Ultra-Homogeneous B0 field for High-Field 
Body Magnetic Resonance Imaging with Unified Shim-RF Coils


 High-field magnetic resonance imaging (MRI, 3.0T and above) offers numerous advantages for imaging the human body over lower-field strengths. However, it suffers from unwanted fast spatially-varying main (B0 ) fields caused by the susceptibility mismatch at the tissue interfaces. When this is combined with the anatomical complexity of the human body, undesirable image artifacts can become damaging as they can compromise potential image contrasts, limit the use of accelerated imaging, and interfere with clinical interpretation. Consequently, these limitations restrict the effective utilization of high-field body MRI and emphasize the need for a major improvement in B0 field homogeneity to take full advantage of the ever increased B0 field. Here we introduce a Unified shim-RF Coil (UNIC) to overcome this existing bottleneck by transcending the conventional low-efficiency, distantly located B0 shim coils. UNIC allows a shim array to be freely allotted and seamlessly integrated into a standard surface RF coil, thus maximizing both the performance of RF receive sensitivity and effective B0 shimming. We demonstrate the capacity of the UNIC approach through detailed characterization of the coil design, prototyping a body coil integrating the UNIC features, and conducting in-vivo imaging of deep organs adjacent to the lung. Our studies provide evidence that UNIC enables homogeneous B0 fields in the liver and the heart, where strong image artifacts are known to occur, and hence facilitate the acquisition of unprecedented image quality in a clinical 3.0T scanner. Further, UNIC’s design is practical as it overcomes one of the most, if not the most, critical limitations of the state-of-the-art high-field MRI with minimal changes to the current MRI hardware architecture. Accordingly, the proposed technique offers opportunities for major advancements in noninvasive imaging of deep organs with high-field imaging in a way it has not been possible thus far.


1
High-field magnetic resonance imaging (MRI, 3.0T and above) offers numerous advantages for 2 imaging the human body over lower-field strengths. However, it suffers from unwanted fast 3 spatially-varying main (B0) fields caused by the susceptibility mismatch at the tissue interfaces. 4 When this is combined with the anatomical complexity of the human body, undesirable image 5 artifacts can become damaging as they can compromise potential image contrasts, limit the use 6 of accelerated imaging, and interfere with clinical interpretation. Consequently, these limitations 7 restrict the effective utilization of high-field body MRI and emphasize the need for a major 8 improvement in B0 field homogeneity to take full advantage of the ever increased B0 field. Here 9 we introduce a Unified shim-RF Coil (UNIC) to overcome this existing bottleneck by 10 transcending the conventional low-efficiency, distantly located B0 shim coils. UNIC allows a 11 shim array to be freely allotted and seamlessly integrated into a standard surface RF coil, thus 12 maximizing both the performances of RF receive sensitivity and effective B0 shimming. We 13 demonstrate the capacity of the UNIC approach through detailed characterization of the coil 14 design, prototyping a body coil integrating the UNIC features, and conducting in-vivo imaging 15 of deep organs adjacent to the lung. Our studies provide evidence that UNIC enables 16 homogeneous B0 fields in the liver and the heart, where strong image artifacts are known to 17 occur, and hence facilitate the acquisition of unprecedented image quality in a clinical 3.0T 18 scanner. Further, UNIC's design is practical as it overcomes one of the most, if not the most, 19 critical limitations of the state-of-the-art high-field MRI with minimal changes to the current 20 MRI hardware architecture. Accordingly, the proposed technique offers opportunities for major 21 advancements in noninvasive imaging of deep organs with high-field imaging in a way it has not 22 been possible thus far. 23 1 Since the advent of magnetic resonance imaging (MRI) in the 1970s, major leaps in image 2 quality and signal-to-noise ratio (SNR) have come from steady increases in strength of the main 3 magnetic field (B0): first from 0.5T and 1.5T in the 20th century, and now to 3.0T and 7.0T in 4 the 21st century 1,2 . However, stronger B0 also comes with increased (and more difficult to 5 correct) field inhomogeneity, which hampers the image quality as homogeneous B0 is imperative 6 for correct spatial representation in MRI images 3 . Although 3.0T scanners have gained high 7 popularity in hospitals, the field inhomogeneity limitations at tissue interfaces in the human body 8 still constitute a central barrier for harnessing the full potential of high-field (3.0T and above) 9 body MRI 4,5 . 10 Magnetic susceptibility mismatch across tissue interfaces can perturb the uniform magnetic 11 field lines and set up fast spatially-varying B0 fields. Since this perturbation scales with the B0 12 field itself, B0 correction (B0 shimming) is particularly challenging at high-field MRI 3-6 . The 13 impact of these field variations is visualized as image artifacts, which can compromise the 14 clinical interpretation of the images. Larger mismatches (e.g., at air-tissue interfaces) are 15 especially challenging and most evident in the images of the heart and liver (particularly at 16 tissue-lung interfaces). For example, 1.5T cardiac MRI is a key modality for determining cardiac 17 function and myocardial tissue characteristics, despite the fact that 3.0T MRI can provide greater 18 capacity (faster image acquisition, novel image contrast, etc.). The inability to fully advance 19 cardiac MRI to 3.0T in the clinical realm is fundamentally tied to the B0 field inhomogeneity at 20 the heart-lung interface. In particular, high SNR readouts using balanced steady-state free 21 precession (bSSFP) 7 , endogenous contrast based on T2* MRI 8 , and fast non-Cartesian readout 22 trajectories 9-11 are unreliable with cardiac MRI at 3.0T. Another example concerns echo-planar 23 imaging (EPI) in body imaging. EPI is frequently relied upon for diffusion MRI 12 , MR 24 spectroscopic imaging (MRSI) 13,14 , and blood-oxygenation-level-dependent MRI (BOLD MRI) 25 15 . These provide information beyond macroscopic morphology, on tissue microstructure, 26 metabolism, and function, offering unique information associated with various pathological 27 states 16 . However, EPI for body imaging at 3.0T is significantly limited by B0 inhomogeneities, 28 which drives image distortion and signal dropout 17 . 29 Hardware-based B0 shim is the most direct approach to rectify B0 inhomogeneity 18-28 . This 30 has been employed in state-of-the-art MR scanners, which utilize 2 nd order spherical harmonic 31 (2 nd SH) shim coils in the magnet bore (Fig. 1A). However, they are distant from the organs of 32 interest and hence ineffective in generating fast spatially-varying shim fields deep in the body. In 33 spite of the recognition of these limitations, this setup has changed little in the past several 34 decades 29 Here we introduce a Unified shim-RF Coil (UNIC) to address the need for far improved B0 3 shimming for high-field body MRI while avoiding the hardware limitations outlined above. 4 Integrating an inherently decoupled surface shim array into a standard RF coil ( Fig. 2A), UNIC 5 maximizes both performances in RF receive and B0 shimming by minimizing the distances 6 between both arrays to the target organs. (Fig. 1A) It allows high freedom of shim loop design 7

Figure 1 Simulations show that unified shim-RF coil (UNIC) provides more effective B0 shimming over the state-ofthe-art second order sphereical harmonic shim. Panel (A)
illustrates the concept of unified shim-RF coil (UNIC) and its advantages over the state-of-the-art scanner equipped in-bore shim coils. A close distance between the shim loops and the target organ can increase the shim efficiency and efficacy by providing a strong and fast spatially-varying shim field to counteract the inhomogeneous field within the target. For magnetic dipoles (shim coils), the magnetic field strength decreases rapidly with distance (D); proportional to D -3 for large distances. When shim coils being placed further away from the target, it would require a dramatic increase of shim current amplitude (I) by I +3 and electric power(P) by P +6 in order to maintain the same field strength. In practice, the requirements for the excessive high current and power cause the shim efficacy to rapidly decrease with the increasing distance. UNIC reduces the coil-target distance by ~5-fold and increases power efficiency by about 4 orders of magnitude and thus improves the shim efficiency and efficacy compared to in-bore shim coils. Panel (B) shows the theoretical simulation of deep tissue shimming with surface shim. The B0 field map of a virtual patient was shimmed by a standard 2 nd order SH in-bore shim coil and 42 channel surface shim coil. The B0 field inhomogeneities under 2 nd SH (black arrows) is substaintially reduced under UNIC shim and a more homogenieous field is achieved in both organs.
for strong, high-order shim field penetration, preserves high RF sensitivity, and can be packed 1 into a compact footprint similar to a standard RF coil. 2 In this study, we prototyped the UNIC coil and demonstrated its utility in body MRI at a 3 state-of-the-art 3.0T scanner. We show that the coil vastly improves the image quality of deep 4 organs that are typically hampered by severe B0 inhomogeneity near the tissue-lung interfaces 5 and successfully reveals pathological lesions that were originally masked by the susceptibility 6 artifacts. 7 8 The Principle of Unified Shim-RF Coil (UNIC) 9 The working principle and circuit diagram of a typical UNIC element is depicted in Fig.  10 2A&B. UNIC is composed of a standard RF receive coil (outer loop) and two shim coils (inner 11 loops). The shim coils are connected into a figure-8 shape circuit using two capacitors C2. RF 12 currents can thus pass through the capacitors and follow a figure-8 path with DC currents 13 restricted to two separate circular loops. Therefore, each of the shim loops can be independently 14 driven by a separate current source. RF currents induced in the symmetric 'figure-8' circuit 15 create magnetic fluxes of equal magnitude but are in opposing directions through the two shim 16 loops, thereby canceling out for the whole circuit. Consequently, the 'figure-8' shim circuit and 17 the RF-receive coil are decoupled, having a 18 negligible mutual inductance. The resonant 19 frequency spectrum of a standard RF coil 20 was measured with and without the figure-8 21 shim loops and presented in Fig. 2D  To investigate the UNIC design, studies were conducted on the bench, in phantoms, and in-2 vivo at 3.0T. First, a detailed characterization of the UNIC's SNR preserving property and the 3 amplified fast spatially-varying shim field components were examined on the bench. 4 Subsequently, we constructed a prototype to validate UNIC's imaging capability in a multi-coil 5 setting. Images of the heart and liver (the two major organs that are exposed to the susceptibility 6 artifacts from tissue-lung interfaces) were acquired with sequences that are clinically important 7 and often suffer from off-resonance artifacts (bSSFP cine in the heart and EPI in the liver). 8 Finally, we explored UNIC's benefits in resolving pathological lesions that were originally 9 hidden by off-resonance induced artifacts. Canine models with hemorrhagic myocardial 10 infarctions(hMI) were studied to image the regional iron deposition within the infarction zone 11 using T2* MRI. These studies showed far improved spatial delineation of hemorrhagic zones 12 that were overwhelmed by the off-resonance artifacts without UNIC shim. All studies were 13 conducted in accordance with the Institutional Review Board and the Institutional Animal Care 14 and Use Committee requirements at Cedars-Sinai Medical Center 15 16 UNIC's coplanar shim and RF coils enable strong, fast spatially-varying shim fields deep in 17 the body without compromising RF performance 18 The UNIC RF receive and B0 shim properties are summarized in Fig. 3. The ratio of unloaded-19 to-loaded quality factors (Q-ratio) was compared in coils with and without UNIC decoupling in 20 panel A. The Q-ratio is a primary parameter of characterizing the RF coil's performance and a 21 main determinant of the image SNR. Two same-size shim loops with 1, 2, and 3 turns were 22 placed in the plane parallel with an RF receive coil. Without UNIC decoupling (naïve coils), Q-23 ratio was substantially compromised. Compared with the RF-only topology, the Q-ratio for 1, 2, 24 and 3-turn shim loops deteriorated, respectively, 71%, 50%, and 51%. On the other hand, with 25 UNIC decoupling, the same metrics were 99%, 97%, and 90% ( Fig. 3A.2). One way of reducing 26 SNR loss from RF coupling is by introducing a buffer distance (gap) between the shim and RF 27 loops 43 . For 2-turn shim loops, Q-ratio variation with buffer distance 0-4 times the shim loop 28 radius Rshim (Gap=0-4 times Rshim) is shown in Fig. 3A.3. Without UNIC decoupling, the Q-ratio 29 was substantially degraded (Q-ratio=50%) when shim loops rested directly on top of the RF loop 30 (Gap=0). At larger buffer distances (Gap>2Rshim), Q-ratio recovered passed 80%, but was still far 31 less than when UNIC-decoupled coils were used, resulting in Q-ratio > 96% for all buffer 32 distances.

33
Shim field strength and its spatial derivatives were compared in a region-of-interest (ROI=1-3 34 times Rshim) 47 for UNIC (Gap=0) versus the naïve coil with 80% Q-ratio (Gap=2Rshim) in Fig.  35 3B. The shim field drops with the inverse third power of the distance. Therefore in the ROI, the 36 UNIC field strength is on average eight times stronger than the naïve field ( Fig. 3B.1). Shim 37 field magnitude spatial derivatives (0 th -6 th order) shown in Fig. 3B.2 are 33-140 times stronger 38 for UNIC than for the naïve coil. Eliminating the need for a buffer distance, UNIC provides 39 strong, fast spatially-varying fields deep in the body for high-efficiency B0 shimming.

13
Prototype UNIC body array provides far improved B0 shimming and comparable RF 14 performance to a state-of-the-art clinical body array coil in phantoms 15 To evaluate UNIC's B0 shim and SNR preserving capability in an array, we evaluated 16 prototypes comprising an in-house built multichannel shim current driver system, 6 RF channels, 17 and 21 evenly distributed two-turn-loop shim coils, against the scanner shimming capability and 18 a similarly-sized standard clinical 6 RF channel body array from the manufacturer (BodyMatrix, 19 Siemens Healthcare). 20 First, absent shim currents, RF receive performances were compared using phantom scans. Inter-21 element coil correlation matrices and SNRs are shown in Fig. 4A. In Fig. 4A.1, the UNIC array 22 shows image quality comparable to the clinical array. Also, the similar coil correlation matrices 23 and SNR histogram distributions show UNIC's shim-RF integration did not noticeably degrade 24 RF reception. 25 Next, we introduced local B0 perturbation with a metallic object (MO), a copper screwdriver 26 ( Fig. 4B.1), to test the B0 shimming capability of the developed coils. Images were acquired 27 with a pair of UNIC arrays at the top and bottom of the phantom. A state-of-the-art image-based 28 shimming procedure was adapted to optimize the 2nd SH shim and UNIC shim. 48 B0 field maps 1 and a bSSFP acquisitions, which are commonly used clinically but are highly sensitive to B0 2 field inhomogeneity, were prescribed ( Fig.4.B.2) with and without UNIC shim. Without UNIC 3 shim (scanner shim only), a strong residual off-resonance B0 is implied in the histogram and 4 evident in the field maps (black arrow). The off-resonance voxels resulted in obvious banding 5 artifacts in the corresponding bSSFP image (blue arrow). In contrast, UNIC shimming markedly 6 reduced the off-resonance regions. B0 homogeneity was significantly improved with the off-7 resonance frequencies' 95% range reduced by half (Scanner shim: -33. 8  UNIC shimming significantly improves B0 homogeneity in the human heart and liver, 1 reducing artifacts and rendering more reliable images in a state-of-the-art 3.0T scanner 2 To validate UNIC's combined shimming/imaging capabilities in-vivo, images of the heart and 3 the liver were acquired in healthy human volunteers (N=8). In the heart, B0 field maps, high-4 resolution bSSFP long-axis anatomical and short-axis cine images (1.0x1.0 mm 2 , commensurate 5 with a prolonged repetition time, TR=5.6ms) were acquired to assess UNIC performance under 6 the highly demanding imaging conditions at 3.0T (Fig. 5). State-of-the-art image-based scanner 7 shim similar to the phantom studies were adopted. Under scanner shim, off-resonance fields 8 were observed in the B0 maps on both posterior and anterior of the heart (black arrows). Banding 9 artifacts at the corresponding locations are shown (blue arrows) and result in significant signal 10 fluctuation in the myocardium. In the short-axis cine images, banding artifacts interacted 11 strongly with the ventricular blood, leading to severe smearing artifacts and corrupting all images

22
We also performed EPI acquisition of the liver to visualize the liver-lung interfaces. Fig. 6  23 shows representative images of the liver in the coronal view from a healthy human volunteer 24 under scanner shim and UNIC shim. Under scanner shim, a strong residual off-resonance field 25 (gray arrows) at the liver-lung interface was evident in all slices, leading to severe artifacts in 26 both EPI and bSSFP images. The EPI images show geometrical distortion and artifactual signal 27 intensity fluctuations in the upper segments of the liver (red arrows). These and the banding 28 artifacts in the bSSFP images (blue arrows) are consistent with the B0 inhomogeneity patterns. In 29 contrast, UNIC shimming successfully corrected the high-order off-resonance pattern at the 30 liver-lung interfaces, substantially reducing artifacts, and leading to improved quality and 31 enhanced clinical utility of both EPI and bSSFP images in all slices. 32 1

Figure 6 UNIC shim markedly improves B0 field in human livers resulting in significant reductions in imaging artifacts at the
2 liver lung interfaces. B0 field maps, bSSFP images, and EPI images in the liver following scanner and UNIC shim are shown.

3
Three slices from the posterior to the anterior of the liver were measured to examine the B0 variations across the whole liver.

4
Following scanner shim, severe B0 inhomogeneity was observed at the liver-lung interface (black arrows). Corresponding bSSFP 5 images with banding artifacts (blue arrows) and EPI images with geometrical distortion (red arrows) in the corresponding regions 6 are also shown. Following UNIC shimming, it was possible to vastly improve the B0 field and hence acquire high-quality liver 7 images.
8 Fig. 7 shows representative histograms of off-resonance B0 in the liver and the heart. B0 maps 9 under central frequency tune-up scanner, 2 nd order SH scanner, and UNIC shim are shown. The 10 95% range for each histogram was measured to characterize and compare the distributions. The 11 central frequency tune-up field distributions were wide due to native field inhomogeneity in both 12 organs. Although 2 nd order SH shim reduced the width, the local fast-varying B0 inhomogeneity 13 at the tissue-lung interfaces remained, reflected in the asymmetric histograms.  UNIC shimming led to significantly tighter B0 field distribution compared to the 2 nd SH scanner shim in the Heart and Liver. Representative histograms of pixel-wise B0 field from the heart and the liver in a human subject are shown, which point to significantly narrower B0 distribution in both the heart and liver following UNIC shim. (Liver: 161Hz (Tune-up) vs. 122Hz (Scanner) vs. 96Hz (UNIC shim); and Heart: 231Hz (Tune-up) vs. 90Hz (Scanner) vs. 57Hz (UNIC)).

UNIC reveals pathological lesions in the heart that were masked by B0 inhomogeneity 1 artifacts under the state-of-the-art scanner shim. 2
To investigate UNIC's ability to improve pathological MRI contrast, canine models with and 3 without hemorrhagic myocardial infarction (hMI) were studied to investigate whether UNIC 4 shimming can improve visualization of focal myocardial iron deposition. 5 Focal iron deposition in the heart caused by hMI can lead to life-threatening adverse events 50-54 . 6 Recent studies have shown that reliable detection of iron deposition post-myocardial infarction 7 can play a crucial role in risk stratification and treatment evaluation 50,51,53 . The ferromagnetic 8 nature of iron has made T2* MRI the desirable noninvasive method to image myocardial iron 9 deposition. However, the susceptibility artifacts at the heart-lung interfaces have made reliable 10 cardiac T2* MRI challenging, particularly at 3.0T 4,55 . Canines (N=3) with and without hMI 11 were imaged before and after UNIC shimming. Fig. 8 shows the T2* images (TE=16.4ms), B0 12 off-resonance map, and late-gadolinium enhanced images (LGE) of the myocardium and the 13 affected territories. T2* images were processed using standard post-processing criteria 8 to 14 identify the iron deposition in the hMI region. The detected hMI territories were highlighted with 15 an olive-color overlay. In all animals, large hypointense artifacts (known as blooming artifacts) 16 at the heart-lung interfaces were evident in the T2*-weighted images under the scanner shim (red 17 arrows). The location of the artifacts corresponded closely to the off-resonance pattern from the 18 B0 maps (red arrows) and compromised the T2* images reliability for identifying the presence of 19 iron deposition. The UNIC shim substantially improved the field homogeneity and T2* image 20 quality. In both healthy and diseased animals, UNIC clearly revealed the left ventricle, originally 21 masked by the blooming artifacts. In animals with hMI, iron deposition in the infarcted territory 22 (blue arrows) was clearly identifiable after UNIC shimming and was delineated consistent with 23 the infarct territory after post-processing. The improved image quality under UNIC shimming 24 eliminated the false positive regions (red arrows) from the state-of-the-art scanner shimming.

25
The results demonstrated UNIC's potential for improving the robustness of T2* images for 26 detecting hMI. 27 28

1
Since the introduction of 3.0T scanners for clinical use two decades ago, there have been high 2 expectations that its SNR advantages and increased spectral separation between chemical 3 substances (e.g., fat, water and metabolites) relative to 1.5T would revolutionize body MRI as it 4 has for neuroimaging. Despite years of efforts and various technical improvements, the adoption 5 of 3.0T body MRI is still hampered by the imaging artifacts. B0 field inhomogeneity is a major, 6 if not the most important, challenge that limits high-field body MRI from advancing to its full 7 potential at 3.0T. The proposed UNIC coil provides an opportunity to impart a strong and fast 8 spatially-varying shim field deep in the body with minimal changes required to the current MRI 9 hardware architecture. Its direct and efficient design allows the utilization of similar mechanical 10 footprints to the standard RF coils, which provides a practical solution for scanner integration. 11 Our studies reveal that UNIC could significantly improve field homogeneity in deep organs that 12 suffer the most from B0 off-resonance artifacts. We demonstrated that adequate B0 homogeneity 13 at a clinical 3.0T scanner is achievable in the heart and the liver to correct for undesirable 14 imaging artifacts, including signal nulling, geometrical distortion, motion smearing, and T2* 15 blooming. The potential of translating the improved image quality for clear visualization of 16 pathological lesions was demonstrated in an animal model with hemorrhagic myocardial 17 infarction.

18
A homogeneous B0 field is critical for the advancement of body MRI at 3.0T and above, 19 particularly for increasing SNR efficiency, enabling image contrasts with advanced functional 20 information, providing reliable quantitative multiparametric maps for tissue characterization, and 21 taking full advantage of image acceleration methods or further improving spatial resolution. 22 Taking cardiac MRI for example, since the heart is constantly moving and located deep in the 23 body, it demands high SNR and short acquisition time. The introduction of the highly SNR 24 efficient bSSFP sequence in the early 2000s 7 revolutionized cardiac MRI at 1.5T. However, due 25 to high sensitivity to B0 inhomogeneity, its use for clinical application at 3.0T has been severely 26 hindered because of inconsistent image quality (banding artifacts and biased relaxation values).

27
It has been reported that with the scanner's 2 nd SH shimming capabilities, the signal variation at 28 1.5T can be reduced to less than 15% with no banding artifacts at the steady-state 6 . However, 29 since the off-resonance frequency doubles at 3.0T 56 , severe banding artifacts can increase the 30 undesirable signal variation to as much as 80% from the truth 57 . These artifacts can be an 31 impediment for fast imaging techniques using non-Cartesian k-space trajectories such as radial 32 and spiral sampling and lead to the inability to deploy accelerated imaging techniques 9-11 . 33 Similar limitations are presented at the liver-lung interface in EPI readouts, which result in 34 misleading geometrical distortion and signal dropouts 58 . Due to these difficulties, GRE 35 acquisitions are usually used at 3.0T for body imaging, which are markedly less SNR/CNR 36 efficient and negate its advantages over 1.5T 55,59 . In addition, inhomogeneous B0 can 37 compromise important image contrast that ties to the higher field strength. For example, T2* and 38 susceptibility contrasts can be enhanced by the increased susceptibility-induced field changes 39 and allow high-field MRI to improve detection of iron deposition 51,60 , myocardial hemorrhage 40 61 , and oxygen consumption 62,63 . Also, the spectral peak separation between metabolites 41 increases proportionally to the field strength, which can boost the CNR in metabolic imaging 42 techniques, including chemical exchange saturation transfer 64 and spectroscopy 65 , and allow 43 better fat-suppression and fat-water separation 5,57 . The amplified B0 inhomogeneity can lead to 44 strong blooming artifacts and spectral corruption, which limits application at high field strengths. 45 Moreover, the growing appreciation of quantitative MRI has further enhanced the importance of 46 B0 field uniformity. For example, accurate myocardial T1 and T2 maps play important role in 1 assessing ischemic and non-ischemic heart diseases 66-69 . Because absolute values are reported, 2 B0 off-resonance can confound the underlying source images and impair the accuracy of the 3 maps and mimic disease 10,70 . We showed that UNIC can reduce the range of B0 off-resonance 4 frequencies by more than 40% in the target organs, which has the potential to translate robust 5 body MRI protocols established at the 1.5T to the 3.0T and allows for exploring the full potential 6 from the increased field strength. 7 Over the years, various attempts have been made to improve B0 homogeneity for high-field 8 MRI. In order to mitigate the loss of efficiency from the cubically decaying magnetic field of 9 magnetic dipoles, methods have been proposed toward bringing the shim coils closer to the 10 subjects to provide stronger shim fields with faster spatial variations. Juchem et al. 43  11 implemented a non-decoupled multi-coil shim array arranged near and with a buffer distance 12 (gap) from the RF receive coil to perform local shimming of the brain. However, the requirement 13 for the buffer distance led to critical limitations (as illustrated in Fig. 3 the dual DC-RF operation in the shared conductor can produce hazardous heat when desired high 20 currents for deep shimming are loaded. It also restricts the shim loop geometry by the 21 corresponding RF loops, which hampered the degrees of freedom in the generated shim patterns. 22 Further, a significant loss in coil Q-ratio and sensitivity was reported 45 by adding extra 23 electronics to the receive loops, which further compromises the image SNR in deep organs. 24 These limitations made the DC-RF setup ineffective in overcoming the challenge of shimming 25 the deep organs. In comparison, UNIC enables strong, deep, high-order B0 shim by eliminating 26 the critical physical limitations associated with DC-RF shim loops, including size, shape, 27 position, the total shim channel number, the loop turns (1-turn), and wire diameter of the shim 28 loops, as well as SNR degradation. 29 In this proof-of-concept study, rigid flat coils were constructed to facilitate the validation of 30 the UNIC theory in a stationary setup. The anterior UNIC coil was unable to closely surround the 31 body contour as a semi-flexible product coil would do, which can create spacing between the 32 coils and the subject at the peripheral of the field of view and result in suboptimal receive and B0 33 shim performances. Yet, owing to the intrinsic decoupling scheme, UNIC is adaptable to semi-34 flexible or flexible RF coils 71-74 , where soft electronics can be employed. A potential 35 complication in using a flexible UNIC coil is that the subject movement may lead to shim field 36 deformation during image acquisition and results in erroneous B0 shim. To correct for the 37 motion-induced error, previously developed motion extraction techniques 10,75 can be adopted to 38 estimate the motion during image acquisition and correct the shim fields. Since the goal of the 39 current study was to build a general-purpose coil, evenly distributed shim loops were adopted. 40 Higher shim efficiency and efficacy for specific applications are expected with further 41 optimization of the coil geometry. 42

Outlook 43
The study addresses a longstanding unmet challenge with 3.0T body MRI by overcoming the 44 primary limitation in B0 homogeneity. The proposed advancement here opens the door for 45 multiple new opportunities in body imaging previously out of reach due to inhomogeneous B0 fields. Notably, it can have a major impact in accelerating the clinical adoption of cardiac MRI 1 and EPI-based body imaging at 3.0T. We envision that the technique is scalable across a range of 2 field strengths (e.g., 3.0T and 7.0T) and organs (e.g., brain, neck, and spine). A numerical 3 simulation of a UNIC 7.0T body array with more shim loop turns is presented in the 4 supplementary material with comparable B0 homogeneity improvement from 3.0T. In addition, 5 UNIC's superior design freedom can enable applications beyond B0 shimming through targeted 6 B0 field shaping. It has the potential of providing a new hardware dimension for image encoding 7 and acceleration 76,77 , B1 correction 78 , and FOV reduction 79 . 8 By adding shimming capability into a conventional RF coil, UNIC functionalizes existing 9 MRI systems with minimal changes required to the current scanner architecture to significantly 10 extend its clinical compatibility. Our findings here are not unlike other historical advances in 11 MRI, which have been enabled by hardware improvements such as multichannel RF detection 31 12 and transmission 80,81 . In the same spirit, we envision that UNIC can be a major advancement 13 over the current (distant) B0 shim coils to facilitate the introduction of a new generation of MR 14 coils with the improved B0 shimming capability and high RF sensitivity.

23
An RF choke inductor (0.6 µH) was soldered into each circular shim loop. The blocking 24 capacitors C2 were each 150 pf. The resonant frequency response was measured using a 25 homemade S12 probe and the Agilent N9923A 4GHz RF vector network analyzer. S12 curve was 26 first measured with the RF loop alone. The UNIC figure-8 shim loop was then overlapped and 27 centered with the RF loop, causing a slight frequency shift to 122.9MHz. The RF loop was tuned 28 back into resonance at 123.2MHz by adjusting the variable capacitor in the loop, and the S12 29 curve was measured again. Coil quality factor (Q) bench measurement: Two sets of 30 measurements were made at the bench to compare the coil quality factors of two shim-RF 31 designs (Fig. 3A), i.e., with and without applying UNIC decoupling. The network analyzer, the 32 9.0 cm circular RF loop, and the UNIC figure-8 shim loops comprising two 2-turn 5.0 cm 33 circular shim loops with an overlapping ratio of 40% were used. For comparison, two turns naïve 34 circular (not decoupled) shim loops were made with the same geometry as the UNIC shim loops 35 (5.0 cm diameter, overlapping ratio of 40% phantom (1% NaCl doped water). The measurement setups were adjusted carefully to ensure a 46 fair comparison between UNIC and naïve coils. Conditions were kept consistent in the 1 comparison, including the S12 probe baseline <-70 dB, the S12 peak (21db), resonance frequency 2 at 123.2 MHz, frequency bandwidths set to 8.0 MHz and 30.0 MHz, respectively, for the 3 phantom unloaded and loaded. The S12 probe location relative to the RF loop and the RF loop 4 position relative to the phantom were kept as nearly identical as possible. Additionally, 5 precautions were taken to ensure that the baseline of the shim loops did not contaminate the S12 6 spectrum. The separation between the shim loops and the RF loop was adjusted by using a small 7 amount of blue tack that had a negligible effect on the measurements. In all cases, the Q was 8 computed as the ratio of the measured resonance frequency divided by the measured frequency 9 bandwidth as 3dB below the resonance peak amplitude. 10 Numerical simulations: 11 Shim field strength and spatial variation: To investigate the shimming performance of UNIC, the 12 strength and spatial variation of the generated shim fields were simulated using the Biot-Savart 13 law for both UNIC coils and naïve coils (Fig. 3). The fields were simulated on an axis 14 perpendicular to the plane of the circular 2-turn shim loops at the center of symmetry of the coils 15 (central axis) and 1A shim currents. To compare coils with comparable RF performance, the 16 naïve coil was adopted with Gap=2Rshim (Q-ratio>80% in Fig. 3A). The shim field strength from 17 each coil was simulated along the central axis for distances from zero up to 6Rshim away from the 18 coil plane. Field strength was normalized by the maximum field from the naïve coil. The 19 distance in the simulations was presented as multiples of the shim loop radius (Rshim). The spatial 20 variability of the shim fields was evaluated with the spatial derivatives (0-6 th order) of the 21 simulated shim fields. The mean magnitude of the field's spatial derivatives in a defined ROI (1-22 3 Rshim from the coil surface) 47 were computed. The ratio between the UNIC and naïve coils for 23 each derivative order (UNIC (n) ÷Naïve (n) ) was reported to evaluate the amplification of high-24 order components with UNIC (Fig. 3). Susceptibility mismatch induced off-resonance frequency 25 map in the human body: Theoretical field inhomogeneity was evaluated using a numerical 26 simulation platform (Medical Interactive Creative Environment; MICE) 82 . Details of the 27 simulation are described in the supplementary material. In our study, a 3D volume covering the 28 torso of a human subject was created with comparable imaging parameters to the in-vivo scans. 29 To define the ROIs for evaluating B0 homogeneity in the target organs, contours of the heart and 30 liver were depicted based on the input images. The root mean square error (RMSE) and median 31 (95% range) of the B0 were measured in the heart and the liver and reported in the supplementary 32 material. All simulations were performed with in-house developed Matlab (Mathworks 2016R) 33 scripts. The simulation setups were applied for a 3.0T and 7.0T to examine the applicability of 34 UNIC in an ultra-high field environment and presented in the supplementary material. At 7.0T, 35 higher shim loop turns (n=3) were adopted to accommodate the amplified B0 inhomogeneity. 36 Prototype UNIC coil and hardware development: 37 Coil design, shim-RF loop layout, and control system: A unified shim-RF coil system for body 38 imaging was built from the ground up. This consisted of a UNIC coil array and a shim current 39 driver and control system. The UNIC array was constructed with two flat sub-arrays, top and 40 bottom, each having 6 RF receive channels and 21 shim channels (Fig. 4A.1), for a total of 12 41 RF and 42 shim channels. The subject would position between the two flat sub-arrays, adjustable 42 in height to closely fit the subject for maximizing RF sensitivity and shim efficiency. The RF 43 array layout was similar to that of a conventional six-channel commercial body array (size 45.0 44 cm × 30.0 cm). Two-turn shim loops (8.5 cm diameter, wire diameter 1.0 mm) were employed 45 for the first prototype, providing a sufficient field strength and reducing the maximum current to 5.0 A for minimal heating. For each of the top and bottom arrays, a total of six shim-RF coil 1 clusters were arranged in 2 rows of 3 clusters. Each cluster contains a 1-ch RF receive (Rx) loop 2 and a 2-ch 2-turn UNIC figure-8 shim loop. The Rx loop is overlapped with the figure-8 shim 3 loop. Additionally, either 4 or 5 shim loops skirt the edge of either row (Fig. 4A1). Between 4 nearest neighboring Rx loops, geometrical decoupling was applied. In addition, preamplifier 5 decoupling was used to decouple non-nearest neighboring Rx loops. Between the overlapping 6 shim and RF loops, the figure-8 geometrical decoupling was used (Fig. 2) in addition to RF 7 chokes to eliminate potential interference between shim loops and wires and their surrounding 8 Rx loops. The outermost nine shim loops had partial overlapping and were thus geometrically 9 decoupled from the respective Rx loops. Coil construction: The RF coil construction procedure 10 was similar to that of building a conventional RF array, which includes unloaded-to-loaded 11 quality factor ratio (Q-ratio) measurements during each step, tuning for active decoupling, 12 adjusting for preamplifier decoupling, tweaking for critical overlap geometrical decoupling, 13 array assembly, tuning and matching, and final bench touch up. The coil loading procedure used 14 a torso phantom mimicking the dielectric properties of the human body. In addition, for shim current-induced ohmic heating, we measured the temperature rise in the 43 shim circuitry on the bench by running the maximum shim currents 5.0 A through individual 44 channels. We also monitored the temperature increase within the entire array with 5.0 A in all 45 channels inside the magnet. The hot spot had a maximum temperature increase of less than 7 °C. 46 Imaging studies: 1 All imaging was performed at a 3.0T clinical scanner (Biograph mMR, Siemens Healthcare, 2 Erlangen, Germany). Image-based shimming was implemented with shim box selection to match 3 the target ROIs. Data analysis, numerical simulation, and image processing were done using 4 customized Matlab scripts. Imaging parameters from all MRI sequences are summarized in the 5 supplementary materials. 6 Phantom scans: To test the RF performance of the prototype array under a controlled 7 environment, experiments were performed in phantoms. A standard American College of 8 Radiology (ACR) phantom (cylindrical, 20.3cm x 17.3cm) and a larger body phantom 9 (rectangular, 40.0cm x 30.0cm x 20.0cm) were used to test the arrays' structure imaging ability 10 and spatial coverage. The constructed array was compared to a six-channel standard surface 11 array (BodyMatrix. Siemens Healthcare, Erlangen, Germany). Arrays were placed over the 12 phantoms using a rectangular rack to support the arrays and to keep the experimental setup 13 identical. The prototype and standard surface array were set up in consecutive imaging sessions. 14 After localization, images were acquired with a 3D proton density-weighted(PDW) gradient 15 recall echo (GRE) sequence to evaluate the SNR for each coil. PDW images were acquired with 16 the clinical RF array, the prototyped RF only array, and the prototyped UNIC array. SNR maps 17 were derived using the difference from two consecutive measurements 84,85 . The SNR maps from 18 different arrays were compared using histogram analysis. To further test the noise characteristics 19 of the prototype arrays, noise covariance information was acquired using the same GRE 20 sequence without the excitation pulse. The coil noise correlation matrix was calculated for the 21 clinical RF array and the UNIC array 35 . To examine the B0 shimming capability of the 22 constructed UNIC prototype, a metallic off-resonance source (MO; a copper screwdriver) was 23 placed in the field of view (FOV) to emulate the high-order field inhomogeneity generated from 24 the tissue interfaces. The screwdriver was placed 5.0cm away from the ACR phantom to induce 25 a high-order off-resonance pattern. B0 maps and clinical bSSFP images were acquired using both 26 the scanner shim alone and with UNIC shim to probe the B0 shimming and imaging capability of 27 the constructed array. During scanning, the scanner 2 nd SH B0 shim and UNIC shim were applied 28 subsequentially. First, scanner shim was applied to the ROIs with a standard image-based 29 shimming procedure 48 . Shim volume was defined with a manually selected shim box to perform 30 shimming. Following the scanner shim and image acquisitions, UNIC shim was applied on top of 31 the scanner shim. The same shim volume and shimming process were used to derive the UNIC 32 shim field. The UNIC shim currents were calculated using linear least-square optimization. The 33 shim current for each UNIC channel was driven by a separate channel of a DC control system. 34 The optimization was constrained by maximum DC current ±5.0A per coil. B0 field maps and 35 bSSFP Images were acquired under each shimming condition. To compare the shimming 36 capability, a pixel-wise off-resonance field map was derived under each shimming condition. 37 Magnitude images from the sequence were used to define the ROIs for B0 field analysis. Since 38 the field distribution is mostly skewed under the influence of the MO, histogram, RMSE, 39 median, and 95% range of the off-resonance maps were all measured. 40 In-vivo human scans: Healthy human volunteers (N=8; five female) were recruited to test the 41 ability of the constructed shim array in deep organs. Subjects were consented and the studies 42 were conducted in accordance with the Institutional Review Board requirements at Cedars-Sinai 43 Medical Center. The volunteers were scanned in the head-first supine position with the UNIC 44 double array (top and bottom). Localizers were acquired to determine the imaging planes and 45 shim volumes for B0 shimming. The heart and liver were scanned in a randomized order in all 46 subjects. Before image acquisition, the scanner table was adjusted to move the target organs to 1 the center of the bore to facilitate optimized B0 field homogeneity. Image Acquisition: The B0 2 field maps were acquired in the target organs before and after applying the UNIC and scanner 3 shims. In addition, off-resonance sensitive sequences (bSSFP and EPI) were acquired to capture 4 the image quality differences under all shimming conditions. In the cardiac scans, high resolution 5 (1 x 1mm 2 ) bSSFP cine images with extended TR (5.6ms) were acquired to evaluate the image 6 quality of the heart. Three short-axis slices (basal, mid, and apical) were acquired and analyzed. 7 In the liver scans, an EPI sequence and a bSSFP sequence were acquired. The images were 8 acquired in the coronal view to illustrate the off-resonance effect at the liver-lung interfaces. 9 Three imaging slabs with six slices each were used to cover different sections of the liver. In all 10 bSSFP images, frequency scouts were performed to optimize the image quality. Identical 11 imaging parameters were applied under the scanner shim and UNIC shim. All images were 12 acquired under breath holds. A similar shimming procedure as described in the ex-vivo studies 13 was used in the in-vivo studies. Shim volumes matching the image position were adopted for 14 each image acquisition. To minimize breathing motion-induced field disturbance, all B0 15 shimming and imaging were performed at the end-expiration. The field of view and slice 16 coverage of the field maps were adjusted to cover the human liver and heart in each subject.

17
Image analysis: To evaluate B0 homogeneity in the target organs, pixel-wise off-resonance maps 18 were derived with the same method described for the phantom study. Contours of the liver and 19 the heart were delineated using the magnitude images to refine the ROIs. The RMSE, median, 20 and 95% range of the off-resonance field in the target organs were derived to assess shimming 21 performances. All B0 fields were tuned up to the central frequency of the target volume. 22 Quantitative comparison of scanner-and UNIC-shimmed fields were made in all volunteers. To 23 evaluate the quality of the images obtained with clinically relevant sequences, the image quality 24 scores were measured for the bSSFP images in both organs and EPI images in the liver. The 25 consensus was reached by two readers with 3+ years of experience. Image quality scores were 26 assigned according to the extent of artifacts using a 5-point ordinal scale as follows: 1 = 27 unreadable images, 2 = severe image artifacts, 3 = moderate image artifacts, 4 = minor image 28 artifacts, 5 = no image artifacts. 29 In-vivo animal scans: Dogs (n = 3, 22 to 26 kg) were studied with and without surgically 30 induced myocardial infarction (MI test the influence of the blooming artifacts in detecting local myocardial iron deposition, T2* 46 images were analyzed with a standard procedure to identify hemorrhagic infarction 8 . Endocardial 1 and epicardial contours were first delineated in the T2* images. Remote myocardium was 2 identified as the region that showed no hyperintensity in the LGE images. The hemorrhagic 3 territories were identified in the myocardium with signal intensity 2 SD lower than the mean 4 signal intensity in the remote ROI. 5 6 Statistical analyses: 7 Differences between the RMSE, Median, and 95% range of the off-resonance fields were 8 analyzed using paired sample t-tests, and the image quality scores were compared using the 9 Wilcoxon rank-sum test. Value of P<0.05 was considered statistically significant. Normality was 10 checked by the Shapiro-Wilk test.