|Home | About | Journals | Submit | Contact Us | Français|
A flexible transceiver array, capable of multiple-purpose imaging applications in vivo at ultrahigh magnetic fields was designed, implemented and tested on a 7 T MR scanner. By alternately placing coil elements with primary and secondary harmonics, improved decoupling among coil elements was accomplished without requiring decoupling circuitry between resonant elements, which is commonly required in high frequency transceiver arrays in order to achieve sufficient element-isolation during RF excitation. This flexible array design is capable of maintaining the required decoupling among resonant elements in different array size and geometry, and is scalable in coil size and number of resonant elements (i.e. number of channels), yielding improved filling factors for various body parts with different geometry and size. To investigate design feasibility, flexibility, and array performance, a multi-channel, 16-element transceiver array was designed and constructed, and in vivo images of the human head, knee, and hand were acquired using a whole-body 7T MR system. 7T parallel imaging with GRAPPA performed using this flexible transceiver array was also presented.
The inherent increase in signal-to-noise ratio (SNR) and favorable changes in MR contrast parameters at high magnetic field strengths have motivated the development of human MRI systems at 7T and above. However, one of the major challenges facing MR imaging at 7T is of the difficulty in designing optimal RF coils for efficient MR signal excitation and reception at ultrahigh operation frequencies. The design of RF coils for high fields is complicated by the fact that at higher RF frequencies there is a more complex interaction of the RF excitation field with the subject (1–11). This leads to excitation inhomogeneity due to wavelength effects and increased mutual coupling among coil elements (12–15). In addition, the lack of a body transmit coil on commercial 7T MR scanners requires ultra-high field RF coils with the ability of both transmission and reception.
A flexible transceiver RF array would offer several benefits for 7T human MR examinations because such a coil array allows for adjustments of the coil geometry to best fit the size of subjects, thus, providing the better filling factors and better RF transmission efficiency as well as improved signal-to-noise ratio (SNR) for different subjects. This could also benefit the applications of parallel imaging, parallel excitation and B1+ shimming at ultrahigh fields. The major technical challenge of designing a multi-purpose flexible array is to develop a geometry-independent decoupling solution for the coil elements. Since the mutual inductance among coil elements varies with the coil size, the requirement of flexibility poses an even more challenging decoupling issue for the flexible array design over the conventional size-fixed transceiver arrays. Thus the commonly used decoupling schemes, for example, coil overlapping, dedicated decoupling circuitry and low impedance pre-amplifiers (10,11,16) are not easily applied. Although the current source RF amplifier method (17–19) can improve the isolation among elements of a transmit array, it is not readily implemented for a flexible transceiver array and it is currently not available for most of the MRI systems. A capacitive decoupling network is commonly applied for transmit arrays especially for most of the microstrip (TEM/strip-line) transmit arrays (3,6,9,20–25). The isolation is improved by placing a capacitor either across the top conductors (9) or grounds (25) of the nearest microstrip array elements. However, the physically connected coil elements through decoupling circuits are not practical for geometry/size adjustment to better fit different human parts. In addition, the decoupling capacitance is a function of gap, angle and loading of coil elements, when coil geometry/size changes, the value of the decoupling capacitors have to be re-adjusted in order to achieve the required decoupling, which is usually undesired and impractical in real MR examinations with patients.
In this work, a flexible transceiver microstrip array was designed for human MR imaging applications at the ultra-high field of 7T. Without using any decoupling circuit, the coil elements are free of electrical connections to each other and easily to be set up for multiple human body parts. In addition to its geometrical adjustability for the human head with various dimensions, this array also has demonstrated applicability for different parts of the human body such as the wrist/palm, knee or torso. The mutual coupling between coil elements was addressed by implementing the proposed microstrip coil array with mixed primary and secondary harmonic elements (22,26). Both the dimensions and the number of coil elements for this coil array may be adjusted for different applications. To investigate its feasibility, MR imaging experiments using the flexible transceiver array were performed on a 7T scanner for multiple anatomic locations. The preliminary images obtained using sum-of-squares (SoS) and GRAPPA parallel imaging reconstructions demonstrate its performance.
In this work, the primary (1st) and secondary (2nd) harmonics of microstrip transmission lines (24,26–28) were chosen to build the coil elements. Two types of microstrip resonators named primary (or 1st) harmonic elements and 2nd harmonic elements were utilized for coil array design. Figure 1a illustrates the placement of the 1st and the 2nd harmonic coil elements in a flexible array and Figure 1b reveals the circuits of the two types of coil elements. The first harmonic coil element is a typical straight-type microstrip coil with two shunt capacitors at the ends of the strip conductor. The second harmonic coil element has an additional shunt capacitor at the center of the microstrip resonator, which is shown in Figure 1b bottom. Two resonance modes are observed in this 2nd harmonic element. The central shunt capacitor is used to tune the frequency of the 2nd harmonic resonance. Different current distributions of the 1st and 2nd harmonic modes result in intrinsic decoupling between these two types of elements if they are placed in parallel. By alternately placing the 1st and 2nd harmonic elements, the nearest neighbors are intrinsically isolated (26). The other elements in this flexible array also possess adequate decoupling at the high frequency of 300MHz range.
Based on this proposed design, a multiple channel flexible array was fabricated for human applications at 7T (Figure 2a). The coil elements were built on Teflon boards 20 cm long, 3 cm wide and 7 mm thick. The dimensions were selected to balance between decoupling and B1 penetration. The strip conductors and ground planes of the microstrip coil were made of 36-μm-thick adhesive-backed copper tape (3M, St. Paul, MN, USA). The width of strip conductors was 0.63 cm, and the width of ground conductor was 2.54 cm. Two tunable capacitors (Voltronics, Denville, NJ) were used at each coil element for tuning and matching. Coil elements were attached parallel one another onto a soft canvas former through back-cohesive Velcro strips. This design made it easy to fine adjust element position for decoupling and to attach/detach those elements for different applications. Up to 16 coil elements, including eight 1st harmonic elements and eight 2nd harmonic elements were constructed and selectable for specific applications. Additional slotted RF shield sheets, each at 5 cm×20 cm, were inserted into the canvas and placed behind the ground plane of coil elements to increase the stability. To protect the human body, a 0.5 mm thick Teflon sheet was inserted between the array elements and subject. Figure 2b shows the flat array with the maximum 65 cm in length loaded with 16 elements. The gap between nearest neighbors was only 0.6 cm ~ 1.2 cm and good isolation was still achieved. This multi-purpose array is flexible enough to be bent into semi-volume and volumes with various dimensions according to practical requirements. Four applications for the human hand, knee, head and liver were tested on the bench and using a 7T MRI scanner. Figure 3a shows the arrangement of this coil array on a volunteer for those four different body parts. Soft canvas and appropriate cushions were employed in this coil array to provide extra patient comfort. The smallest elliptical volume with eight coil elements was 12 cm ×9 cm in diameter which fits the human hand or foot. More compact coil sizes than this might cause unacceptable coupling among the coil elements, especially among the 1st harmonic elements. A larger size of approximately 16 cm in diameter with eight coil elements is applicable for the human knee scan. All sixteen elements were implemented for human head imaging and the dimensions typically used were approximately 18 cm × 22 cm for our volunteer studies. This array may also be placed at the right side of the human torso to cover the liver. The eight coil elements thus placed as a semi-volume shape with 22cm × 35 cm in diameter (refer to Figure 3b). We confirmed the decoupling of any two coil elements by measuring reflection and transmission coefficients (S11 and S21, respectively) using a network analyzer (HP model 4396A), while connecting the other coil elements with 50 Ohm terminators.
Specific Absorption Rate (SAR) evaluation of the proposed flexible transceiver array was performed numerically by using FDTD method (Remcom Inc, State College, PA). Average SAR and maximum local SAR were calculated and compared with the FDA limits. SAR distribution in a head model generated by the flexible coil array was quantitatively mapped.
MR imaging experiments with this array coil were performed on a 7T/90cm magnet (GE Healthcare, Milwaukee, WI). This scanner provided two transmit channels with 0° and 90° phases. Each channel included a T/R switch. To test the transceiver arrays on this system, MR imaging scans were performed by connecting two coil elements into the transmit channels each time, and combining all sub-images offline. To avoid the signal cancellation from phase difference of the channels, the two coil elements scanned at each time were non-adjacent 1st harmonic elements or 2nd harmonic elements. All other coil elements without scanning were terminated with 50 ohm terminators. We acquired typical gradient echo (GRE) and SPGR images in axial and sagittal orientations using this flexible microstrip array. To avoid the potential amplifier damage caused by mismatching of coil elements, the scan was performed with individual transmit optimization and the peak power to each coil element was ~125 W and was controlled below 160 W for all the imaging experiments. RF pulse used was the standard sinc pulse with two side lobes for 7T imaging. The pulse width was 1ms. Repetition time (TR) was 100ms for head and hand images, 150ms for knee images. After the acquisition, the images and raw data from each coil element were saved and reconstructed with specialized MATLAB code.
To investigate the parallel imaging performance of this transceiver array, GRAPPA reconstructions (29,30) were performed. Partial k-space data were extracted from the full dataset to simulate 2× and 3× accelerations. A total of 16 auto-calibrating signal (ACS) lines were used for human head fitting process and 10 ACS lines for palm and knee scans.
Each element of the proposed flexible transceiver arrays was tuned and matched for 7T imaging. In all loading cases, as exemplified in Figure 4, the transceiver arrays demonstrated excellent decoupling performance. The resonance peaks for each coil element remained distinct without splitting in the different loading situations for imaging the head, knee, hand and liver performed in this study. Even with the coil elements placed very densely with 0.6 cm to 1.2 cm inner gaps for human hand and head imaging, the obtained isolations were better than −12 dB without causing any detectable resonance peak split. S21 for human knee and liver were better than −16dB due to the larger gap between elements and the flatter coil arrangements. It is worth noting that the decoupling performance varies with the coil geometry; showing a better performance in a planar array than in a volume array. Tuning and matching were easily performed even when the loading and coil shapes were significantly changed. For all the applications, the loaded Q factors of the primary harmonic elements were approximately 40 to 90, and the loaded Q factors of the second harmonic elements were in the range of 110 to 130.
In the SAR evaluation of the proposed flexible transceiver array, FDTD analysis and calculation with a human head model demonstrated that the average SAR and maximum local SAR in the head were 0.1 W/kg and 2.5 W/kg, respectively, when the feeding peak power was 160W with a duty cycle of 1%, as shown in Table 1 and Figure 5. As indicated in Table 1, both average SAR and maximum local SAR were under the FDA limits of 3.0 W/kg (average) and 8.0 W/kg (local). In the safety consideration, previous investigations (2,3,5,8,9,21,31,32), the E-field of microstrip resonators is highly concentrated within the substrate between the ground plane and strip conductor of the coil element. The SAR or E-fields of such two-conductor structures in the image area is usually low. The dimension of the microstrip resonators used in this work is similar to the ones in the publications (3,9,20,21,25). Thus, the local SAR was well below FDA guidelines even when the power was high enough to provide a 90 degree flip in the center region. Our experiments on a straight type microstrip resonator showed that a nominal 90 degree (even a nominal 180 degree) could be safely reached without triggering the RF power monitor.
Figure 6 shows the images of human hand, knee and head, which were combined offline with sum-of-squares (SoS) method and GRAPPA reconstructions with parallel imaging reduction factors of up to 3.
In the MR imaging experiments, all images were acquired using small flip angles to reduce the excitation power and to avoid any possible RF system damage caused by impedance mismatch or mistuning due to unexpected breakdown of capacitors or other elements during the experiments. The B1 penetration in low flip-angle images might be insufficient for deep tissues. In addition, the isolation lower than −16dB might not be sufficient for a transmit array in practice, especially for large samples where higher RF power is required. In future studies, we aim to optimize the design of each element to improve B1 penetration as well as coil isolation. Combining the use of a current source RF amplifier may offer better performance, including deeper image coverage and better coil isolation.
Due to the limitation of the transmit channel number of our 7T system, it was not possible to acquire images using all 16 channels at one time. Therefore a B1+ shimming application of this coil array applying transmit signals with different phases and amplitudes to each element has not been implemented at this stage. To investigate the decoupling performance and image coverage of the flexible transceiver array, multiple acquisitions were performed by connecting two elements once and then combined all the sub-images offline. Since the B1 fields of the 1st harmonic elements and 2nd harmonic elements differ, specific phase and amplitude arrangements can be applied to achieve homogeneous images and reduce RF “hot spots”. As an option in the use of the proposed coil array, only the 1st harmonic elements can be used for spin excitation during transmit phase. As such, B1+ shimming can follow the methods described in previous works (33,34). During receive phase, both the 1st harmonic elements and the 2nd harmonic elements are utilized for signal reception (therefore, the number of coil elements is doubled) to gain higher SNR and improved parallel imaging performance. This option of operation is particularly useful to the MR systems that are not equipped with a large number of transmit channels.
A multiple purpose flexible transceiver array for 7T MR applications in humans was successfully designed, constructed and tested. The performance of the proposed transceiver array was investigated on the bench and with in vivo human imaging of the hand, knee, and head. This multichannel transceiver array demonstrated unique capability of maintaining the required decoupling among resonant elements in different array size and shape and excellent SAR performance at 7T. These features make it possible and practical for this flexible array design to efficiently image different body parts which usually have a large variation in size and shape, at the ultrahigh field of 7T in transmit/receive manner. Besides the imaging examples presented here, this flexible transceiver array has also been successfully used to image the ankle and liver, and would be applicable to other possible applications in humans at 7T.
This work was supported in part by NIH grants EB004453, EB008699, EB007588-03S1, UL1 RR024131-01 and P41EB013598, and a UC discovery grant ITL-Bio04-10148 and a QB3 research award.
A model was created and the ABCD method (commonly used in microwave theory and application) was used to calculate the frequencies of the 1st and 2nd harmonic elements.
Figure 7 shows the circuits and current distributions of these resonators. The resonator in Figure 7a is a typical straight microstrip coil with two shunt capacitors at the ends of the strip. To perform the tuning of 2nd harmonic resonance peak, additional shunt capacitor was placed at center of the microstrip resonator, which is shown in Figure 7b. Without the shunt capacitors on the strip, the second harmonic peak has twice the resonance frequency of the primary peak, while after employing shunt capacitors, the second harmonic frequency is decreased and tunable by varying the shunt capacitances C2. Current distributions along two resonators are also shown in Figure 7.
To calculate the second harmonic frequency, the schematic structure in Figure 7b is repeated in Figure 8a. For convenience, the calculations assume that the capacitors at the two ends have the same value C1, and the central capacitor is C2. Its symmetric equivalent structure along the dash line is shown in Figure 8b. According to “magnetic/electric wall” analysis in microwave theory, the symmetry circuit can be analyzed by inserting an open circuit (magnetic wall) and a short circuit (electric wall) into the dash line. Any bisection magnetic and electric circuit, which is shown in Figure 8c is used to calculate the resonant frequencies with 1st and 2nd harmonics.
For primary resonance mode shown in Figure 8c, the ABCD matrix is expressed as follows.
where ω and Z0 are angular frequency and characteristic impedance, respectively. θ denotes the electric length of the microstrip,
where λ is wave length within dielectric material and l is the length of the microstrip. Eq.  only included the capacitor C1 part and transmission line part of the model shown in Figure 8c. The short circuit in this circuit will be treated as a load ZL in the following calculation.
For resonance condition Yin = 0, the primary resonant frequency is obtained.
where f 1st = ω/2π. Similarly for the circuit shown in Figure 8c (right), its ABCD matrix may be represented in the form:
Since the loading impedance ZL = ∞ (open circuit), and input admittance Yin is:
The second harmonic resonant frequency can be obtained:
In a special case when C1 and C2 is not equal to 0, the 2nd harmonic frequency can be expressed as
where f 2nd = ω/2π. Note that the second harmonic frequency f 2nd is not double of the primary frequency f 1st due to the involved shunt capacitors C1 and C2. For distributed-element microstrip resonators (C1 = C2 = 0), Equations  and  can be simplified and we have f 2nd = 2 f 1st, which has been indicated in reference (24).