The Fractionated Dipole Antenna: A New Antenna for Body Imaging at 7 Tesla

Transcription

1 HARDWARE AND INSTRUMENTATION- FULL PAPER Magnetic Resonance in Medicine 75: (2016) The Fractionated Dipole Antenna: A New Antenna for Body Imaging at 7 Tesla Alexander J.E. Raaijmakers, 1 * Michel Italiaander, 2 Ingmar J. Voogt, 1 Peter R. Luijten, 1 Johannes M. Hoogduin, 1 Dennis W.J. Klomp, 1 and Cornelis A.T. van den Berg 3 Purpose: Dipole antennas in ultrahigh field MRI have demonstrated advantages over more conventional designs. In this study, the fractionated dipole antenna is presented: a dipole where the legs are split into segments that are interconnected by capacitors or inductors. Methods: A parameter study has been performed on dipole antenna length using numerical simulations. A subsequent simulation study investigates the optimal intersegment capacitor/inductor value. The resulting optimal design has been constructed and compared to a previous design, the single-side adapted dipole (SSAD) by simulations and measurements. An array of eight elements has been constructed for prostate imaging on four subjects (body mass index ) using 8 2 kw amplifiers. Results: For prostate imaging at 7T, lowest peak local specificabsorption rate (SAR) levels are achieved if the antenna is 30 cm or longer. A fractionated dipole antenna design with inductors between segments has been chosen to achieve even lower SAR levels and more homogeneous receive sensitivities. Conclusion: With the new design, good quality prostate images are acquired. SAR levels are reduced by 41% to 63% in comparison to the SSAD. Coupling levels are moderate (average nearest neighbor: 14.6 db) for each subject and prostate B þ 1 levels range from 12 to 18 mt. Magn Reson Med 75: , VC 2015 Wiley Periodicals, Inc. Key words: Ultrahigh field; Transceive array; Dipole antenna; Body imaging; Prostate imaging INTRODUCTION Ultrahigh field imaging has demonstrated unprecedented imaging resolutions with additional advantages in the realm of functional contrast mechanisms. However, ultrahigh field imaging of deeply located targets in the body is compromised by the reduced penetration depth of B1 signals and interference effects caused by the reduced wavelength. Therefore, initial body imaging experiments using a conventional birdcage body coil revealed signal voids in 1 Department of Radiology, UMC Utrecht, Utrecht, The Netherlands. 2 MR Coils B.V, Drunen, The Netherlands. 3 Department of Radiotherapy, UMC Utrecht, Utrecht, The Netherlands. *Correspondence to: Alexander Raaijmakers, HP Q00.118, Heidelberglaan 100, 3584 CX, Utrecht, The Netherlands. a.raaijmakers@umcutrecht.nl Received 6 August 2014; revised 9 December 2014; accepted 10 December 2014 DOI /mrm Published online 2 May 2015 in Wiley Online Library (wileyonlinelibrary.com). VC 2015 Wiley Periodicals, Inc the images and low efficiency, requiring large amounts of power for acceptable levels of B þ 1 (1). The signal voids caused by interference can be diminished using a multitransmit setup and radio-frequency shimming (2,3). The reduced power efficiency has led to a dominance for on-body transmit arrays in 7T body imaging to assure sufficient B þ 1 levels (1,3). With these techniques, body imaging at ultrahigh fields is being explored using a wide variety of transmit or transceive surface arrays. These arrays consist of microstrip elements, (3 4) loop coils, (5) dipole antennas (6,7), or combinations thereof (8). Various studies show that dipole antennas may have advantages over more conventional coil array elements when imaging deeply located body targets at ultrahigh fields (6,9,10). The use of dipoles as coil array elements was first introduced by Raaijmakers et al. in 2009 (6,11). These dipoles were mounted on a ceramic substrate to keep the conservative E-fields associated with dipole antennas outside the subject tissue. The resulting elements were called single-side adapted dipole antennas (SSAD antennas). The advantages were a larger signal penetration and a more homogeneous excitation/sensitivity pattern, particularly for higher field strengths. Winter et al. introduced a variation on the single-side adapted dipole antenna, using a bowtie-shaped antenna on a water-filled substrate (7). Wiggins et al. investigated the potential of dipole antennas (without substrate) for head imaging (8,12). A combination of dipoles and loop coils is advocated by Wiggins et al.(8) and Lee et al. have introduced the folded dipole antenna (13). More recently, other variations combining the designs of dipoles and loop coils also have been presented (14,15). In succession of the SSAD antenna, Raaijmakers et al. have introduced the fractionated dipole antenna (16,17). The design consists of a 30-cm dipole antenna split into several segments, and these segments are connected by lumped elements (capacitors or inductors). These lumped elements modify the currents and voltages on the antenna to enhance the performance. It was shown that a fractionated dipole antenna with inductors between the segments has lower specific-absorption rate (SAR) levels than the SSAD antenna. However, although the improvements are apparent, it remains unclear whether they are caused by the segmentation or by the increased length of the dipole. A plain, elongated dipole antenna without segmentation may also have advantages over previous designs. To separate the gain from the increased antenna length from the gain

2 The Fractionated Dipole Antenna 1367 FIG. 1. Simulation setup (a) Study on dipole length. Investigated values: 5, 10, 15, 20, 25, 30, 35, 40, and 45 cm. (b) Study on fractionated dipole intersegment impedance values. Investigated values Z: 1, 2, 5, 10, 20, 50, 100, 500, and 1000 pf and 10, 20, 50, 100, and 200 nh. (c) Schematic representation of the simulated dipole antenna on a phantom with indicated dimensions. provided by the segmented design, two separate parameter studies are performed by numerical simulations. First, the effect of antenna length is investigated. After that, the effect of interelement lumped elements is determined. Then, a more thorough analysis and imaging results of the new type of antenna array are presented. This will include numerical simulations of an array setup, typical MR imaging examples, and a series of scatter parameter (S-parameter) and B þ 1 measurements on four imaging subjects with widely varying body mass indices (BMIs). METHODS Single Element: Dipole Antenna Length To investigate the imaging performance of a dipole antenna without a dielectric spacer, finite-difference time-domain (FDTD) simulations have been performed using the software package SEMCAD X (Speag, Zurich, Switzerland). The simulation geometry contains a 0.5-m cubic phantom with a permittivity of 34 and a conductivity of 0.4 S/m (Fig. 1c). These phantom properties represent the average tissue properties in human tissue (18). A dipole antenna is modeled by two PEC (perfect electric conductor) planes of variable length and 10 mm width, separated by a 2 mm gap and placed at a 20 mm height over the phantom. The dipole is fed in the center with a 50 ohm voltage source. Harmonic simulations (single frequency) are carried out at 298 MHz for antenna lengths ranging from 5 to 45 cm, with 5 cm increments (Fig. 1a). The resulting B þ 1 and local SAR distributions in the phantom, normalized to 1W accepted power, are used for evaluation. Results were compared to the performance of a SSAD antenna (6) (substrate with permittivity 36 and dimensions of cm 3 ). Throughout this work, we assume that these single element studies for dipole antennas largely reflect the performance of such elements in an array because interelement coupling is generally not an issue for dipole antennas close to tissue, and peak SAR locations are directly underneath each antenna so they do not overlap (which otherwise would cause severe deviations between single-element SAR levels and SAR levels in an array setup). Single Element: Dipole Antenna Segmentation To investigate the potential benefit of a segmented dipole antenna, a subsequent study is performed. Results presented in Section 3 will show that for 7T, a dipole antenna length of 30 cm is optimal for the B þ 1 over SAR max ratio, where SAR max is the peak value in the 10g averaged SAR distribution. The previously described simulation setup with a dipole antenna length of 30 cm is used where each leg of the dipole antenna is split in three pieces with 2 mm gaps in between, resulting in the simulation geometry, as indicated in Figure 1b. The number of three segments per leg is an arbitrary choice; it could be the subject of an optimization study, which was beyond the scope of this article. The dipole is fed in the center with a 50 ohm voltage source. The dipole segments are interconnected by lumped elements that are modeled as either capacitors or inductors. Harmonic simulations (single frequency) are carried out at 298 MHz repeatedly with varying lumped element values. Tested values are 1, 2, 5, 10, 20, 50, 100, 500, and 1000 picofarad (pf) capacitors and 10, 20, 50, 75, 100, and 150 nanohenry (nh) inductors. The resulting B þ 1 and 10g local SAR (SAR 10g ) distributions in the phantom, normalized to 1W accepted power, are used for evaluation. Note that receive performance for single elements is also revealed by the B þ 1 distribution, which is a mirror symmetry copy of the B 1 distribution that, per unit power, corresponds to the intrinsic signal-to-noise ratio (SNR) of a single element (19). Results presented in section 3 will show that an inductor value of nh has the most beneficial B þ 1 over SAR max ratio. High power inductors with little effective resistance can be realized with a meander structure. The required dimensions for a 75 nh inductor were estimated to correspond to the structure indicated in Figure 5a (75 mm conductor length). A physical realization with four of such meanders was realized by printed circuit board (PCB) grinding. The PCB board with the copper structures was mounted on a 20 mm PMMA substrate. Matching was realized by two series capacitors of 12 pf. Note that this design is only optimal at 7 Tesla and aiming at optimal B þ 1 over SAR max at 10 cm depth. The physical element as described above was placed on a pelvis-shaped PVC phantom containing ethylene glycol with 35 g/l salt. This mixture was chosen to achieve the permittivity/conductivity values of 34 and 0.4 S/m. MR measurements were performed in a 7 Tesla system (Philips Healthcare, Cleveland, OH). B þ 1 maps were acquired using the actual flip-angle imaging (AFI) method (20). Results were compared to a SSAD antenna (6) with a ceramic substrate with permittivity 36 and dimensions of

3 1368 Raaijmakers et al. FIG. 2. Simulated B þ 1 and 10g averaged SAR distributions for plain dipole antenna with lengths of 5, 30 and 45 cm and distributions for fractionated dipole antennas with 1 pf or 150 nh between the segments. Setup as indicated in Figure 1. Presented B þ 1 distributions are from the central yz-plane; SAR distributions are from the top xz-plane, where the SAR levels are highest. Scaling for B þ 1 distributions is logarithmic; scaling for SAR distributions is linear. Green dashed lines indicate profiles as presented in Figure 3 and cm 3. Both measurement setups (fractionated dipole and SSAD antenna) were implemented by means of numerical electromagnetic simulations (SEM- CAD X, Speag) to verify the outcome of the measurement comparison. Array Setup After achieving good results with one single element, eight copies of the element were created where the antenna shape was implemented on predesigned printed circuit boards using FR4 material (Eurocircuits N.V., Mechelen, Belgium). Two series capacitors of 12 pf ensured sufficient impedance matching. The elements were placed around a subject, four at the posterior and four at the anterior side. Polycarbonate covers were fabricated for each element to safeguard imaging subjects from touching electric parts and to prevent elements from being damaged when placed upside down with the imaging subject on top. Before use on volunteers, the B 1 /Bþ 1 efficiency and potential SAR exposure were determined by FDTD simulations. The setup as described above was implemented in the simulation environment using the human model Duke from the Virtual Family (21). For each element, two balanced series capacitors were used to realize a real impedance of the antenna ( tuning of the antenna). Then, a matched condition was obtained by choosing the impedance of the source equal to the impedance of the antenna. Eight simulations were performed, for which each time only one of the elements was active. Electric, current density and B 1 -fields were exported for further processing (MATLAB, Mathworks, Natick, MA). The same procedure was repeated for an array of single-side adapted dipole antennas. To achieve the required abutment of the SSAD antennas, the elements were either slightly sunk into the tissue or minor gaps between the elements and tissue were filled with fat. Note that a previous study on SSAD antennas has shown that such an approach does not result in deviating SAR predictions (22). Each array setup performance is characterized by its relative intrinsic SNR (taking sum of magnitude of B 1 per unit power), its transmit efficiency (average B þ 1 in region of interest after B1 shimming) and SAR exposure. In a multitransmit setup, the SAR depends heavily on the applied phase settings. The SAR exposure can be characterized by the maximum potential SAR distribution (for each voxel, the highest 10-gram averaged local SAR value that can possibly be achieved with the most disadvantageous phase settings). This distribution is presented for both arrays. However, the maximum value in this distribution is much higher than the highest local SAR value that will typically be reached when using shim settings aimed to maximize the B þ 1 in a region of interest. Therefore, the local SAR distribution for a B þ 1 - shimmed phase setting is also presented for both setups. To demonstrate the performance of the array in vivo, MR prostate measurements are being performed on four healthy volunteers (age 22 34, BMI ). All volunteers filled out a written form of informed consent. Measurements are performed on an 8-channel Philips 7T platform, equipped with 8 2 kw amplifiers (CPC, Hauppauge, NY). The antenna elements are connected to custom made Tx/Rx switches. Total power loss in the chain from amplifier to the Tx/Rx switch is around 57% (average over all channels). The antenna array is fixedtuned. To demonstrate the robustness of fixed tuning for various loading conditions, the scattering parameter (Sparameter) matrix is measured for each volunteer, providing reflection and coupling values. Also, for each volunteer AFI B þ 1 measurements are performed to establish the efficiency of the array. RESULTS Single Element: Dipole Antenna Length Figure 2 shows the simulated B þ 1 and 10g averaged SAR distributions inside the phantom for dipole antennas with the lengths of 5, 30, and 45 cm. Figure 3 shows B þ 1 and SAR profiles, as indicated in Figure 2. Figure 3a shows in-depth B þ 1 profiles for the investigated dipole lengths and the SSAD antenna. These graphs show that, close to the surface of the phantom, short dipoles have the strongest signal. All signals decay exponentially along the depth of the phantom, and the individual graphs are hard to distinguish at large depths. Therefore, Figure 3b shows the same profiles relative to the SSAD antenna. For each depth, which length is most favorable is clearly shown. This is portrayed even more clearly by Figure 3c, where the B þ 1 level at three distinct depths (5, 10, and 15 cm) is plotted as a function of dipole antenna

4 The Fractionated Dipole Antenna 1369 FIG. 3. Simulation results of B þ 1 and SAR for a plain dipole antenna with varying length. (a) In-depth Bþ 1 profiles per unit power. (b) Relative in-depth B þ 1 profiles per unit power, normalized to the profile of the 25 cm antenna. (c) Relative B þ 1 signal per unit power as a function of dipole antenna length. Graphs indicate B þ 1 field amplitudes at 5, 10, or 15 cm depth, normalized to each graph s maximum. (d) 10g averaged SAR profile per unit power (profile along the longitudinal axis in the top of the phantom where SAR levels are highest). (e) Relative in-depth B þ 1 profiles divided by the square root of the maximum SAR 10g level. All graphs normalized to the 25 cm antenna. (f) Relative B þ 1 signal per square root of the maximum SAR 10g level, as a function of dipole antenna length. The B þ 1 / SAR max ratio is shown for 5, 10, or 15 cm depth relative to each curve s maximum. Location of all profiles is indicated in Figure 2 by green dashed lines. length. Note that each graph is normalized to its maximum. Optimal antenna lengths for target depths of 5, 10, and 15 cm are 10, 17.5, and 22.5 cm in terms of transmit efficiency and receive sensitivity. The picture changes when SAR is taken into account. Figure 3d shows the axial profiles of the 10g averaged SAR distributions along the top slice of the phantom where the SAR levels are highest, as indicated in Figure 2. Strikingly, for short dipoles (5 and 10 cm) the profile shows two separate maxima, indicating that the SAR levels at those locations originate from the high E-fields at the endings of the dipole antenna. However, for longer dipoles the SAR level is maximal in the center, indicating that these E-fields originate from induction due to the local time-varying B1-fields. Therefore, the SAR profiles (except for the separate maxima in the 5 cm and 10 cm profiles) reflect the current distribution along the antenna. Clearly, longer antennas have more shallow current profiles with lower maxima. Longer antennas also provide lower SAR levels and higher B þ 1 levels for large depths. There is a tradeoff for shallow and intermediate depths. In Figure 3e, the B þ 1 profiles divided by the square root of the maximum SAR level are plotted relative to the SSAD antenna, similar to Figure 3b. This figure again shows at what depth which antenna length performs best. Lastly, Figure 3f shows the B þ 1 SAR max ratio at three distinct depths as a function of dipole antenna length (similar to Fig. 3c). Again, each graph is normalized to its maximum. Now, the optimal antenna length for a target depth of 5 cm is 25 cm; whereas for target depths of 10 and 15 cm the maximum is reached at around 30 and 40 cm, after which the graph seems to plateau. Single Element: Dipole Antenna Segmentation After investigating the influence of dipole length, the potential of dipole segmentation is studied. A 30 cm dipole antenna is chosen because of its beneficial B þ 1 over SAR max for a target depth of 10 cm (Fig. 3), which is the approximate prostate depth. Figure 2 shows the simulated B þ 1 and 10g averaged SAR distributions for a 30 cm fractionated dipole on a large phantom (Fig. 1), for two intersegment lumped elements: 1 pf and 150 nh. In Figure 4a, the B þ 1 profiles along the central y-axis are presented for each lumped element value, along with the profile for a plain dipole with the same length (30 cm). Figure 4b shows the same profiles, normalized to the plain dipole antenna profile. This graph shows more clearly the relative performance of each design, particularly for larger depths. Figure 4c shows the relative B þ 1 at the depths of 5, 10, and 15 cm as a function of the impedance of the lumped element values. Note that the negative impedance values correspond to using capacitors as intersegment lumped element components, whereas positive impedance values correspond to using

5 1370 Raaijmakers et al. FIG. 4. Simulation results of B þ 1 and SAR for a fractionated dipole antenna of 30 cm, with capacitors or inductors between the segments. (a) In-depth B þ 1 profiles per unit power. (b) Relative in-depth Bþ 1 profiles per unit power, normalized to the profile of a plain dipole antenna. (c) Relative B þ 1 signal per unit power as a function of intersegment impedance value (negative impedance corresponds to intersegment capacitors, positive impedance corresponds to intersegment inductors). Graphs indicate B þ 1 field amplitudes at 5, 10, or 15 cm depth relative to the plain dipole antenna (0 Ohm inter-segment impedance). (d) 10g averaged SAR profile per unit power (profile along the longitudinal axis in the top of the phantom where SAR levels are highest). (e) Relative in-depth B þ 1 profiles divided by the square root of the maximum SAR 10g level. (f) Relative B þ 1 signal per square root of the maximum SAR 10g level as a function of intersegment impedance value (negative impedance corresponds to intersegment capacitors, positive impedance corresponds to intersegment inductors). The B þ 1 / SAR max ratio is shown for 5, 10, or 15 cm depth relative to the plain dipole antenna (0 Ohm inter-segment impedance). Location of the profiles is indicated in Figure 2 by green dashed lines. inductors as intersegment lumped element components. All profiles are normalized to the B þ 1 level for 0 Ohm impedance, which corresponds to a plain dipole without segmentation. These graphs clearly show that the use of intersegment capacitors provides a gain in B þ 1 strength, particularly at shallow depths (< 10 cm). At 15 cm depth, there is no noticeable influence on the B þ 1 strength when using capacitors. The use of intersegment inductors results in a reduced B þ 1 strength, again particularly for the shallow depths but the larger depths also experience some B þ 1 reduction. These results would favor the use of capacitors as intersegment lumped element components because of a gain in B þ 1 at shallow depths without a penalty in B þ 1 at larger depths. However, we also evaluated the influence of the particular choice of intersegment lumped element values on the SAR distributions (Fig. 4d). These distributions show a clear advantage for the use of inductors as intersegment lumped elements. However, as demonstrated by Figures 4a c this reduction in SAR goes accompanied by a reduction in B þ 1 efficiency. Therefore, figure 4e shows the B þ 1 level at each particular depth, divided by the square root of the maximum SAR 10g value, normalized to the plain dipole profile. Note that a 10% higher value in this diagram corresponds to a 20% lower SAR value for the same B þ 1 level. Figure 4f shows, similar to Figure 4c, the relative B þ 1 SAR max ratio for the depths of 5, 10, and 15 cm as a function of the impedance of the intersegment lumped element components. Figure 4f shows that the lowest effective SAR level is achieved when an inductor is chosen for the intersegment lumped element value with a value of approximately 100 nh. However, for higher inductor values, the performance steeply declines. We therefore aimed at a somewhat more moderate 75 nh inductor between the segments. This value was achieved by a meander structure (7.5 cm length) due to low losses associated with such structures and their easy manufacturing and reproducibility on PCBs. Figure 5a shows a constructed version of the design, which has been used for benchmark tests against our previous body array element, the SSAD (Fig. 5b). Figures 5c and d show the simulated B þ 1 distributions for both elements in the corresponding columns, and Figures 5e and f show the measured B þ 1 distributions for both elements in the corresponding columns. Figure 5g shows the in-depth B þ 1 profiles for both the SSAD antenna and the new fractionated dipole antenna. These results show that the measurements correspond with the simulations. Array Setup Figure 6 shows simulation results for an array of the new fractionated dipole antennas (Fig. 6d) in comparison to an array of SSAD antennas (Fig. 6a). Figures 6b and e show the relative receive performance (sum of magnitude of B 1 per unit power). Figures 6c and f show the Bþ 1 efficiency after phase-shimming on the prostate. Results

6 The Fractionated Dipole Antenna 1371 FIG. 5. Simulated and measured B þ 1 distributions in a phantom for the fractionated dipole (a) and the single-side adapted dipole antenna (b). (c d) Simulated B þ 1 distributions for both antennas. (e f) Measured Bþ 1 distributions for both antennas. (g) Measured and simulated profiles along the pink dotted lines, as indicated in c f. B þ 1 show that both the SNR and the B þ 1 efficiency of the arrays are expected to be similar. For the same simulation setups, Figure 7a and c show the maximum potential SAR distributions (worst-case) as explainedinsection2.figures7banddshowthesar distributions when the elements are driven with the phaseshimmed settings from Figures 6c and f. The maximum SAR level for each distribution is indicated below each figure. These results clearly show much lower SAR levels for the new fractionated dipole antenna array in comparison to the array of single-side adapted dipole antennas. An array of eight fractionated dipole antennas was constructed and used for prostate imaging (Fig. 8). Typical imaging performance is demonstrated by Figure 9, showing a sagittal survey image to depict the field of view and a transverse T2w turbo spin echo (TSE) image, being the most clinically relevant imaging sequence for diagnosis and treatment planning of prostate cancer. The scattering matrices for four volunteers are depicted in Figure 10. The BMI of these volunteers ranges from 20 to Matching networks have been built and adjusted on volunteer (c) which explains the lowest reflections on this subject. In general, reflections are moderate: less than 10 db for most elements and most subjects. The upper elements (5 8) in general show less reflection ( 18.3 db on average) than the lower elements ( 12.5 db on average). Coupling to nearest neighbors shows the same behavior: All elements less than 10 db. Upper elements ( 19.1 db on average) perform better than bottom elements ( 12.1 db on average), probably due to larger interelement distances. These levels of reflection and coupling are moderate and will result in power losses of no more than 10%, depending on the phase settings (largest eigenvalue of S-parameter matrices is ). Phase-shimming was performed on each subject individually, resulting in a B þ 1 level in the prostate of (a) 18 mt, (b) 14 mt, (c) 14 mt, and (d) 12 mt. Clearly, the efficiency drops with increasing BMI. B þ 1 FIG. 6. Simulated B 1 distributions for the single-side adapted dipole antenna array (a c) and the fractionated dipole antenna array (d f). Simulation geometries (a and d). Sum of magnitude of 1W normalized B 1 distributions (b and e). Phase-shimmed Bþ 1 distributions for 8 x 1W (c and f). Pink contour shows the prostate location; indicated below are the mean B 1 values in the prostate.

7 1372 Raaijmakers et al. FIG. 7. Simulated SAR distributions for the single-side adapted dipole antenna array (a b) and the fractionated dipole antenna array (c d). (a and c) show, for each voxel, the maximum 10g averaged SAR level that could possibly be obtained with 8 x 1 W input power using the worst possible shim settings. (c and d) show the 10g averaged SAR distribution for the prostate shimmed phase settings of Figures 6c, f. The maximum value for each distribution is indicated below. DISCUSSION Classically, dipole antennas are tuned to their resonance frequency by adjusting the length. The length should be approximately equal to half the wavelength in the target medium. Because the wavelength in tissue is 12 to 16 cm, the dipole antenna should have a length of 6 to 8 cm. Such antennas do indeed cause high SAR levels (Figs. 2 and 3). A previously presented dipole antenna design (SSAD antenna) avoids these SAR levels by using a high-dielectric spacer. However, the antenna length does not need to correspond to half the target wavelength. The resulting nonreal impedances can be adjusted by a matching network. This provides a degree of freedom that has been explored in this study. If the antenna is chosen longer, the increased target volume results in an increased loading of the antenna. This results in more dampened current amplitudes on the antenna and, consequently, more dampened SAR levels at the surface. Although the dampening results in lower B þ 1 levels at the surface, the B þ 1 level at depth is increased with increasing length. Beyond a length of 20 cm, the B 1 level at depth starts to decrease, but the B1 per unit SAR max increases further and stabilizes at a length of 30 cm. These results show that a dielectric spacer is not essential when using dipole antennas for transmit purposes in MRI. After optimization of the length, the performance of the antenna can be further improved by the separation of the antenna legs into segments that are interconnected by lumped elements. This design is referred to as a fractionated dipole antenna. It reduces the SAR level by another 20% if the antenna elements are interconnected by large inductors. (Note that a 10% increase in B þ 1 / SAR max corresponds to a 20% decrease in SAR max /(B þ 1 )2 ). Apparently, inductors increase the effective length of the antenna. Note that inductors can be realized by meander structures that were used by Orzada et al. to stretch the current profile of microstrips (4), resulting in similar advantages (lower SAR, less coupling). Also Wiggins et al. use meander structures to reach the required electric length of their antennas (23). If the inductors become very large (150 nh), the effective length of the antenna becomes large enough to support two wavelengths. This is indicated by the appearance of two separate maxima in Figures 2 and 4d. This also causes the trend of lower SAR levels with increasing inductor value to collapse, as is indicated clearly by Figure 4f; for intersegment impedance values of 280 Ohm and higher (¼ 150 nh), the increasing B þ 1 / SAR max ratio decreases sharply. Reversely, the use of capacitors could be regarded as a means to decrease the effective length of the antenna, resulting in a more peaked current distribution in the center of the antenna, with consequently more peaked B þ 1 and SAR distributions around the center as one might expect for short dipole antennas. However, the fact that the B þ 1 level at depth remains unchanged in comparison to a plain dipole with the same length (Figure 4c) distinguishes a fractionated dipole antenna with capacitors from a short dipole antenna. FIG. 8. (a) Fractionated dipole elements with 20 mm PMMA spacer and covers. (b) Volunteer imaging setup with eight fractionated dipole elements around the pelvis. (c) Schematic representation of the element placement.

8 The Fractionated Dipole Antenna 1373 FIG. 9. (a) Sagittal survey image showing longitudinal extent of the field of view, (b) T2w TSE healthy prostate image. TR/TE ¼ 8,000/70 ms, 1 x 1 x 2 mm3, SENSE2, NSA ¼2, TSE-factor 9. From these results, we have developed a novel type of antenna for body imaging at 7T: the fractionated dipole antenna. The design consists of a segmented dipole with meander inductors in between. The elements are fixedtuned; they do not need manual adjustment of tuning capacitors after placing them on an imaging subject. Although the resulting matching conditions are not optimal, they are by all means acceptable. For four volunteers with widely varying BMI values, all reflection levels are below 10 db except one. Note that 10 db means that 90% of the power is still arriving at the antenna; consequently, 95% of the potential B1 is being achieved. Also, the coupling between the elements can be suboptimal (up to 11 db). However, the coupling to next-nearest neighbors is always much less. The resulting total power losses through reflection and coupling are below 10%. In general, fractionated dipole antennas with inductors show lower SAR levels per unit B þ 1 in comparison to plain dipole antennas. In addition, their shallower indepth B þ 1 profile (Fig. 4a) causes a more homogeneous flip angle distribution. Drawback is a slightly lower transmit efficiency at the target depth ( 6% at 10 cm; Figs. 4b,c). In receive, this translates into lower SNR at depth ( 6% at 10 cm) and much more at the surface (Figs. 4b,c). For deeply located imaging targets, the SNR reduction is acceptable. In addition, it gives the images (not corrected for coil sensitivities) a more homogeneous intensity distribution because the dropoff of element sensitivities is not as steep (Fig. 4a). These figures apply to 75 nh inductors, the estimated inductance of the meanders in the presented design. However, the real inductance of such a meander structure may be lower due to parasitic self-capacitances. Alternatively, introducing capacitors between the antenna segments would increase SNR at the surface without compromising SNR at larger depths (Fig. 4c). This would motivate the realization of a switchable transceive element that switches between inductors and capacitors using PIN diodes to arrive at the optimal performance for respectively transmit and receive operation. Enhanced coil sensitivity inhomogeneities would then have to be corrected for. This concept will be further explored as a subsequent improvement on the current design. An array of fractionated dipole antennas shows 40% to 60% lower SAR levels than an array of previously developed SSAD antennas. In spite of this improvement, sequence parameters are still severely constrained by SAR restrictions. At this stage, we use the worst possible SAR level to calculate the maximum time-averaged input power to the array. Using a SAR limit of 20 W/kg (24), the maximum allowed average input power is W, based on results of Fig. 7. On top of this, we limit the maximum input power by an additional factor of 2 to W to account for variations between our simulation model and imaging subjects. We realize that these combined limitations of the time-averaged input power will most likely result in severely lower SAR levels in the imaging subject than the actual SAR limits allow. More accurate SAR predictions are the subject of ongoing research, which will result in less restricted average input powers. FIG. 10. Scattering matrices for an array of fractionated dipole antennas on four different subjects with a BMI of (a) 20, (b) 22.5, (c) 23.7, and (d) 27.

9 1374 Raaijmakers et al. With this new array, many imaging targets within the body are being explored, including the heart, prostate, kidneys, cervix, and spine. Already, studies are being performed that compare intrinsic SNR at 7T with 3T. Further improvements of the current array are being developed, including combination with a separate receive array underneath the antennas or switchable antennas for optimal transmit and receive performance. Also other institutes are working on improved arrays. Our expectation is that body imaging at 7T will demonstrate advantages over 3T within the upcoming year. CONCLUSION Dipole antennas in MRI can be used for both transmit and receive arrays without the need for a dielectric spacer to avoid high SAR levels. Lowest SAR levels are achieved if the antenna is 30 cm or longer. We have introduced a new type of coil array element: the fractionated dipole antenna. The element consists of a 30 cm dipole antenna where the legs of the antenna are split into segments. The segments are interconnected by capacitors or inductors to steer the resulting B 1 field distribution. Introducing inductors reduces effective SAR levels by 20% (per unit B þ 1 ) at the expense of 6% receive performance at depth. At the surface, the receive sensitivity is diminished much stronger providing inherently more homogeneous images. If high SNR at the surface is desired, capacitors between the segments are more beneficial, but this will elevate SAR levels. A design with meander structures as inductors has been chosen for its ability to reduce SAR levels. An array of eight of these elements has been constructed. EM simulations show that SAR levels are reduced by 41% to 63% in comparison to the single-side adapted dipole antenna. The new array provides good quality images, with a large longitudinal field of view. B þ 1 levels in the prostate range from 12 to 18 mt for a volunteer BMI range of 20 to 27.5, using 8 2 kw amplifiers. References 1. Vaughan JT, Snyder CJ, DelaBarre LJ, Bolan PJ, Tian J, Bolinger L, Adriany G, Andersen P, Strupp J, Ugurbil K. Whole-body imaging at 7T: preliminary results. Magn Reson Med 2009;61: Adriany G, Van de Moortele PF, Wiesinger F, Moeller S, Strupp JP, Andersen P, Snyder C, Zhang X, Chen W, Pruessmann KP. and others Transmit and receive transmission line arrays for 7 Tesla parallel imaging. Magn Reson Med 2005;53: Metzger GJ, Snyder C, Akgun C, Vaughan T, Ugurbil K, Van de Moortele PF. Local B1þ shimming for prostate imaging with transceiver arrays at 7T based on subject-dependent transmit phase measurements. Magn Reson Med 2008;59: Orzada S, Quick HH, Ladd ME, Bahr A, Bolz T, Yazdanbakhsh P, Solbach K, Bitz AK. A flexible 8-channel transmit/receive body coil for 7 T human imaging. In: Proceedings of the 17th Annual ISMRM Scientific Meeting & Exhibition in Honolulu, Hawai i, USA, 2009, p Graessl A, Renz W, Hezel F, Dieringer MA, Winter L, Oezerdem C, Rieger J, Kellman P, Santoro D, Lindel TD, and others. Modular 32- channel transceiver coil array for cardiac MRI at 7.0T. Magn Reson Med 2013;72: Raaijmakers AJ, Ipek O, Klomp DW, Possanzini C, Harvey PR, Lagendijk JJ, van den Berg CA. Design of a radiative surface coil array element at 7 T: the single-side adapted dipole antenna. Magn Reson Med 2011;66: Winter L, Ozerdem C, Hoffmann W, Santoro D, M uller A, Waiczies H, Seemann R, Graessl A, Wust P, Niendorf T. Design and evaluation of a hybrid radiofrequency applicator for magnetic resonance imaging and RF induced hyperthermia: electromagnetic field simulations up to 14.0 Tesla and proof-of-concept at 7.0 Tesla. PLoS One 2013;8(4). 8. Wiggins GC. Mixing loops and electric dipole antennas for increased sensitivity at 7 Tesla. In: Proceedings of the 21st Annual ISMRM Scientific Meeting & Exhibition in Salt Lake City, Utah, USA, 2013, p Raaijmakers AJE, van den Berg CAT. Antennas as surface array elements for body imaging at ultrahigh field strengths. emagres. Hoboken, NJ:John Wiley & Sons; Lattanzi R, Sodickson DK. Ideal current patterns yielding optimal signal-to-noise ratio and specific absorption rate in magnetic resonance imaging: computational methods and physical insights. Magn Reson Med 2012;68: Raaijmakers AJE, Lagendijk JJ, Klomp DJ, van den Bergen B, Possanzini C, Harvey PR, van den Berg CA. Boosting B1þ efficiency for RF transmit surface elements by a radiative antenna design. In: Proceedings of the 17th Annual ISMRM Scientific Meeting & Exhibition in Honolulu, Hawai i, USA, Wiggins GC, Zhang B, Lattanzi R, Sodickson D, inventors; New York University, assignee. Dipole array arrangement. US Patent WO A1. April 19, Lee W, Cloos MA, Sodickson D, Wiggins GC. Parallel Transceiver Array Design Using the Modified Folded Dipole for 7T Body Applications. In: Proceedings of the 21st Annual ISMRM Scientific Meeting & Exhibition in Salt Lake City, Utah, USA, 2013, p Lakshmanan K, Cloos M, Lattanzi R, Sodickson D, Novivkov D, Wiggins GC. The circular dipole. In: Proceedings of the 22nd Annual Meeting of ISMRM Meeting & Exhibition in Milan, Italy, Lakshmanan K, Cloos M, Lattanzi R, Sodickson D, Wiggins GC. The loopole antenna: capturing magnetic and electric dipole fields with a single structure to improve transmit and receive performance. In: Proceedings of the 22nd Annual Meeting of ISMRM Meeting & Exhibition in Milan, Italy, Raaijmakers AJE, Voogt I, Klomp DWJ, Luijten PR, van den Berg CAT. Dipole Antenna Without Ceramic Substrate and Still Low SAR: The fractionated dipole antenna. In: Proceedings of the 21st Annual Meeting of ISMRM Meeting & Exhibition in Salt Lake City, Utah, USA, Raaijmakers AJE, Voogt I, Klomp DWJ, Luijten PR, van den Berg CAT. Prostate Imaging at 7 Tesla with Fractionated Dipole Antennas: A new type of radiative coil array element with lower SAR levels. In: Proceedings of the 21st Annual Meeting of ISMRM Meeting & Exhibition in Salt Lake City, Utah, USA, van den Bergen B, van den Berg CA, Klomp DW, Lagendijk JJ. SAR and power implications of different RF shimming strategies in the pelvis for 7T MRI. J Magn Reson Imaging 2009;30: Wright SM, Wald LL. Theory and application of array coils in MR spectroscopy. NMR Biomed 1997;10: Yarnykh VL. Actual flip-angle imaging in the pulsed steady state: a method for rapid three-dimensional mapping of the transmitted radiofrequency field. Magn Reson Imaging 2007;57: Christ A, Kainz W, Hahn EG, Honegger K, Zefferer M, Neufeld E, Rascher W, Janka R, Bautz W, Chen J and others. The virtual family development of surface-based anatomical models of two adults and two children for dosimetric simulations. Phys Med Biol 2010;55:N23 N Ipek O, Raaijmakers AJ, Lagendijk JJ, Luijten PR, van den Berg CA. Intersubject local SAR variation for 7T prostate MR imaging with an eight-channel single-side adapted dipole antenna array. Magn Reson Med Wiggins GC, Zhang B, Lattanzi R, Chen G, Sodickson D. The electric dipole array: an attempt to match the ideal current pattern for central SNR at 7 tesla. In: Proceedings of the 20th Annual ISMRM Scientific Meeting & Exhibition in Melbourne, Australia, 2012, p Medical electrical equipment Part 2 33: Particular requirements for the basic safety and essential performance of magnetic resonance equipment for medical diagnosis. In: IEC, editor ed p 36.