
Note 1
Practically, a slight offset of k-space center subsequent to the digital receiver triggering interface was responsible for a linear phase shift along the readout direction in reconstructed images, which was systematically removed before any other processing step. This term was not explicitly include in the equations in order to avoid overloaded expressions.
Note 2
When transmitting RF through the eight coil elements, each is expected to contribute for the same amount of B1+ in the center of the phantom, i.e., each element contributes to ⅛th of the total B1+ amplitude. Keeping the same RF power per coil element, the nominal flip angle of 3° in the center of the phantom obtained with eight transmitting coil elements becomes about 0.375° with only one transmitting coil element. Note that when transmitting through one element only, B1+ amplitude becomes much stronger in the periphery, close to the coil, because there is no more B1+ cancellation from neighboring elements. Considering this feature, together with the fact that our B1 transmit approach assumes a very small flip angle all over the phantom for the fast GE images following the MPP, it was decided to keep the same RF power per coil element for the small flip angle in the fast GE images used to map B1+ (i.e., reducing RF power by 9 dB). (Note also the reduction in the RF power by 9 dB for MPP for the purpose of keeping a constant B1+ amplitude delivered to the sample per coil element). Because of the resulting low SNR, some degree of spatial filtering was used to process the transmit B1 mapping data (2D median filter, kernel size=3 pixels), which can explain the apparent smoother pattern for transmit B1 than for receive B1 mapping (see
Highest Field Human MR Imaging
Abstract—Magnetic resonance imaging (MRI) is a tool for observing the human body in health and disease. Its speed and resolving power increase with the strength of the magnetic field required for the MR signal transduction. The present subject matter includes body imaging at 7 T, and of head imaging at 9.4 T.
New radio frequency methods and technologies for high frequency MRI signal stimulation and detection, human imaging at these unprecedented field strengths are disclosed. Preliminary results of 7 T body images and 9.4 T head images together with the new RF technologies and methods to achieve them are included in this report.
I. Introduction
Whole body imaging and its clinical applications are developed at 3 T. Body imaging can be conducted at 4 T. A number of 7 T MRI systems can be used in the development of human head imaging. Consider using 7 T systems for human body imaging. Consider human head imaging at higher field strengths. Results from selected tests of 7 T human body imaging and 9.4 T human head imaging are presented and discussed.
There are technological and physical challenges to human imaging at these high field strengths and accompanying Larmer resonance frequencies. The Larmor wavelengths in the human tissue dielectric at 300 MHz (7 T) and 400 MHz (9.4 T) are on the order of 12 cm and 9 cm respectively in high water content tissues such as muscle and brain. These wavelengths may challenge safe and successful human scale imaging. RF interference patterns from uniform field volume coils can create severe RF field inhomogeneity in the anatomy. RF losses to the tissue conductor and the tissue dielectric can result in severe heating for conventional pulse protocols. The present subject matter includes methods and technology to solve some of these problems, use the short wavelengths to advantage. By controlling the currents in individual RF coil elements, in phase, gain, frequency, and time, the RF field can be manipulated to optimize signal from a targeted region of interest for SNR, SAR, CNR, homogeneity, or other criteria. Such RF field shimming can be automated much like magnetic field shimming. 7 T body imaging and 9.4 T head imaging is presented.
II. 7 T Human Body Images
Methods for conducting 7 T body imaging entails high frequency modeling, hardware development, and data acquisition.
A. Modeling
Modeling can use a Finite Difference Time Domain (FDTD) method to numerically predict RF field equipotentials in the National Library of Medicine's Visual Human digital atlas with segmented anatomy adjusted to conductivity and permittivity of 300 MHz. The RF field, circularly polarized and uniform in the unloaded condition, was generated by a TEM body coil of the circuit dimensions: i.d.=57.5 cm, o.d.=62.5 cm, active element length=33 cm, cavity length=100 cm. FIG. J 1 shows a scale model of a body loading a TEM body coil as pictured in FIG. J 2.
B. Hardware Development
The equipment and sequences specific to 7 T human body imaging are presented. The hardware, controls and pulse protocols are presented. The MRI system can include a Magnex 7 T, 90 cm bore magnet with Magnex whole body gradients and shims (i.d.=63 cm). A Varian (Palo Alto, Calif.) Inova console can be used, together with an 8 kW solid state RF power amplifier from Communications Power Corporation (Hauppauge, N.Y.). An actively detuned, 300 MHz RF body coil of the model dimensions can be configured with PIN detuning circuits and system control interface as shown in FIG. J 2. 300 MHz receiver circuits such as the eight element strip line array, half of which is shown in FIG. J 4a, and the four element strip line dual breast array shown in FIG. J 4b are configured along with PIN decoupling circuits and multi-channel, digital receivers.
C. Measurement
Human measurements can be performed. In light of the peak power (approximately 4 kW at the coil) and high frequency load losses predicted, certain measurements used low tip angle sequences of long pulse widths. For mapping of the RF landscape in the body at 300 MHz, coronal, saggital, and transaxial scout images can be acquired with the body coil activated in transmit and receive mode, without the use of additional receiver coils. The parameters used for acquiring the gradient echo, whole body images of FIG. J 3 were: 256×256 matrix, 3 mm thick slice, 2 ms windowed sinc pulse, flip angle=25 degrees. Specific to the coronal images: TR/TR=60/4 ms, FOV=50×35 cm, NT=4, time=110 sec. The sagittal images can be acquired by: TR/TE=50/4 ms, 50×35 cm, NT=2, scan time=55 sec., and the transaxial images: TR/TE=50/4 ms, 35×35 cm, NT=4, and scan time=110 sec.
The body coil can be used with an actively detuned, homogeneous transmitter for imaging with local receiver arrays. Two blocks of four TEM strip line elements each (FIG. J 4, left) can be placed above and below the torso to receive the signal summed by simple magnitude addition to create the image shown in FIG. J 5, right. In one example, to facilitate the evaluation of RF non uniformities, no intensity correction was applied. This higher signal, receive only image compares to the body coil transmit and receive image shown in FIG. J 5, left. A breast coil configured of four strip line loops in a pair of quadrature arrays is shown in FIG. J 4, right. This dual breast array can be used to acquire the breast image of FIG. J 6, right. The acquisition parameters for the gradient echo sequence can be: 256×128 matrix, 42 cm/25 cmY 3 mm thick slice, 2 ms windowed sinc RF pulse, TR/TE=50/5 ms, flip angle=20 degrees, scan time 12 sec. No intensity corrections were applied to this image. A transverse chest image can be acquired with the body coil only in FIG. J 5, left for comparison.
D. Results and Discussion for 7 T Body Imaging
The data predicted significant RF artifacts with at least one sharp line running longitudinally through the body, primarily due to destructive interference of the short (12 cm) wavelengths in the high water content tissue dielectrics at 300 MHz. Imaging proved these predictions generally correct. See FIG. J 3. Also shown in the FIG. J 3 images are what appear to be susceptibility artifacts possibly due to gas pockets in the bowel. Apparent susceptibility artifacts were also noted near the body, between the breasts in female subjects. See FIG. J 6, left. B0 phenomena were not modeled for this investigation. The RF power (SAR) used to acquire the images was estimated by: average SAR=996 W peak power×0.2789 (sinc Se power ratio)=277.7 Watts×2 ms/50 ms=11.1 W per slice in each TR. 11.1×5 slices=55.5 W, or less that 1.0 W/kg for most adults, well within the FDA guideline for human torso imaging.
Observed artifacts and RF power requirements do not appear to preclude the possibility of useful whole body imaging for at least some applications at 7 T. Improvement in both signal and homogeneity appear to be gained by using the body coil for excitation together with separate receiver coils for higher sensitivity, local image signal reception. FIG. J 5, right acquired with a local receiver for example recovers the signal void seen in the center of the body from body coil only images of
III. 9.4 T Human Head Images
A 9.4 T system has been used for human mind/brain and primate research. The system includes a 65 cm bore Magnex magnet, a Varian Inova console, a High Field NMR Systems patient table, and a multi-channel RF transceiver designed and developed with Communications Power Corp (CPC) power amplifiers. In one example, the parallel transceiver incorporates 16,500 W transmit channels, each channel with programmable gain, phase, and time (switching) control of 16 independent TEM coil elements which in turn feed 16 diode protected receive channels. Results from an eight channel version are reported here.
RF non uniformities due to extremely short tissue wavelengths (9 cm) of 9.4 T (400 MHz) MRI preclude the possibility of homogeneous head images by conventional means. Both constructive and destructive RF interference over the whole head FOV create highly non uniform excitation and reception from a conventional, homogeneous, circularly polarized head coil. Treatment of these high field non uniformities by manipulating and compensating RF field interference patterns through impedance control (phase and magnitude) of independent current elements, used at 4 T, can be modeled and demonstrated in head imaging at 7 T and 9.4 T. The short wavelength relative to anatomic geometry enhances the effectiveness of RF shims and gradients to localize ROIs and optimize selected criteria therein by selected RF protocols and feedback driven algorithms. To this end, individual elements in the TEM coils or loop arrays are independently controlled unlike the end ring-bound cages. Implementation of RF phase and gain control is performed with multi-element coils interfaced to the MR system by a parallel transceiver described herein.
Results of programmable RF phase and gain control are shown in FIG. J 7. These images were acquired with an eight element TEM coil (FIG. J 2) driven by an eight channel parallel transceiver with 500 W per channel RF power amplification. The solid state RF power amplifiers are broad banded to facilitate the addition of programmable frequency shifters for multinuclear control. Behind TR switches, eight digital receivers are used. Phase and gain values are entered on a web table for the coil-transceiver channel matrix. High frequency coil optimization methods employing RF field gradients together with RF field magnitude and phase shimming are used to produce homogeneous head images at 9.4 T. FIG. J 7 shows 9.4 T gradient echo images from an 8 element TEM coil and acquired by the parameters: TR/TE=40/5 ms, TI=1.55 sec, Thk=3 mm, matrix=256×128, SAR=0.4 W/kg. The images are eight channel magnitude addition composites without intensity correction.
IV. Conclusion
Results demonstrate whole body imaging at 7 T and human head imaging at 9.4 T.
Fast Mapping of Relative B1+ Phase in the Human Head at 9.4 Tesla with a 14 Channel Transceive Coil Array
Introduction
In the human brain, at 9.4 T for example, mapping of the relative phase of Transmit B1 (B1+), for each coil element of a head array coil, can be done in just a few tens of seconds. This mapping can be used to reduce destructive B1+ interferences. For illustration, a straightforward correction of any local low B1+ spot chosen in a brain image can be obtained (for one location at a time) with proper B1+ phase correction derived from this mapping. In this study, only RF phases were modulated, RF power being constant for any given RF amplifier when used.
Methods
Data were obtained with healthy volunteers at 9.4 Tesla (Magnex, Varian). A 16 channel cylindrical stripline transceiver array was used, as well as a 16 channel digital receiver. RF power was delivered through fourteen RF 500 Watt-Amplifiers (RFA) (CPC™). Only 14 RFA's were available for this study, thus only 14 transceivers were used for RF transmission. A programmable unit (CPC™) allowing individual phase/amplitude modulation was used to split RF input through 14 channels. B1+ phase mapping was based on complex vector formalism which requires Receive B1 (B1−) phase mapping as well. Fifteen consecutive FLASH images of an axial plane (matrix: 128×80, TR=50 ms) were obtained: one image while pulsing through all 14 RFA simultaneously, and 14 images while pulsing with each RFA (one at a time). The resulting 210 complex images (14 channels×15 repetitions) were utilized to obtain common, relative B1− and B1+ phase maps.
Results and Discussion
Note that the B1+ phase mapping is a subcomponent within the fast |B1+| magnitude mapping technique for multiple channels. The actual B1+ mapping technique is described elsewhere in this document.
Calibration Tools for RF Shim at Very High Field with Multiple Element RF Coils: Ultra Fast Local Relative Phase to Absolute Magnitude B1+ Mapping
Introduction:
At very high magnetic field, strong B1 heterogeneities occur in human MR images as RF wavelength becomes equal to, or smaller than, the size of the imaged target, interacting with lossy and dielectric properties of biological tissues. Different RF Shim techniques can be used in order to mitigate B1 inhomogeneities, which have considerable differences in their respective hardware and B1 calibration requirements, especially with large numbers of elements in a transmit coil array. In this document, a comprehensive B1 phase shim strategy is disclosed. In addition, this document considers gradual solutions for increasing B1 Shim demands and for different hardware setups as shown in Table N1. The present subject matter introduce fast B1 calibration tools utilizing some linear algebra formalism with transmit coil arrays, ranging from simple (relative B1 phase) to more demanding measures (absolute B1 magnitude). This document shows that local B1 Phase Shim can be performed in less than one minute with a 16 elements transmit coil array, with dramatic gain in local B1+ magnitude.
Principle, Methods and Results:
Considering K transmit coils k and J receive coils j, the complex signal Sk,j obtained in receive coil j in a GE image obtained when transmitting with only coil k is shown in Eq. N1, where B1− and B1+ are complex transmit and receive B1 vectors with phase φj (receive) or φk (transmit), and Mz0 reflects proton density. Rj, Tk and C are respectively Receive, Transmit and Common (non coil dependant) terms. Flip angle is given by θk=V|B1k*|γτ where τ is a square pulse equivalent duration. Note the distinction between relative phase, relative magnitude and absolute magnitude of B1k+, corresponding to measures A, B and C in Table N1. Note that in this section, B1+ phase refers to the phase in the transverse XY plane.
TABLE N1
Table 1. Required Measures and Hardware in RF Shim Techniques
NEEDED RF
MINIMUM MEASUREMENT
HARDWARE PER COIL
Relative B1+
Absolute
Phase
Magnitude
Shaped Pulse +
Mapping
B1+
Shifter
Modulation
RF Amplifier
Phase
Magnitude
Magnitude
GOAL
Mapping B1+ Destructive Interferences
+
+
LOCALIZED
B1+ Phase Only
+
+
B1+ SHIM
w/B1+ Magnitude
+
+
+
+
REGIONAL
B1+ Phase Only
+
+
+
B1+ SHIM
w/B1+ Magnitude
+
+
+
+
FLIP ANGLE SHIM
Single RF Coil
+
(2D/3D RF PULSE)
Parallel
+
+
+
Excitation
A
B
C
Measure A: Relative B1+ Phase mapping and Local B1 Phase Shim: If one acquires a quick GRE image when transmitting only with one coil, for each coil k (k:1→K), and taking k=1 as a reference, from Eq. 1 it will be noted that for any receive coil j the phase of the complex ratio Sk,j÷S1,j=Tk÷T1 is independent to T1, TE, TR or θ(θ<π). Thus, each image can be acquired in less than 3 seconds at Ernst angle for best SNR. Local B1 phase shim then includes choosing an ROI (a dedicated Matlab™ GUI was developed) and computing from the K complex ratios the phase to add to each coil k for all B1k+'s to be in phase at this location. The top row of
Measure B: Relative B1+ Magnitude and avoiding “dark holes”: Mapping B1+ destructive interference ratio (DIR) is useful to determine where, and by which fraction, complex B1k+'s are cancelling to each other. This measure can be obtained with DIRj=|ΣkSk,j|÷ΣkSk,j|, which simplifies into |ΣkB1k+|÷|k|B1k+|. DIR can take values from zero (total cancellation) to unity (no cancellation) (note that Rj and C terms vanish in DIR so that, ∀j, DIRj˜DIR). However, two conditions must now be obtained: 1) small flip angles (<10°) must be used to be in a regime where signal intensity is ≈ proportional to flip angle (and thus to |B1+|), and 2) TR chosen accordingly to reduce T1 relaxation bias. In such conditions, superposition algebra allows to predict relative destructive interference patterns for any set of B1k+ phases and magnitudes, and iterative algorithms can be used to avoid or minimize “dark or black holes” in resulting B1+ profiles. At 7 T, low resolution images can be obtained in this regime in about 10 seconds per coil in order to not introduce T1 relaxation bias in this index (3 times longer than relative phase mapping only, still about only ≈3 minutes for 16 coils). If the conditions above are fulfilled, then, ∀j: SKj=Rj.C.Σk(Tk), which is the sum through K of Sk,j, is comparable to an image obtained when actually pulsing through all coils together with the reference set of phases an magnitudes and we suggest this verification as an integrant part of standard B1 shim calibration.
Measure C: Hybrid Absolute B1+ Magnitude Mapping for multiple Transmit Coils: Although useful, other solutions may not provide flattening B1+| solutions over larger spatial scales (regional/global B1 Shim), which require absolute |B1+| mapping. At high field there is no homogenous B1+ volume coil reference. Opposite to the previous requirements for superposition formalism, absolute |B1+| mapping requires sinusoidal dependence between signal intensity and |B1+|, thus large flip angles, making |B1+| mapping typically very long because T1 relaxation bias has to be avoided (furthermore, T1's are longer at higher field). For a faster technique, the present subject matter includes a hybrid approach, including merging the previously collected relative information for each coil (small flip angle linear regime), with a unique, classic |B1K+| mapping acquisition obtained when pulsing through all K coils together (large flip angle sinusoidal regime). Indeed, for a given set of phase and magnitude: |B1K+|=|ΣkB1k+|=|Σk{|B1k+|.ejφk}|. Having measured |B1K+| and all Sk,j's, it becomes possible to generate each individual coil magnitude with, ∀j: |B1k+|={|Sk,j|÷|ΣkSk,j|}·|BK1+|. The image acquisition time to map B1k+ magnitude in a phantom with this technique at 7 Tesla was 16 times 10 s (the 16 Sk,j data) plus two times 6 minutes (two flip angles for |B1k+|), which is considerably shorter than mapping |B1k+| 16 times. Practically, any B1+ mapping technique could be utilized for the large flip angle case. Knowing |B1k+| as well as each relative phase φk, one can now calculate phase and/or magnitude sets for larger scale B1 shim with appropriate optimization algorithms. If equipped with multiple shaped RF pulse channels, it is also then possible to utilize Parallel Transmission techniques to obtain a homogenous flip angle within a predefined target with multidimensional RF pulses.
Additional Multichannel Methods
Rather than developing an RF coil and then utilize model and experiment to evaluate performance, the present subject matter operates in reverse in tools which design multiple transmitter RF coil based on target field configuration.
RF coils which are producing a B1 field (for example, the head or the body) and the object is, in some cases, to produce a relatively flat B1 field through the region to be imaged (homogenous).
At very high field, the homogeneity can be compromised. In one example, rather than developing RF coil designs with a lot of mathematical tools, the present subject matter uses the target field parameters to develop parameters for configuring the RF coil.
In one example, based on a desired field parameter, (for example a homogenous field within the coil), design the coil to satisfy that field parameter.
Typically, coils are designed to produce a homogenous field in the empty space. Coils are not built as a function of the field distortion that arises from tissue effects (when a human or other specimen) located inside the coil.
For each single RF coil, define the complex B1 profile and SAR, and then define the gain, and derive from it with optimization which combination fits best.
To summarize those writing—there are four concepts.
First, a faster technique to make B1 mapping for multiple coil elements. Mapping in the sense of calibrating. For each coil, measure for each individual coil what is the produced B1, transmit B1 field in the head or the body. There are a number of techniques to do that with one RF coil, any one of which requires at least a few minutes. In the case of multiple channel RF coils, such techniques are too time consuming.
Also, for certain multiple channel high field techniques, mapping of the B1 field will be helpful.
In one example, an acquisition technique includes a combination of relative and absolute measurements.
For magnetic resonance, the present subject matter includes making an absolute measurement of the transmit B1 field as well as a relative measure of, for example, for example, 16 or 32 coils. For example, if coil number 1 is selected as a reference, then for every single voxel, it's can be determined what is the contribution for each other coil. For example, one coil may have 3 times more or 2 times less B1 than another coil. The relative contribution of each of the coils, relatively quick to generate. Then knowing the contribution of each, final absolute amount for every single voxel in one location, generate, for example, a total of 100, and for each of the 16 coils, knowing the percentage of the contribution, then multiply this percentage by the final total to generate the absolute value for each of the individual.
In one view, the calculation scales the relative measures based on the relative contributions of each coil.
A single absolute measure allows such scaling of all other coils when all the coils are pulsing together.
Also allows determination of the field when the pulses are together, relative. There can be a number of differences. For example, when the individual coils are not in phase, the B1 of one coil can destroy the B1 of another coil. The present system allows measurement of this contribution. Knowing the absolute measurement, the contribution of each, as a percentage of that when pulsing all the coils together, able to determine the contribution of each of the coils will have contributing.
There are various options to obtain the information of one coil at a time. In
In
In
In
In sum, the various methods described above are directed to increase the signal to noise ratio of quick images.
The first data set is directed to small flip angle, quick images (one coil transmitting at a time). For the second data set,
The B1+ for the large flip angle provides a scaling factor to rescale the relative measure from the small flip angle. Note that the images generated based on a sin function as noted in the figure. This corresponds to a difference between small and large flip angle. In small flip angle it is possible to make like no sin and large flip angle it is not.
The disclosure concerns a faster method to make the global map.
The present subject matter can be used to expedite T1 mapping techniques when used with B1+ mapping techniques as described herein.
Knowing the local T1 parameter and the transmit B1, then the relative quick images can be expedited.
To make it utilizable, or like for example in a clinical field, those techniques really need to go very fast.
Utilizing quick T1 mapping with a small flip angle in the steady state (based on high SNR at high field and parallel imaging) in order to be able to map B1+ coil profile with short TR and large flip angle (where T1 impact is strong).
This will enable techniques utilizing long acquisition to be run much faster.
One application example includes the ability to acquire multislice absolute B1+ for all coils transmitting together if a non-selective magnetization preparation module is used.
In one example, a database is used for B1. A database is built based on acquired data from a subject or patient. The data corresponds to transmit B1 map obtained with a given coil for, for example, the head or the body. To optimize, in light of length calculation times, begin with the best guess based on building a data base. This will reduce the amount of time to calculate a new solution. This is an iterative process.
With a comprehensive and accurate database, estimates can be made without measurements. With sufficiently good data, the estimated parameters may be adequate.
For example, data may correspond to a particular shape of a head or the body, and based on a categorization, the effects on the coil can be estimated. The specimen is classified based on stored data in the database.
3. Utilizing a priori knowledge (database) from experiments and/or from modeling studies to anticipate B1+ adjustment sets (phase and amplitude for each element of a coil array) for a given patient and a given organ/target. For example, a head is classified according to size or shape and body mass for coil load as well. Fuzzy logic, expert systems and other technologies can be used to classify and determine parameters.
The coil load itself has an impact and can be derived from each coil's Q measured with the patient in position.
Next, consider B1 mapping filtering. To expedite mapping of B1 transmit, and when using the minimum SNR. The faster, then the less signal to noise. With noisy data, based on this database, able to guess such information. This may include information as to how to operate the array based on the signal to noise.
This may allow operation with less signal to noise ratio
For example, if there are relatively reproducible shapes in the distortion, and with two heads that are the same, then, knowing two or three quarter of the transmit B1 profile, the present subject matter can guess what the rest of it.
B1+ mapping filtering including in the a prior array knowledge.
Expected lower SNR for each relative measure in (a) areas removed from the coil (what is measured very close to the one coil has a strong SNR and that measured very remotely has a weak SNR), (b) no tissues within an image. For example, a slice of the brain that crosses a cavity, there will be a location where there is just air and it does not produce signal). Some algorithms allow acquisition for a location having noise.
To expedite with low SNR for the relative measure, there will be a lot of noise either when there is no tissue. “A” is specific—to measure the transmit B1 of a given coil (if you are far away from it in the brain, for example on the other side), the signal is low which ordinarily requires more time. However, if the database of previously measured the B1 map for the same kind of head and coil arrangement, then able to predict, with a fairly reasonable accuracy, the completed extent of B1 profile if able to measure a portion (for example, two-third or three-quarter of the map).
Then there was masking. Masking with morpho-mathematical (mathematic utilizing morphology information).
At bottom, the location having a low signal to noise ratio is enhanced based on a prior knowledge of the database and allow filling in to replace the noisy part with valuable information.
Database and/or model—the source of data (for interpolating)
Next, consider reciprocity. Utilizing reciprocity B1+/B1− in coils with relatively regularly distributed channels (for example, head) to estimate peripheral center in order to derive images.
Chart showing what coil elements, depending on their size, placement.
For RF coil design, objective is generally in generating a homogenous B1 field on the bench when the coil is empty. With a body in place, in a very high field, there is a lot of distortion of the transmit B1 field. It is possible to adjust the phase and amplitude of individual coils to improve the homogeneity. The present subject matter includes a method to reverse this procedure and use information that such distorted B1 is produced by a particular coil. Then, assume the data is in a database, the task becomes that of modeling which is based on a large number of experimental data. The analysis includes determining which combination will make the transmit B1 profile more homogenous in order to determine how to configure the RF coil.
Coil parameters include the size, each element and the distance to the biological target. Rather than building/modeling a multi-element coil as whole, rather (measure/model) complex B1+ and SAR profile for each single RF element as a function of its size, (thickness, width, length) and of its distance to the target.
Then, define the imaging/spectroscopic aim (means how to configure B1 approximately) and device from it, with optimization algorithms, which combination fits best considering a given number of available independent transmit channels.
Optimization can be either constraining or not constraining the final transmit B1 phase through space. In some case such constrain should not be needed yet.
Optimization tools constraining or not constraining on the phase are different tools.
Defining successive sets of RF amplitude/phase with different focuses for different temporal sections of an MR sequence. See the timeline noted. Times go from the left to right
Next, consider
Then switch and, as shown in the next box, illustrates a rather strong B1 one in the image box to be excited (target). These represent different spatial targets.
In sum, use fat suppression to mitigate the skin coming up, and then target a specific location (localized spectroscopy, (ellipse represents the head and the box represents the target location for spectroscopy)
At right is an oval region for imaging (good B1 everywhere).
An additional routine one more step now.
A and B which is either spectroscopy or imaging, and now we consider 3 different events for localized spectroscopy (fat suppression, and B is able to saturate, want a B1 strong everywhere else except for the target box). The surrounding tissue is uniform (OVS is outer volume suppression) so then you want to target what is not the goal and, in C generate a good B1 in the box which is in the center of sequence.
These are general schemes of sequences which are used in spectroscopy and imaging.
The next section considers skin tissue classification.
A derivative is used here.
The previous material in this section is directed to transmit B1. The following material in this section is directed to receive B1 (at very high field).
Signal intensity correction in multi-coil images based on multi-directional derivatives.
When considering a location in the brain, for example the right hemisphere in small box, the received B1 profile of the coil with vary in space when traversing the box. It will not have the same direction—there will be peaks and valleys having different orientation and slope. The underlying information, if the signal exhibits a staircase, then those staircases will still be there so from every single coil, the results will include a signal having the combination of the staircase plus homogenous sensitivity profile of the coil. For example, at a first step in the staircase, the results may be lower than it should be with one coil but perhaps another step will be stronger than it should be. By utilizing the information of multiple channels, (multiple receive coils) able to derive what resists to the signal variation due to the sensitivity profile of the coils.
Ordinarily unable to measure absolute sensitivity profile of the coils.
The technique includes calculating, in 2-D space, derivative information obtained from the same target (part of the brain) using different coils, and then remove from the data, all components that do not yield the same derivative. In other words, removing derivative information that is not collaborated by other coils for that same target.
The derivative technique herein includes calculating gradients (2-dimensional) of the image and then, for each receiver coil, a certain map in space of the local derivative. For each, there will be a certain direction and certain slope (positive or negative) and with a certain intensity.
So, for each coil, a derivative map is generated. From the multiple derivative maps, the present subject matter extracts the common elements and the differential elements. The differences between the maps then relates to the coil sensitivity profile. Those portions that are the same for each coil will correspond to the structure of the object being imaged. This can be viewed as a masking function where masking is in the sense of keeping or removing some value of special variation of the signal intensity.
This routine facilitates removal of intensity variation which is due to the coils instead of due to the tissue imaged by that coil.
So for each receive coil channel, there is a separate derivative map. This can include 2-dimensional or 3-dimensional images.
The procedure described here will remove what is different between the individual coil profiles. But even those profiles do have, to certain extent, some resemblance to each other.
The procedure includes removing the common (the shared) components between the coils. This entails using a priori knowledge regarding the kind of functional shapes which are shared by such coils
High pass 2D filtering (actually a derivative process) the image obtained from each coil elements through
In one example, the present subject matter includes parameterizing. Other information includes the gradient information (direction of the derivative and its intensity)
Each different coil will be expected to have a different local gradient, so expect to be able to detect at least as many different components as the number of coils.
Brain tissue classification includes two components.
The high pass filter (derivative) is part of this classification technique.
High pass filtering—based on the idea of classification (grey and white matter).
Based on the spatial variation of the signal intensity, it is difficult to have an algorithm which automatically detects those edges. To utilize the information of multiple coils, a high pass 2D filtering—the addition of the structure being imaged and local variation of the RF coils. This is a matter of correlation. In other words, where edges are found between two tissues with all the coils, even though all the coils don't see the image with the same local intensity, the consistency between finding edges from all the different coils shall provide a much more robust classification. This routine can be used for detecting edges with different detectors that have different sensitivities and based on whatever the detector, some local edges will show for some substantial (may be not 100% of the time but perhaps 70% of the time). This will facilitate classification when relying on a global image. This routine benefits from the use of multiple receive coils.
An example of an image having boxes is in a figure and each box has an edge between those different components.
The upper portion is directed to intensity correction.
The lower portion is directed to brain classification.
Transmit B1 mapping for coil arrays at high field in vivo.
B1+ Mapping for Each Coil Element
The following describes mapping of the B1 transmit field for each coil elements in coil arrays in order to address B1+ inhomogeneity at very high field, no matter which technique is used:
a) RF coil design improvement or alteration
a) B1 Shim (static phase and amplitude modulation)
b) Dynamic B1 modulation with different temporal B1+ profile for each coil element (this includes Transmit SENSE)
As the number of transmit RF coils increases (e.g. 16 channels for some coils, and up to 32 channels transmit channels at 7 T, for example), the time needed to map B1+ for each individual coil element becomes a limiting factor. Typical B1+ mapping requires about 5 minutes or more for one coil, because T1 is long, especially at high field, which is burdensome for numerous coil elements.
Furthermore, B1+ mapping techniques rely on fitting with trigonometric functions, which have a reasonable sensitivity to flip angle changes only in a limited range of flip angles (typically for example, sin(x) based functions will provide reliable numbers mostly in the range of 60 degrees to 120 degrees, then from 240 to 300 degrees etc. . . .). This is acceptable for a volume coil with a reasonable flat B1+ profile but not for individual coil elements of a coil array which typically have a very large dynamic range of B1+ amplitude through space.
One Relative B1+ Mapping for Each Coil Element and One Absolute B1+ Mapping for all Coils Transmitting Together
The present subject matter uses a measure of relative complex B1+ mapping for each coil element (altering one transmitting coil element at a time), and one absolute B1+ only at once, with all coils transmitting simultaneously.
Note that this concept extends to the B1+ amplitude the formalism introduced for the phase of B1 profiles: the common absolute amplitude information shared by all coils in a given location in space can be ignored when altering one coil element at a time and comparing each coil element contribution to the complex sum of all coil elements. With |B1+| the measured MR signal is linearly proportional to the |B1−| coil profile, it varies trigonometrically with the flip angle, the latter being proportional to |B1+|. This requires large flip angles to make accurate B1+ absolute measurements. Also, as noted above, each coil element, like a surface coil, requires a large range of flip angle to establish the entire B1 profile through space.
The Present Subject Matter Addresses:
A) Low Flip Angle for Fast Relative B1+ Profile of Each Coil Element (Altering One Transmit Channel at a Time) (Still Making Sure with Simulations that T1 Effect is Limited)
b) Large Flip Angles for Absolute B1+ Profile of all Coil Transmitting Together
The signal obtained with very small flip angle is almost proportional to the flip angle which itself is directly proportional to B1+ (simulate T1, TR and flip angle within a range where T1 effect is limited). Thus comparing the signal obtained with each channel for one pixel at small flip angle (i.e. a few seconds scan) provides a relative B1+ estimation. The absolute magnitude of the complex B1+ sum can be measured in a classic B1+ mapping technique (i.e. a few minutes scan) obtained by transmitting through all channels. Then, to obtain individual quantitative mapping, multiply the individual relative B1 profile of each coil element with the amplitude of the global B1 profile corresponding to the complex sum of the individual profiles.
The high signal to noise ratio at UHF provides reasonable SNR for very small flip angle utilized for relative transmit B1 mapping without spending too much time in such acquisitions.
Alternatives:
The relative B1+ profile of each coil element typically requires acquiring one image with all channels transmitting together, and then one image for each channel transmission altered individually. The image where all channels are transmitting together can be used to confirm the equivalence between numeric and hardware complex summation, thus validating the small flip angle approximation, and/or as a reference of the sum in the calculation.
Several options exist for the individual altered channels. The following are some examples: 1) to acquire one single channel transmitting at a time, 2) to acquire all channels transmitting together except for the concerned channel that is blanked, 3) to acquire all channels transmitting together, with twice the current amplitude for the concerned channel, 4) to acquire all channels transmitting together with phase rather than amplitude modulation for the concerned channel.
The technique selected will depend on SNR issues as well as on how to constraint the acquisition to stay within small flip angles in order to keep the linear approximation valid. One practical advantage of the single element at a time is the absence, by definition, of any destructive B1 interference, but even if black holes were forming in the case of transmitting all coils all the time, they might be moving in each different configuration, thus not disturbing too much the calculation.
Example of an Acquisition Scheme:
Data is acquired with all channels and complex data from each is collected for the analysis. This way, the phase and amplitude of individual B1− (receiver) will not compromise the results.
1) very fast low flip angle relative B1+ calibration.
2) Calculation from the later of a set of phases and amplitudes that will provide a more flat global B1+ profile. This requires optimization routines that in most cases will not need phase constraints for the solution. In some cases, (such as in imaging), the pixel size will be small enough to neglect significant intra-voxel signal cancellation due to the combination of B1+ phase and B1− phase.
Thus, non linear least square approaches for example can be used.
2) with this set of phase that is needed to avoid “black holes” or too dark areas in the final images, acquire B1 mapping sequence covering large flip angles where trigonometric fits can be obtained with good accuracy.
Alternative Technique for the Large Flip Angles: Utilizing Quick T1 Measurement in the Steady State and with Small Flip Angle in Order to Map B1+ with Large Flip Angle but with Short TR
fast, small flip angle T1 measurement at high field since
a) higher SNR and b) better parallel imaging performances.
This can be extended to multiple slices B1+ mapping at high flip angle if a non selective preparation module is utilized.
RF Field Localization Through Convex Optimization
Abstract
As MRI B0 field strength increases, the Larmor frequency increases causing the assumption of uniform B1 field magnitude throughout an imaged load to become a poor approximation. Because of non-uniformity, ad hoc methods can be used to adjust antenna drive element's phase and magnitude with the goal of improving uniformity. Other approaches to controlling B1 field magnitude to achieve a desired B1 magnitude distribution are considered. Desired distributions can be chosen to improve uniformity, or can be used to achieve field localization.
Introduction
In the MRI process a nuclear magnetic resonance is used to probe a load (a human body for example) in a noninvasive and nondestructive way. The received signals from different atoms are slightly different allowing imaging of the load. To setup a resonance, a z directioned DC field called the B0 field is created with a powerful electromagnet. The B0 field strength assumed in this project is 9.4 Tesla. Other values of DC magnetic field may also be used ranging from 1.5 Tesla to 11.7 Tesla and beyond. Additionally a transverse RF field called the B1 field is used to drive the resonance from various RF antennas. The B1 field lies principally in the x, y plane. z directional components are neglected in this project as their effect is small compared with the much larger B0 field strength. Additionally, several gradient (quasi DC) coils are used in the imaging process and are neglected here.
B1 field frequency is related to B0 field strength through the Larmar relationship: ω=γgB0 where γg is the gyrometric ratio and ω is the angular frequency. Thus B0 field strength determines the required B1 frequency and the wavelength decreases. The B1 field phase and magnitude are assumed to be uniform for simplicity. However as frequency increases, this assumption breaks down. This can manifest itself as spurious contrast in the MRI image due to constructive and destructive interference of B1 fields.
To alleviate this problem, installations have employed ad hoc techniques to adjust phase and magnitude of the RF signal strength fed to the RF antenna elements. This technique can involve increasing signal strength to antenna elements near dark parts of the load as seen in the image to attempt to brighten the image. These techniques are collectively known as RF shimming.
Problem Formulation
The present subject matter makes use of additional data obtained from RF field simulation or measurement. Using an RF field simulation or measurement technique (examples include FDTD, Finite Element, Integral Equation, Analytical Solution, B Field Mapping, etc.) or combination thereof, the B1 response to a single RF antenna element is calculated for a x, y slice of the load. Repeating this process for each antenna element gives the B1 field effect data for each point in the x, y slice caused by each antenna element. The combine effect of all antennas at all signal strength levels and phases can then be calculated at each point in the load. Then, using a desired x, y field map as the objective target, an optimization formulation can be made to control phase and magnitude of the signal to each antenna element.
A convex optimization technique is used to determine the magnitude and phase of each antenna element to obtain a desired field distribution. The advantages of convex optimization over other types of optimization formulations are numerous. Problems that can be cast as convex problem have efficient and accurate solution techniques that lead to global optimality with polynomial effort. These results can also be found to a desired level of accuracy.
Problem Solution
The non-convex quadratically constrained quadratic program is an NP (nondeterministic polynomial)-hard problem and thus intractable. However, through relaxation and/or formulation manipulation a good solution can be found with a deterministic suboptimality in polynomial time. This relaxation may be solved by the interior point method. Through suitable manipulation, the problem can be recast for solution by, for example, SeDuMi. Any code capable of solving a Semidefinite Program could, in principle, be used.
Problem Size
In practice, m can be 2, 4, 8, 16, 32, etc. but for the sake of this project an RF antenna set (RF coil) including 16 elements was examined. n is allowed to vary to fit computer capacity. Data can be calculated for arbitrarily many points inside the load with arbitrary geometry with a practical upper limit of 40000 points (corresponding to 1 mm point spacing on a grid 20 cm square grid) or more. However, given that the wavelength of B field variation is on the order of 9 cm at 400 MHz in a human load, one point every 1-2 cm should be sufficient to satisfy the sampling criterion.
Results Considered
As time progresses, the trend toward higher B0 field strength and higher B1 field frequency becomes necessary. Unlike attempts to maintain uniformity exclusively, the present subject matter benefits from added spatial freedom to use in the optimization allowing better conformity to desired distributions with sharp variations. The present subject matter also benefits from RF coils with increased numbers of uncoupled radiating elements. These too introduce additional degrees of freedom increasing the size and dimension of the algorithm's search space.
This technique is not limited to antenna weight design for an x, y directional plane. Weights can be found for skew planes or, more generally, any 3 dimensional volume or surface. The procedure allows for a ‘closed loop’ imaging process not previously possible. Closed loop imaging involves taking an image of a load using a standard MRI imaging process. From the image, using suitable mathematical manipulation, the Bij matrix can be derived. Using this data as input to the algorithm, antenna drive weights can be calculated improving resolution in parts of the image. Using the improved resolution images, a more accurate Bij matrix can be derived. In turn, more accurate drive weights can be calculated iteratively improving resolution. Designing antenna drive weights essentially closes the loop.
Highest Field Human MR Imaging
Abstract—Magnetic resonance imaging (MRI) is a tool for observing the human body in health and disease. Its speed and resolving power increase with the strength of the magnetic field required for the MR signal transduction. The present subject matter includes body imaging at 7 T, and of head imaging at 9.4 T.
New radio frequency methods and technologies for high frequency MRI signal stimulation and detection, human imaging at these unprecedented field strengths are disclosed. Preliminary results of 7 T body images and 9.4 T head images together with the new RF technologies and methods to achieve them are included in this report.
I. Introduction
Whole body imaging and its clinical applications are developed at 3 T. Body imaging can be conducted at 4 T. A number of 7 T MRI systems can be used in the development of human head imaging. Consider using 7 T systems for human body imaging. Consider human head imaging at higher field strengths. Results from selected tests of 7 T human body imaging and 9.4 T human head imaging are presented and discussed.
There are technological and physical challenges to human imaging at these high field strengths and accompanying Larmer resonance frequencies. The Larmor wavelengths in the human tissue dielectric at 300 MHz (7 T) and 400 MHz (9.4 T) are on the order of 12 cm and 9 cm respectively in high water content tissues such as muscle and brain. These wavelengths may challenge safe and successful human scale imaging. RF interference patterns from uniform field volume coils can create severe RF field inhomogeneity in the anatomy. RF losses to the tissue conductor and the tissue dielectric can result in severe heating for conventional pulse protocols. The present subject matter includes methods and technology to solve some of these problems, use the short wavelengths to advantage. By controlling the currents in individual RF coil elements, in phase, gain, frequency, and time, the RF field can be manipulated to optimize signal from a targeted region of interest for SNR, SAR, CNR, homogeneity, or other criteria. Such RF field shimming can be automated much like magnetic field shimming. 7 T body imaging and 9.4 T head imaging is presented.
II. 7 T Human Body Images
Methods for conducting 7 T body imaging entails high frequency modeling, hardware development, and data acquisition.
A. Modeling
Modeling can use a Finite Difference Time Domain (FDTD) method to numerically predict RF field equipotentials in the National Library of Medicine's Visual Human digital atlas with segmented anatomy adjusted to conductivity and permittivity of 300 MHz. The RF field, circularly polarized and uniform in the unloaded condition, was generated by a TEM body coil of the circuit dimensions: i.d.=57.5 cm, o.d.=62.5 cm, active element length=33 cm, cavity length=100 cm. FIG. J 1 shows a scale model of a body loading a TEM body coil as pictured in FIG. J 2.
B. Hardware Development
The equipment and sequences specific to 7 T human body imaging are presented. The hardware, controls and pulse protocols are presented. The MRI system can include a Magnex 7 T, 90 cm bore magnet with Magnex whole body gradients and shims (i.d.=63 cm). A Varian (Palo Alto, Calif.) Inova console can be used, together with an 8 kW solid state RF power amplifier from Communications Power Corporation (Hauppauge, N.Y.). An actively detuned, 300 MHz RF body coil of the model dimensions can be configured with PIN detuning circuits and system control interface as shown in FIG. J 2. 300 MHz receiver circuits such as the eight element strip line array, half of which is shown in FIG. J 4a, and the four element strip line dual breast array shown in FIG. J 4b are configured along with PIN decoupling circuits and multi-channel, digital receivers.
C. Measurement
Human measurements can be performed. In light of the peak power (approximately 4 kW at the coil) and high frequency load losses predicted, certain measurements used low tip angle sequences of long pulse widths. For mapping of the RF landscape in the body at 300 MHz, coronal, saggital, and transaxial scout images can be acquired with the body coil activated in transmit and receive mode, without the use of additional receiver coils. The parameters used for acquiring the gradient echo, whole body images of FIG. J 3 were: 256×256 matrix, 3 mm thick slice, 2 ms windowed sinc pulse, flip angle=25 degrees. Specific to the coronal images: TR/TR=60/4 ms, FOV=50×35 cm, NT=4, time=110 sec. The sagittal images can be acquired by: TR/TE=50/4 ms, 50×35 cm, NT=2, scan time=55 sec., and the transaxial images: TR/TE=50/4 ms, 35×35 cm, NT=4, and scan time=110 sec.
The body coil can be used with an actively detuned, homogeneous transmitter for imaging with local receiver arrays. Two blocks of four TEM strip line elements each (FIG. J 4, left) can be placed above and below the torso to receive the signal summed by simple magnitude addition to create the image shown in FIG. J 5, right. In one example, to facilitate the evaluation of RF non uniformities, no intensity correction was applied. This higher signal, receive only image compares to the body coil transmit and receive image shown in FIG. J 5, left. A breast coil configured of four strip line loops in a pair of quadrature arrays is shown in FIG. J 4, right. This dual breast array can be used to acquire the breast image of FIG. J 6, right. The acquisition parameters for the gradient echo sequence can be: 256×128 matrix, 42 cm/25 cmY 3 mm thick slice, 2 ms windowed sinc RF pulse, TR/TE=50/5 ms, flip angle=20 degrees, scan time 12 sec. No intensity corrections were applied to this image. A transverse chest image can be acquired with the body coil only in FIG. J 5, left for comparison.
D. Results and Discussion for 7 T Body Imaging
The data predicted significant RF artifacts with at least one sharp line running longitudinally through the body, primarily due to destructive interference of the short (12 cm) wavelengths in the high water content tissue dielectrics at 300 MHz. Imaging proved these predictions generally correct. See FIG. J 3. Also shown in the FIG. J 3 images are what appear to be susceptibility artifacts possibly due to gas pockets in the bowel. Apparent susceptibility artifacts were also noted near the body, between the breasts in female subjects. See FIG. J 6, left. B0 phenomena were not modeled for this investigation. The RF power (SAR) used to acquire the images was estimated by: average SAR=996 W peak power×0.2789 (sinc Se power ratio)=277.7 Watts×2 ms/50 ms=11.1 W per slice in each TR. 11.1×5 slices=55.5 W, or less that 1.0 W/kg for most adults, well within the FDA guideline for human torso imaging.
Observed artifacts and RF power requirements do not appear to preclude the possibility of useful whole body imaging for at least some applications at 7 T. Improvement in both signal and homogeneity appear to be gained by using the body coil for excitation together with separate receiver coils for higher sensitivity, local image signal reception. FIG. J 5, right acquired with a local receiver for example recovers the signal void seen in the center of the body from body coil only images of
III. 9.4 T Human Head Images
A 9.4 T system has been used for human mind/brain and primate research. The system includes a 65 cm bore Magnex magnet, a Varian Inova console, a High Field NMR Systems patient table, and a multi-channel RF transceiver designed and developed with Communications Power Corp (CPC) power amplifiers. In one example, the parallel transceiver incorporates 16,500 W transmit channels, each channel with programmable gain, phase, and time (switching) control of 16 independent TEM coil elements which in turn feed 16 diode protected receive channels. Results from an eight channel version are reported here.
RF non uniformities due to extremely short tissue wavelengths (9 cm) of 9.4 T (400 MHz) MRI preclude the possibility of homogeneous head images by conventional means. Both constructive and destructive RF interference over the whole head FOV create highly non uniform excitation and reception from a conventional, homogeneous, circularly polarized head coil. Treatment of these high field non uniformities by manipulating and compensating RF field interference patterns through impedance control (phase and magnitude) of independent current elements, used at 4 T, can be modeled and demonstrated in head imaging at 7 T and 9.4 T. The short wavelength relative to anatomic geometry enhances the effectiveness of RF shims and gradients to localize ROIs and optimize selected criteria therein by selected RF protocols and feedback driven algorithms. To this end, individual elements in the TEM coils or loop arrays are independently controlled unlike the end ring-bound cages. Implementation of RF phase and gain control is performed with multi-element coils interfaced to the MR system by a parallel transceiver described herein.
Results of programmable RF phase and gain control are shown in FIG. J 7. These images were acquired with an eight element TEM coil (FIG. J 2) driven by an eight channel parallel transceiver with 500 W per channel RF power amplification. The solid state RF power amplifiers are broad banded to facilitate the addition of programmable frequency shifters for multinuclear control. Behind TR switches, eight digital receivers are used. Phase and gain values are entered on a web table for the coil-transceiver channel matrix. High frequency coil optimization methods employing RF field gradients together with RF field magnitude and phase shimming are used to produce homogeneous head images at 9.4 T. FIG. J 7 shows 9.4 T gradient echo images from an 8 element TEM coil and acquired by the parameters: TR/TE=40/5 ms, TI=1.55 sec, Thk=3 mm, matrix=256×128, SAR=0.4 W/kg. The images are eight channel magnitude addition composites without intensity correction.
IV. Conclusion
Results demonstrate whole body imaging at 7 T and human head imaging at 9.4 T.
FAST B1+ Mapping Technique for Multiple Transmit Coil Elements and Fast Local B1+ Phase Shim Even in the Absence of Approximately Uniform Reference B1+ or B1− Map.
Theory and Overview
Presented below are MR acquisition and calculation tools that obtain fast absolute B1+ magnitude mapping and fast relative B1+ phase mapping for multiple transmit RF coils in the absence of an approximately uniform transmit B1+ or flat receive B1− reference map. This is typically encountered when imaging humans with Magnetic Resonance at very high magnetic field.
The tools allow for subsequently computing phase and/or magnitude modulation of the RF input for each transmit RF coil in order to obtain a more homogeneous transmit B1+ field within a slice, or within a volume, or within a user defined spatial target.
The tools also allow for subsequently computing independent shaped RF pulses to feed each RF transmit coil when utilizing parallel excitation techniques for multidimensional RF pulse and for Transmit SENSE such as techniques to obtain a predetermined 2D or 3D spatial pattern of excitation flip angle.
Consider an MR system equipped with K Transmit RF Coils (index k=1, . . . M) and J RF Receive Coils (index j=1, . . . J):
Required Measures and Hardware in B1 Shim and Flip Angle Profile control
NEEDED RF
MINIMUM
HARDWARE PER COIL
MEASUREMENT
Shaped
Relative B1+
Absolute
PROPOSED
Phase
Magnitude
Pulse + RF
Mapping
B1+
B1 SHIM
Shifter
Modulation
Amplifier
Phase
Magnitude
Magnitude
METHOD
B1 SHIM
LOCAL
Phase only
I
B1+ SHIM
Phase/Magnitude
BIASED B1 RELATIVE
Phase Only
II A
INTERFERENCE SCORE
Phase/Magnitude
UNBIASED B1 RELATIVE
Phase Only
II B
INTERFERENCE SCORE
Phase/Magnitude
FULL
Phase only
III
B1+ SHIM
w/B1+ Magnitude
FLIP ANGLE PROFILE
Single RF Coil
(2D/3D RF PULSE)
Parallel
Excitation
A
B
C
Table summarizing the context for utilizing the techniques presented hereby. The fast relative B1 phase and B1 magnitude mapping techniques correspond to measure A and measure B in the table. The absolute B1+ magnitude mapping technique for multiple transmit RF coils correspond to measure C. The “proposed B1 Shim methods” I, IIA, IIB and III correspond to actual B1 Shim
First Step: Quick Relative Measurement of Relative B1+ Phase and Relative B+ Magnitude.
Data acquisition: fast Gradient Echo Images with one coil transmitting at a time. This Acquisition can be acquired within a few seconds per coils. For each acquisition, all raw images in the complex domain are collected from all receive channels.
With K transmit RF coils, K+2 images are acquired:
The K series acquisition yields K×J complex images. The complex signal Sk,j, which can be represented as a triple product:
Rj varies only as a function of the receive coil j
Tk varies only as a function of transmit coil k
C common term independent to the coils.
One application example would be to acquire multislice |B1+| for all coils transmitting together if a non selective magnetization preparation module is utilized.
(5)
Data based choices for B1,k+ adjustments
For a given set of multi-element RF coil
(load could be obtained from measurement?) in order to utilize previously optimized sets of amplitudes and phases.
Also data based average ∝|B1−| profiles in order to estimate the actual |ΣB1+| profiles from a quick measurement.
Relative sensitivity for |B1−| can also be utilized to validate the morphologically defined coil behavior.
(if needed, fine adjustments could be obtained with reduced amount of calculation)
(3)
Utilizing a priori knowledge (database) from experiments and/or from modeling studies to anticipate B1+ adjustment sets (phase and amplitude for each elements of coil arrays) for a given patient and a given organ target.
Example: Head (considering its size/shape, and body mass for coil load as well)
Coil load itself has an impact and could be derived from each coil's Q measured with the patient.
(2A)
B1+ mapping filtering: including a priori knowledge
Expected lower SNR for each relative measure in (a) areas remote from coil (b) no tissues
(2B)
Utilizing reciprocity B1+/B1+ in coils with relatively regularly distributed channels (head) to estimate ratio peripheral/center of Σ|B1+/−| in order to derive from images a centripetal intensity correction based on Σ|B1+|×Σ|B1−|
(might be usable for quick calibration purposes)
(4) Computerized RF coil design for multi-element RF coils utilizing individual B1+ (complex) and SAR profiles for lightly coupled coil elements (“LEGO BUILDER”)
NB: optimization can be either constraining or non constraining the final b1+ phase through space (in most cases such constrain should not be needed)
Successive sets of RF amplitude/phase with different focuses for different temporal section of a sequence.
Examples: see
(6) Brain tissue classification
Principle: given a tissue signal, this structure may have higher spatial frequencies than the different coil |B1−| profiles such as the 3 color plot.
High pass filtering the images obtained from each coil element through different directions (typically as many directions as coil elements) would consistently find edges that would allow to classify brain images into subsets of (ρ,T1,T2) class.
See
All publications, patents and patent applications noted are incorporated herein by reference. While in the foregoing specification the subject matter has been described in relation to certain examples thereof, and many details have been set forth for purposes of illustration, it will be apparent to those skilled in the art that the subject matter is susceptible to additional embodiments and that certain of the details described herein may be varied considerably without departing from the basic principles described.
The above description is intended to be illustrative, and not restrictive. For example, the above-described embodiments (or one or more aspects thereof) may be used in combination with each other. Other embodiments will be apparent to those of skill in the art upon reviewing the above description. The scope of the invention should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. In the appended claims, the terms “including” and “in which” are used as the plain-English equivalents of the respective terms “comprising” and “wherein.” Also, in the following claims, the terms “including” and “comprising” are open-ended, that is, a system, device, article, or process that includes elements in addition to those listed after such a term in a claim are still deemed to fall within the scope of that claim. Moreover, in the following claims, the terms “first,” “second,” and “third,” etc. are used merely as labels, and are not intended to impose numerical requirements on their objects.
The Abstract is provided to comply with 37 C.F.R. §1.72 (b), which requires that it allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. Also, in the above Detailed Description, various features may be grouped together to streamline the disclosure. This should not be interpreted as intending that an unclaimed disclosed feature is essential to any claim. Rather, inventive subject matter may lie in less than all features of a particular disclosed embodiment. Thus, the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment.
Snyder, Carl, Adriany, Gregor, Ugurbil, Kamil, DelaBarre, Lance, Gopinath, Anand, Van de Moortele, Pierre-Francois, Tian, Jinfeng, Strupp, John, Vaughan, J Thomas, Olson, Chris E, Orser, Heather, Akgun, Can, Andersen, Peter M
Patent | Priority | Assignee | Title |
10281538, | Sep 05 2012 | General Electric Company | Warm bore cylinder assembly |
10761164, | Apr 28 2017 | SIEMENS HEALTHINEERS AG | Generating a spatially resolved magnetic resonance dataset |
10890631, | Jan 19 2017 | Ohio State Innovation Foundation | Estimating absolute phase of radio frequency fields of transmit and receive coils in a magnetic resonance |
10942232, | Mar 23 2017 | GE Precision Healthcare LLC | RF coil array and MRI transmit array |
11219384, | Oct 08 2019 | Trustees of Boston University | Nonlinear and smart metamaterials useful to change resonance frequencies |
11596309, | Jul 09 2019 | COMET TECHNOLOGIES USA, INC. | Hybrid matching network topology |
11896344, | Jul 09 2019 | COMET TECHNOLOGIES USA, INC. | Hybrid matching network topology |
11906655, | Mar 02 2017 | Symeo GmbH | Method and apparatus for capturing the surroundings |
Patent | Priority | Assignee | Title |
3199021, | |||
4899109, | Aug 17 1988 | TOSHIBA AMERICA MRI, INC | Method and apparatus for automated magnetic field shimming in magnetic resonance spectroscopic imaging |
5027070, | Dec 20 1988 | Hitachi Medical Corporation | MR imaging system and method |
5311132, | Jul 28 1992 | The Board of Trustees of the Leland Stanford Junior University | Method of enhancing the focus of magnetic resonance images |
5318025, | Apr 01 1992 | General Electric Company | Tracking system to monitor the position and orientation of a device using multiplexed magnetic resonance detection |
5614827, | Mar 30 1995 | Siemens Healthcare GmbH | Method and apparatus for shimming a magnet system of a nuclear magnetic resonance tomography system |
5644234, | Mar 04 1995 | U S PHILIPS CORPORATION | MR method and MR device for implementing the method |
5754047, | Apr 19 1994 | Hitachi Medical Corporation | Inspection method and apparatus using nuclear magnetic resonance |
6696838, | Feb 26 1999 | Purdue Research Foundation | Nuclear magnetic resonance analysis of multiple samples |
6711430, | Oct 09 1998 | EKAM IMAGING INC | Method and apparatus for performing neuroimaging |
6900636, | May 19 2001 | Koninklijke Philips Electronics N V | Transmission and receiving coil for MR apparatus |
6906518, | Nov 21 2001 | Koninklijke Philips Electronics N V | RF coil system for magnetic resonance imaging apparatus |
6969992, | Oct 03 2003 | MINNESOTA, REGENTS OF THE UNIVERSITY OF | Parallel transceiver for nuclear magnetic resonance system |
6975114, | Nov 20 2002 | NOVA MEDICAL, INC | Methods for transmit excitation in magnetic resonance imaging using a transmit pulse with time varying spatial characteristics |
6977502, | Nov 04 2002 | Fonar Corporation | Configurable matrix receiver for MRI |
6982554, | Apr 29 2004 | TEXAS A&M UNIVERSITY SYSTEM, THE | System and method for operating transmit or transmit/receive elements in an MR system |
6995561, | Apr 01 2002 | GE Medical Systems Global Technology Company, LLC | Multiple channel, microstrip transceiver volume array for magnetic resonance imaging |
20080309334, | |||
20110213235, | |||
WO2013135251, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Sep 20 2012 | Regents of the University of Minnesota | (assignment on the face of the patent) | / | |||
Apr 03 2013 | Regents of the University of Minnesota | NATIONAL INSTITUTES OF HEALTH NIH , U S DEPT OF HEALTH AND HUMAN SERVICES DHHS , U S GOVERNMENT | CONFIRMATORY LICENSE SEE DOCUMENT FOR DETAILS | 030175 | /0468 |
Date | Maintenance Fee Events |
Feb 09 2018 | BIG: Entity status set to Undiscounted (note the period is included in the code). |
Mar 08 2021 | SMAL: Entity status set to Small. |
Mar 21 2022 | M2553: Payment of Maintenance Fee, 12th Yr, Small Entity. |
Date | Maintenance Schedule |
Sep 04 2021 | 4 years fee payment window open |
Mar 04 2022 | 6 months grace period start (w surcharge) |
Sep 04 2022 | patent expiry (for year 4) |
Sep 04 2024 | 2 years to revive unintentionally abandoned end. (for year 4) |
Sep 04 2025 | 8 years fee payment window open |
Mar 04 2026 | 6 months grace period start (w surcharge) |
Sep 04 2026 | patent expiry (for year 8) |
Sep 04 2028 | 2 years to revive unintentionally abandoned end. (for year 8) |
Sep 04 2029 | 12 years fee payment window open |
Mar 04 2030 | 6 months grace period start (w surcharge) |
Sep 04 2030 | patent expiry (for year 12) |
Sep 04 2032 | 2 years to revive unintentionally abandoned end. (for year 12) |