Tao Zhang¹, Michael Lustig^1,2, Shreyas Vasanawala³, and John Pauly^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Electrical Engineering and Computer Science, University of California Berkeley, Berkeley, CA, United States, ³Radiology, Stanford University, Stanford, CA, United States

Array compression is a technique to reduce data size and reconstruction computation for large coil arrays. In this work, a data-driven array compression for 3D Cartesian sampling is proposed. A slice-by-slice array compression method with autocalibrating parallel reconstruction using 3D synthesis kernels is designed. Faster reconstruction and similar image quality is achieved compared with reconstruction results using the original large arrays.

Philip James Beatty¹, Atsushi Takahashi², Kevin M Johnson³, and Jean H Brittain^4
1Global Applied Science Laboratory, GE Healthcare, Thornhill, Ontario, Canada, ²Global Applied Science Laboratory, GE Healthcare, Menlo Park, California, United States, ³Medical Physics, University of Wisconsin - Madison, Madison, Wisconsin, United States, ⁴Global Applied Science Laboratory, GE Healthcare, Madison, Wisconsin, United States

Feng Huang¹, Wei Lin¹, George Randy Duensing¹, and Arne Reykowski^1
1Invivo Corporation, Gainesville, Florida, United States

In MRI, imaging using receiving coil arrays with a large number of elements is an area of growing interest. With increasing channel numbers, longer reconstruction times have become a significant concern. Channel compression has been proposed to reduce the processing time. However, channel compression technique has to balance speed and preservation of signal. In this work, a novel technique using relative sensitivity maps is proposed for faster channel-by-channel compressed sensing. The proposed method is much faster than conventional channel compression technique, and preserves the signal significantly better.

Feng Huang¹, Wei Lin¹, George Randy Duensing¹, and Arne Reykowski^1
1Invivo Corporation, Gainesville, Florida, United States

The combination of partially parallel imaging (PPI) and compressed sensing has shown great potential for fast imaging. Fourier transform of the partially acquired data is conventionally used as the initialization of the iterative reconstruction algorithm. A good initialization is crucial for the convergence speed and accuracy of iterative algorithms. In this work, it is proposed to use GRAPPA operator to efficiently generate initialization for multi-channel compressed sensing. Using self-feeding Sparse SENSE as a specific example of multi-channel compressed sensing algorithm, experimental results show the advantages of the proposed method over conventional scheme.

Daniel Stuart Weller¹, Jonathan R Polimeni^2,3, Leo Grady⁴, Lawrence L Wald^2,3, Elfar Adalsteinsson¹, and Vivek Goyal^1
1EECS, Massachusetts Institute of Technology, Cambridge, MA, United States, ²A. A. Martinos Center, Dept. of Radiology, Massachusetts General Hospital, Charlestown, MA, United States, ³Dept. of Radiology, Harvard Medical School, Boston, MA, United States, ⁴Dept. of Image Analytics and Informatics, Siemens Corporate Research, Princeton, NJ, United States

SpRING combines compressed sensing (CS) with GRAPPA to recover a sparse image from multi-channel, undersampled k-space data. The combined method operates in the nullspace of the observation matrix, holding the acquired data fixed without resorting to complicated procedures for constrained optimization. Whereas GRAPPA amplifies the noise present in the image and CS over-smoothes non-sparse details, the combined method strikes a balance to improve SNR and preserve details. We analyze the noise amplification properties of the combined algorithm using g-factors computed using Monte Carlo trials to illustrate its superiority over GRAPPA and CS alone for real data.

Kang Wang¹, Philip Beatty², James Holmes², Reed Busse², Jean Brittain², and Frank Korosec^1
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ²Global Applied Science Laboratory, GE Healthcare

This abstract presents a framework that combines compressed sensing (CS) with auto-calibration parallel imaging (acPI) reconstruction for undersampled Cartesian acquisition. In data acquisition, a two-step undersampling scheme is used. For reconstruction, an acPI method is integrated into the CS L1 norm minimization process, such that both the coherent and incoherent aliasing artifacts associated with the undersampling can be suppressed in the iteration. The feasibility of the new method was validated using 3D contrast-enhanced peripheral MR angiography data sets

Anja Brau¹, Peng Lai¹, Srihari Narasimhan², Babu Narayanan³, and Vijaya Saradhi^2
1Global Applied Science Laboratory, GE Healthcare, Menlo Park, CA, United States, ²Computing & Decision Sciences Lab, GE Global Research, Bangalore, India, ³Medical Image Analysis Lab, GE Global Research, Bangalore, India

As part of the calibration step for Compressed Sensing & Parallel Imaging algorithms like ESPIRiT and L1-SPIRiT, computation of kernel weights involves obtaining a least squares fit for predicting target points in the calibration region using a set of source points in their neighborhood, separately for each coil. If we do not use the coil neighbors of the target location, the linear system needs to be solved only once. We observe that using this approach, we get significant computational benefit and still obtain similar image quality for high channel count reconstructions.

Mariya Doneva^1,2, Peter Börnert¹, and Alfred Mertins^2
1Philips Research Europe, Hamburg, Germany, ²University of Luebeck, Luebeck, Germany

The synthesis of CS and parallel imaging is of consierable interest. Several works have proposed a combination of SENSE and CS as an iterative sparsity constrained SENSE reconstruction. Typically a variable density pseudo-random sampling is used as a compromise between regular and irregular sampling. Based on the properties of CS and SENSE, we propose a two-step CS-SENSE reconstruction, in which the two reconstruction steps are used to recover distinct parts of k-space and apply a two-level variable density sampling pattern.

satoshi Ito¹, Hirotoshi Arai¹, and Yoshifumi Yamada^1
1Research Division of Intelligence and Information Sciencs, Utsunomiya University, Utsunomiya, Tochigi, Japan

In this paper, we propose a novel image reconstruction method in which CS and PI are executed using only single set of signal. Since the distribution of PSFT signal strongly reflects the distribution of the object, the application of a weighting function to the PSFT signals has a similar effect as the application of the weighting function to the object. Therefore SENSE reconstruction using a single signal is feasible by producing another folded image having another weighting function. Here, we propose a new imaging method which combine CS and PI using single signal to achieve higher reduction factor.

Ching-hua Chang¹, and Jim Ji^1
1Texas A&M University, College Station, Texas, United States

The combination of Compressed Sensing (CS) with multiple receiver coils has drawn great attentions because of their potentials to significantly reduce acquisition time in MRI. The former emerged by exploring the sparsity of MR images, and the latter can reduce scan time by resorting to parallel MRI (pMRI). CS can be used as regularization method and is integrated to take advantages of the sensitivity information from multiple receiver coils. In this abstract, we propose to use a reweighted nonlinear conjugate gradient L1 minimization method to enhance the reconstruction of parallel CS-MRI.

Feng Huang¹, Wei Lin¹, George Randy Duensing¹, and Arne Reykowski^1
1Invivo Corporation, Gainesville, Florida, United States

In MRI, imaging using receiving coil arrays with a large number of elements is an area of growing interest. With increasing channel numbers for parallel acquisition, longer reconstruction times have become a significant concern. Channel compression, direct virtual coil (DVC) and synthetic coil (ST) have been proposed to reduce the processing time of channel-by-channel partially parallel imaging (PPI) techniques [4]. In this work, a technique called flexible virtual coils (FVC) is proposed to enable faster and more accurate reconstruction than these existing techniques.

Philip James Beatty¹, James H Holmes², Scott B Reeder³, and Jean H Brittain^2
1Global Applied Science Laboratory, GE Healthcare, Thornhill, Ontario, Canada, ²Global Applied Science Laboratory, GE Healthcare, Madison, Wisconsin, United States, ³Departments of Radiology and Medical Physics, University of Wisconsin - Madison, Madison, Wisconsin, United States

Combining parallel imaging with reduced field-of-view acquisitions is challenging since such acquisitions violate underlying assumptions of many parallel imaging reconstruction methods. Direct Virtual Coil (DVC) parallel imaging is a data-driven technique that delivers similar image quality to coil-by-coil approaches, such as GRAPPA, while reducing memory and compute requirements by over 10X for high channel count coil arrays. This work compares images reconstructed with DVC channel combination and sum-of-squares (SoS) channel combination, revealing subtle differences within the wrapped region.

Philip James Beatty¹, Ananth Madhuranthakam², Shaorong Chang³, Ersin Bayram³, and Jean H Brittain^4
1Global Applied Science Laboratory, GE Healthcare, Thornhill, Ontario, Canada, ²Global Applied Science Laboratory, GE Healthcare, Boston, Massachusetts, United States, ³GE Healthcare, Madison, Wisconsin, United States, ⁴Global Applied Science Laboratory, GE Healthcare, Madison, Wisconsin, United States

Partial k-space acquisition is an important technique to reduce scan time or shorten echo time and is often used together with parallel imaging. In this work, we combine a partial k-space acquisition with the Direct Virtual Coil (DVC) parallel imaging reconstruction method. In vivo imaging results compare coil-by-coil reconstruction results to DVC.

Meihan Wang¹, Weitian Chen², Micheal Lustig³, Peng Hu⁴, Michael Salerno⁵, Christopher Kramer⁵, and Craig Meyer^1
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, ²GE Healthcare, ³UC Berkeley, ⁴University of California, Los Angeles, ⁵University of Virginia

Synthetic Target (ST) is a rapid self-calibrated parallel reconstruction method. The original ST works well for spiral and Cartesians, yet suffers from aliasing for radial trajectories. In this abstract, we improved our original Synthetic Target algorithm by adding a phase constraint in the training process. The resultant images have higher SNR than the original ST for radial trajectories. We also compared our method with another iterative parallel reconstruction algorithm in terms of image quality and computation time.

Michael John Allison¹, and Jeffrey A Fessler^1
1Department of Electrical Engineering and Computer Science, The University of Michigan, Ann Arbor, Michigan, United States

Accurate coil sensitivity estimates are required to avoid artifacts in many parallel imaging techniques. Existing regularized methods provide sufficiently accurate estimates, but often at a high computational cost. We propose an iterative technique that uses Augmented Lagrangian principles to efficiently compute sensitivity estimates. Our method generated sensitivity estimates for a challenging breast phantom dataset in half the time required by a conjugate gradient algorithm. We therefore conclude that our AL approach provides an efficient strategy for accurately estimating coil sensitivities.

Simon Bauer¹, Bernd Andre Jung¹, Alexey A Samsonov², Matthias Honal¹, and Michael Markl^1
1Department of Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany, ²Department of Radiology, University of Wisconsin, Madison, Wisconsin, United States

Parallel imaging is based on the measurement of the MR signal with multiple receiver coils, where each coil modulates the signal with its sensitivity profile. The motion of an object can also be seen as a modulation of the object. Therefore, a time resolved measurement of a moving object can be considered as an acquisition with multiple (virtual) receiver coils along the temporal dimension. In this work we used a phantom and an in-vivo measurement to compare GRAPPA to a parallel imaging reconstruction which uses a combined virtual coil array consisting of the coil array and all time frames.

Florian Knoll¹, Kristian Bredies², and Rudolf Stollberger^1
1Institute of Medical Engineering, Graz University of Technology, Graz, Austria, ²Institute for Mathematics and Scientific Computing, University of Graz, Graz, Austria

Constrained iterative image reconstruction of undersampled data from multiple coils has shown its high potential to deliver images with excellent image quality from highly accelerated measurements. To eliminate aliasing artifacts, regularization methods are facilitated, which introduce a-priori knowledge about the structure of the desired solution. Usually, regularization is applied only in 2D, and data is reconstructed slice by slice. While this approach reduces the size of the problem and therefore the amount of memory that is needed in the computation, it neglects the potential of introducing a-priori information in the third dimension. This work introduces an approach which treats a whole 3D volume of images as a single data set, and also includes 3D regularization. Results are presented for undersampled spiral angiography data from the 2010 ISMRM image reconstruction challenge.

Shuo Feng¹, and Jim Ji^1
1Texas A&M University, College Station, Texas, United States

Large arrays make it possible to use parallel imaging to significantly accelerate MR imaging speed. However, the need for auto calibration signals (ACS) limits the actual acceleration factors achievable with large arrays. This paper presents a novel method for parallel imaging with large arrays. The method uses Sinc kernels for interpolation that only requires one phase parameter to be estimated. Phantom experiments using a 64-channel system show that the new method provides higher acceleration factors, comparable reconstruction quality, and fast reconstruction speed.

Yiqun Xue¹, Jiangsheng Yu¹, Hyun Seon Kang¹, Sarah Englander¹, Mark A Rosen¹, and Hee Kwon Song^1
1Department of Radiology, University of Pennsylvania, Philadelphia, PA, United States

In radial MRI, streaking artifacts arising from regions of bright signal intensity can contaminate an entire image. The situation is exacerbated in large FOV imaging using coil arrays where the high sensitivities of local coils can detect signal from peripheral regions of the FOV in which gradient field distortions often cause signal intensities to become highly concentrated. In this abstract, we describe an algorithm which can automatically identify those coils whose images contain streaking artifacts and by excluding these coils, we demonstrate marked improvements in image quality.

Gigi Galiana¹, Jason Stockmann², Leo K. Tam², and Robert Todd Constable^1,2
1Diagnostic Radiology, Yale University, New Haven, CT, United States, ²Biomedical Engineering, Yale University, New Haven, CT, United States

Since surface coil profiles are not usually arranged for optimal encoding along the undersampled direction, we address this with a sequence that uses RF excitation to create virtual coil profiles. These virtual coil profiles can be chosen to optimally complement the undersampled data, and previous work showed that encoding with bar-like RF profiles significantly improved the achievable acceleration of Cartesian undersampled data. With the increasing availability of high powered nonlinear gradient sets, other shapes are possible that can better complement the azimuthal arrangement of multichannel coils. Exploiting this flexibility, we present an implementation of ring-like RF profiles (BARings) and the acceleration enhancements they can bring to multichannel radial acquisitions.

Kie Tae Kwon¹, Holden H Wu^1,2, Michael Lustig^1,3, and Dwight G Nishimura^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Cardiovascular Medicine, Stanford University, Stanford, CA, United States, ³Electrical Engineering and Computer Science, University of California at Berkeley, Berkeley, CA, United States

The 3D concentric cylinders trajectory is an efficient and robust trajectory with fewer excitations than a comparable 3DFT trajectory and benign off-resonance effects. In this work, we applied an efficient parallel imaging strategy to 3D concentric cylinders, which decomposes the 3D non-Cartesian parallel imaging reconstruction into a series of 2D Cartesian reconstructions. Combined with the centric-ordered sampling scheme, this trajectory can be used to efficiently capture transient responses in a magnetization-prepared sequence.

Daniel Neumann¹, Felix A. Breuer¹, Peter M. Jakob², Gregory Lee³, Mark A. Griswold³, and Nicole Seiberlich^3
1Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Germany, ²Dept. of Experimental Physics 5, University of Würzburg, Würzburg, Germany, ³Dept. of Radiology, Case Western Reserve University, Cleveland, Ohio, United States

Non-Cartesian imaging has many known advantages over Cartesian imaging; however, complex parallel imaging techniques are required for such non-standard trajectories. In this work, we first use Calibrationless Parallel Imaging (CPI) to iteratively reconstruct fully-sampled and undersampled variable density spiral data. This method requires no explicit calibration data or coil sensitivity maps and offers robust reconstructions using in-vivo spiral data for acceleration factors up to 4. Our results suggest that CPI with GROG may overcome the problems of reconstructing images from accelerated non-Cartesian measurements using parallel imaging.

Jian Zhang¹, and Ajit Shankaranarayanan^2
1Global Applied Science Lab, GE HealthCare, Bethesda, MD, United States, ²Global Applied Science Lab, GE HealthCare, Menlo Park, CA, United States

Pseudo-Cartesian GRAPPA in conjunction with GROG (PCG-GROG) is an effective algorithm to reconstruct non-Cartesian parallel imaging data. However, the PCG reconstruction is usually a cumbersome task as different fitting patterns need to be determined point by point. PRUNO is a generalized auto-calibrated reconstruction algorithm, in which local nulling kernels are independent of the acceleration factor and undersampling patterns. We can thus replace the PCG step with PRUNO to bring out a more convenient and accurate algorithm, termed PRUNO-GROG. Promising preliminary results are shown on both phantom and in vivo images.

Thomas Oberhammer¹, Markus Weiger^2,3, Franciszek Hennel³, and Klaas Paul Pruessmann^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland, ²Bruker BioSpin AG, Faellanden, Switzerland, ³Bruker BioSpin MRI GmbH, Ettlingen, Germany

ZTE is an MRI technique with 3D radial centre-out encoding and zero echo time, particularly suited for imaging samples with short T2. ZTE data is inherently incomplete in the k-space centre, which can be addressed by algebraic reconstruction, however, possibly entailing noise amplification and correlation. In this work, the principles of parallel imaging are applied for improving the noise behaviour in ZTE imaging. By means of 1D and 2D simulations it is demonstrated, that, beyond the original purpose of accelerated data acquisition, sensitivity encoding is also capable of handling the ZTE-specific problem of incomplete sampling in central k-space.

Nicholas Ryan Zwart¹, and James Grant Pipe^1
1Neuroimaging Research, Barrow Neurological Institute, Phoenix, Arizona, United States

Parallel image reconstruction methods synthesize data to replace undersampled or non-sampled gaps in k-space. The SENSE parallel imaging algorithm is generalized for the reconstruction of non-cartesian k-space trajectories through the use of a gridding step within each iteration. In an effort to reduce the number of computations, a method of masking k-space was introduced as a replacement for this step. This work compares the two techniques.

Kenneth Otho Johnson¹, and Craig H Meyer^1
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States

Pretty Easy Parallel Imaging (PEPI) provides a fast framework for non-Cartesian reconstruction of undersampled data. Here a phase constraint is added that helps reduce aliasing artifacts and improve convergence.

Leo K Tam¹, Jason P Stockmann¹, Gigi Galiana², and Robert Todd Constable^1,2
1Biomedical Engineering, Yale University, New Haven, Connecticut, United States, ²Diagnostic Radiology & Neurosurgery, Yale University, New Haven, Connecticut

Recent advances in accelerating MRI scans include compressed sensing and the utilization of nonlinear magnetic encoding fields. The two methods collect data in a targeted manner and disperse residual aliasing artifacts to make them less apparent. In the current work, a parallel nonlinear magnetic encoding method using first and second order in-plane spherical harmonics, Null Space Imaging (NSI), is combined with a compressed sensing algorithm to reconstruct highly accelerated images with fidelity. The work suggests that compressed sensing and parallel imaging with higher order gradients may be a synergistic approach towards robust reconstructions of accelerated scans.

Shr-Tai Liou¹, Hsiao-Wen Chung¹, Wei-Tang Chang², Wen-Kai Tsai², and Fa-Hsuan Lin^2,3
1Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan, Taiwan, ²Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan, Taiwan, ³MGH-HST Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States

Single-shot volumetric MR inverse imaging (InI) can achieve 100 ms temporal resolution and 5-10 mm spatial resolution with the whole head coverage. Different approaches have been proposed to reconstruct InI images with high spatial resolution or high computational efficiency. Here we propose the virtually independent parallel Gaussian channel nulling (VIPGen) reconstruction. The group analysis in our visuomotor experiments shows that VIPGen provides good spatial resolution similar to the sources reconstructed by the eigenspace L1 norm minimization and high statistical values, while the computing time is comparable to the L2 norm minimization.

Berkay Kanberoglu¹, Lina J. Karam¹, and David Frakes^1,2
1School of Electrical, Computer and Energy Engineering, Arizona State University, Tempe, AZ, United States, ²School of Biological and Health Systems Engineering, Arizona State University, Tempe, AZ, United States

An alternative method to ABSINTHE (Atlas based sparsification of image and theoretical estimation) is proposed. ABSINTHE achieves sparsification by performing a principle component analysis (PCA) on the aliased undersampled image. Better sparsification can be achieved by using K-SVD. K-SVD provides flexibility of dictionary design parameters which can be important for the image approximation. The proposed method shows that K-SVD is able to reconstruct similar quality images in comparison to the traditional ABSINTHE method while using half the number of basis images required by PCA.

Eric Pierre^1,2, Nicole Seiberlich³, Vikas Gulani⁴, Pierrick Bourgeat², Olivier Salvado², and Mark Griswold^3
1Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States, ²ICT Centre, CSIRO, Brisbane, QLD, Australia, ³Radiology, Case Western Reserve University, Cleveland, Ohio, United States, ⁴Radiology, Case Western Reserve University, Cleveland, United States

In order to improve the GRAPPA reconstruction of an undersampled object at high reduction factors, the previously introduced ABSINTHE technique required a large training set of MR signal with matching coil configuration. This study seeks to further increase the effectiveness and applicability of ABSINTHE by allowing the addition of any MR image to the training set regardless of its coil configuration. Furthermore, it introduces key image preprocessing steps which noticeably increase the relevance of each image in the training set. An improved image quality is shown in simulated and in vivo data compared to GRAPPA and the previous ABSINTHE technique.

Jun Shen^1
1NIMH, Bethesda, Maryland, United States

A new method is introduced here for parallel magnetic resonance imaging reconstruction that is based on the concept of two-step spectral editing in magnetic resonance spectroscopy (MRS). It uses a unitary 2×2 Fourier matrix (1 1;1 -1) for encoding and decoding. Although image editing is performed in the image domain, it does not suffer from the deficiency associated with SENSE when the prescribed field of view in phase-encoding direction is smaller than the region occupied by the object. Instead, the appearance of the aliasing in the reconstructed image is exactly the same as in the unaccelerated image.

William A Grissom¹, Laura Sacolick¹, and Mika W Vogel^1
1GE Global Research, Munich, Germany

3D parallel spokes excitation pulses have been proposed for compensating flip angle inhomogeneities induced by inhomogeneous B1+ at high field. While flip angle homogeneity improves with increasing spokes pulse duration and an increasing number of parallel excitation channels, competing desires to limit off-resonance sensitivity and system complexity limit these pulses' effectiveness in practice. We show that nonlinear gradient phase encoding is well-suited to this application and that simultaneous linear and nonlinear phase encoding can significantly improve spokes pulse performance, enabling shorter pulse durations and fewer transmit channels for a desired level of flip angle homogeneity.

Chao Ma¹, Kevin F. King², Dan Xu², and Zhi-Pei Liang^1
1Electrical and Computer Engineering, University of Illinois, Urbana, Illinois, United States, ²Global Applied Science Lab, General Electric Healthcare, Waukesha, Wisconsin, United States

This paper presents a method to design RF pulses for reduced FOV excitation using second-order gradients and spatial-spectral pulses. By taking advantage of the spatial dependence introduced by the second-order gradients, a thin disk can be excited using a 2D spatial-spectral pulse, i.e. covering 2D excitation k-space instead of 3D excitation k-space in the case of using the first-order gradients. The proposed method was validated using Bloch equation simulation.

Weiran Deng¹, Cungeng Yang¹, and V. Andrew Stenger^1
1Medicine, University of Hawaii John A. Burns School of Medicine, Honolulu, HI, United States

Multi-dimensioanl refocusing RF pulses for parallel transmission are useful for spin-echo imaging at high magnetic fields. The design of such pulses is challenging because it requires a nonlinear approach to "pancake-flip" the spins with different in-plane phases. We designed a refocusing pulse with 2D spiral k-space trajectory for eight transmitters using a two-process optimal control approach that flips the Y-component of spins from the +Y to -Y direction and leaves the X-component intact. The performance of the pulse was demonstrated using Bloch equation simulation.

Tomasz Dawid Lindel^1,2, Frank Seifert^1,2, Martin Dietterle^1,2, Thoralf Niendorf², and Bernd Ittermann^1,2
1Physikalisch-Technische Bundesanstalt, Braunschweig und Berlin, Germany, ²Berlin Ultrahigh Field Facility, Max-Delbrück-Centrum für Molekulare Medizin, Berlin, Germany

2D spatial excitation using Tx-SENSE is frequently combined with a conventional, slice selective refocusing pulse to restrict the signal in the third dimension. Severe B₁⁺ inhomogeneities of the refocusing pulse at 7T compromise the resulting image quality even if B₁ shimming is applied. We present phantom experiments with adapted Tx-SENSE pulses to address this problem.

Daehyun Yoon¹, Jeffrey A Fessler¹, Anna C Gilbert², and Douglas C Noll^3
1Electrical Engineering, University of Michigan, Ann Arbor, MI, United States, ²Mathematics, University of Michigan, Ann Arbor, MI, United States, ³Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States

We present a novel method for selecting and ordering efficient phase encoding(PE) locations in an Echo-Volumar trajectory for parallel excitation considering B0 field inhomogeneity. Recently several parallel excitation algorithms were presented enforcing sparsity on the number of chosen PE locations, but none of them considered off-resonance during excitation. We observed from simulation that this could lead to significant degradation in excitation accuracy where strong off-resonance is present. We developed a fast greedy algorithm to select and order efficient PE locations considering B0 field inhomogeneity, and demonstrate from simulation that our algorithm can significantly improve the excitation accuracy.

Peter Ullmann¹, Jeff Snyder², Martin Haas², Johannes Thomas Schneider^1,2, and Wolfgang Ruhm^1
1Bruker BioSpin MRI GmbH, Ettlingen, Germany, ²Dept. of Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany

Spatially-selective excitation (SSE) offers great potential for volume-selective MR-spectroscopy by allowing the excitation of arbitrarily shaped voxels which can be used to mitigate partial-volume effects and to increase SNR. In order to excite arbitrary voxels with high spatial resolution and large spectral bandwidth this study combines segmented SSE pulses with parallel RF transmission which enables significant reduction of the number of segments. Experiments on a 9.4T animal scanner demonstrate that by using segmented parallel excitation pulses complex-shaped voxels can be excited with good spatial selectivity and spectra with broad spectral range and high resolution can be acquired from these voxels.

Ling Xia¹, Tingting Shao¹, Minhua Zhu¹, Guofa Shou¹, Feng Liu², and Stuart Crozier^2
1Department of Biomedical Engineering, Zhejiang University, Hangzhou, China, People's Republic of, ²School of Information Technology & Electrical Engineering, University of Queensland, Brisbane, Australia

This work presents a novel approach to account for the limited coverage of RF energy in ultra high field MRI. Based on the recently developed parallel transmission technology, an optimized 3D tailored RF (TRF) pulse has been proposed to upgrade the RF excitation. The pulse is designed with an adaptive stack-spiral trajectory that is tailored according to the high-weight k-space area, which is most responsible for the desired excitation pattern. An iterative RF pulse design method is employed to ensure the excitation accuracy. Test simulations show that the proposed scheme optimally upgrades the excitation over the whole imaging area which includes deep domain where RF energy can be difficult to penetrate at ultra high field.

Sebastian Schmitter¹, Xiaoping Wu¹, Edward J Auerbach¹, Michael Hamm², Josef Pfeuffer³, Kamil Ugurbil¹, and Pierre-Francois Van de Moortele^1
1Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, United States, ²Siemens Healthcare, Charlestown, MA, United States, ³MR Application Development, Siemens Healthcare, Erlangen, Germany

One of the main challenges in ultra-high-field imaging is the spatial inhomogeneity of transmit B1. In cerebral time-of-flight (TOF) angiography this inhomogeneity results in spatially varying background suppression and thus suboptimal contrast. Parallel transmission (pTX) proved to generate homogeneous excitation by using independent RF waveforms on multiple transmit channels. So far, however, pTX methods have not been used at a large scale in clinical applications, and it remains to be demonstrated that this approach provides reliable results in clinical MR protocols. In this report we demonstrate the successful implementation of Spoke 3D RF pulses in multi-slab TOF angiography at 7T.

Manuel Walther¹, Kay Nehrke², Ingmar Grässlin², Ulrich Katscher², Markus Eblenkamp¹, Erich Wintermantel¹, and Peter Börnert^2
1Chair of Medical Engineering, Technische Universität München, Garching, Germany, ²Philips Research Laboratories, Hamburg, Germany

Parallel transmission has been used to overcome B1 homogeneity limitations and to improve the performance of multi-dimensional RF pulses by shortening their duration. One of the simplest multi-dimensional RF pulses is the pencil beam which finds application as a navigator to sense respiratory motion. As modern high-field systems (3T and beyond) are especially prone to off-resonance effects, shorter pulse durations are necessary to overcome this problem. However, accelerated pulses also cause aliasing artifacts. Transmit SENSE, replacing gradient encoding with transmit sensitivity encoding, has been found to significantly improve the performance of these accelerated navigator pulses, allowing pulse durations of 1.7ms.

Priti Balchandani¹, Mohammad Mehdi Khalighi², Scott Sigao Hsieh^1,3, Kawin Setsompop⁴, John Pauly³, and Daniel Spielman^1
1Radiology, Stanford University, Stanford, California, United States, ²Global Applied Science Laboratory, GE Healthcare, Menlo Park, California, United States, ³Electrical Engineering, Stanford University, Stanford, California, United States, ⁴A.A. Martinos Center for Biomedical Imaging, Harvard Medical School, Massachusetts General Hospital, Charlestown, Massachusetts, United States

Adiabatic B₁ shimming is a hybrid imaging approach which exploits the flexibility offered by multiple transmit channels to ensure the B₁-immunity of adiabatic RF pulses over the entire spatial region of interest. We developed a simulated annealing optimization algorithm to determine the amplitude and phase adjustments to adiabatic RF pulses played on individual transmit channels. The values were chosen to maximize uniformity of the flip angle while limiting RF amplitude and minimizing worst-case SAR. The final shimming algorithm was applied to 2-channel and 8-channel 7T B₁ transmit maps and the resultant values were tested in simulation. Highly uniform flip angles were achieved.

Adam B. Kerr¹, Duan Xu², Peder E.Z. Larson², Daniel B. Vigneron², and John M. Pauly^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States

Minimum-duration adiabatic spectral-spatial refocusing pulses are designed by using an alternate frequency sweep than the conventional hyperbolic secant pulse. A 55% reduction in pulse duration from 26 to 11.7 ms is demonstrated, at the cost of higher SAR and larger spectral transition bands. This approach will provide a significant decrease in the minimum TE for PRESS-localized MRSI sequences that use two of these pulses.

Adam B. Kerr¹, Duan Xu², Peder E.Z. Larson², Daniel B. Vigneron², and John M. Pauly^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States

A novel adiabatic full-passage pulse has been designed demonstrating a 33% peak power reduction compared to a classic hyperbolic secant (HS) pulse, while maintaining the spectral selectivity at one frequency edge. The new pulse is used to design the two adiabatic spectral-spatial pulses for a PRESS-localization, but with spectral profiles offset and reversed to achieve full HS selectivity. The peak power reduction will enable the higher spectral bandwidths demanded at high field without compromising selectivity.

Mayur Narsude^1,2, José Marques^1,2, Florent Eggenschwiler¹, and Rolf Gruetter^1,3
1Laboratory for Functional and Metabolic Imaging, Ecole Polytechnique Fédéral de Lausanne, Lausanne, Vaud, Switzerland, ²Department of Radiology, University of Lausanne, Lausanne, Vaud, Switzerland, ³Department of Radiology, University of Geneva, Geneva, Switzerland

Transmit B1 field inhomogeneous distribution encountered at high magnetic field (7T and above) makes it difficult to achieve homogeneous inversion in the whole brain region. In the presented study, few inversion RF pulses are studied for their inversion efficiencies using in vivo B1 mapping and in-vitro measurements. It is demonstrated that whole brain inversion is possible at 7T at the cost of somewhat higher SAR. The observed inversion efficiencies in the CSF regions were as predicted by the model with slight deviations, which are likely due to T1rho during the inversion, in the white matter regions.

William A Grissom¹, Graeme C McKinnon², and Mika W Vogel^1
1GE Global Research, Munich, Bavaria, Germany, ²GE Applied Science Lab, GE Healthcare, Milwaukee, Wisconsion, United States

The Shinnar-Le Roux (SLR) RF pulse design algorithm is currently the most widely-used method for designing one-dimensional RF pulses on constant gradient waveforms due to its ease of use and optimality. Presently, no analogous method exists for designing multidimensional RF pulses on non-Cartesian gradient trajectories. In this work the SLR pulse algorithm is generalized to nonuniform and multidimensional excitation pulses via a novel problem parameterization and design problem framework. The method is validated and compared to methods conventionally used for large-tip-angle spiral and spectral-spatial pulse design.

Jay Moore^1,2, Marcin Jankiewicz^1,3, Adam W Anderson^1,4, and John C Gore^1,4
1Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States, ²Physics and Astronomy, Vanderbilt University, Nashville, TN, United States, ³Department of Radiology and Radiological Sciences, Vanderbilt University, ⁴Department of Biomedical Engineering, Vanderbilt University

Non-selective composite RF pulses are numerically optimized for insensitivity to static and RF field variations and subsequently transformed into slice-selective pulses via RF shape modification and execution of an oscillating gradient waveform. Pulse designs accommodate the gradient performance and maximum RF amplitude limits of a commercial 7 T human scanner while significantly reducing signal variations due to B₁⁺ inhomogeneities in the human brain.

Martin A Janich^1,2, Rolf F Schulte², Markus Schwaiger³, and Steffen J Glaser^1
1Chemistry, Technische Universität München, Munich, Germany, ²GE Global Research, Munich, Germany, ³Nuclear Medicine, Technische Universität München, Munich, Germany

Broadband RF pulses are of great interest for improving selectivity and reducing chemical shift displacements in localized spectroscopy. Slice-selective broadband refocusing pulses with immunity to B₁ variations are designed using optimal control theory. Pulses are optimized under energy constraints, are compared to a broadband SLR pulse, and applied in a PRESS experiment.

Marcin Jankiewicz^1,2, Jay Moore^1,3, Adam Anderson^1,4, and John Gore^1,4
1Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States, ²Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, United States, ³Physics and Astronomy, Vanderbilt University, United States, ⁴Biomedical Engineering, Vanderbilt University

We evaluate a sample of short, low-SAR, non-selective refocusing pulses using a 3D SE-EPI sequence at 7 T. With only a two-fold increase in SAR relative to a block pulse, several refocusing options were found to increase signal in low B₁⁺ regions by a factor of 2.

Galen D Reed¹, Adam B Kerr², Peder E.Z Larson³, Eugene Ozhinsky³, John Kurhanewicz³, and Daniel B Vigneron^3
1Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, California, United States, ²Electrical Engineering, Stanford University, Palo Alto, California, ³Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, California

Very Selective Saturation (VSS) pulses are critical for PRESS localized spectroscopy by providing outer volume suppression and reducing chemical shift artifacts. Since these bands are placed in parallel pairs, cosine modulation can be applied to excite two bands in a single pulse, thereby saving time in the suppression pulse train by eliminating redundant crusher gradients. This abstract demonstrates the design of an increased bandwidth dual band selective saturation pulse based on a convex optimization filter design (for optimizing selectivity) and biased-probability Monte Carlo root flipping algorithm (for minimizing computation time). These pulses were designed and tested in patient exams incorporating prostate MRSI

Christine Law¹, and Sonal Josan^2
1University of Oxford, Oxford, Oxfordshire, United Kingdom, ²SRI International

We demonstrate a convex optimization technique to design minimum peak RF saturation pulses. This method achieves global minimum peak RF amplitude without the need of exhaustive search over possible phase profiles. Our technique accept passband/stopband ripple and time-bandwidth product as inputs. But to formulate as a convex problem, we make use of autocorrelation function of the rf waveform and employ the technique of convex iteration. We compare our technique with exhaustive search in small time-bandwidth case and show a result for large time-bandwidth design.

Xiaocheng Wei¹, and Yongchuan Lai^1
1MR, GE Healthcare, Beijing, Beijing, China, People's Republic of

In this study, a new Spatial-Spectral (SpSp) pulse, called as Dynamic Gradient SpSp (DG-SpSp) pulse, is designed to achieve better minimum slice thickness performance under certain peripheral nerve stimulation (PNS) limitations. And the payment of this improvement is tiny increment of TR time.

Gregory R Lee¹, Jean A Tkach^1,2, and Mark A Griswold^1,3
1Radiology, Case Western Reserve University, Cleveland, OH, United States, ²Radiology, Imaging Research Center, Cincinnati Children's Hospital Research Foundation, Cincinnati, OH, United States, ³Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States

At 3T and above, spectral-spatial (SPSP) subpulse durations of less than 1 ms are needed in order to selectively excite either water or fat spins. With such short subpulses, RF power limits prohibit the excitation of sharp spatial slabs when using traditional SPSP pulse designs. In the present work, a new algorithm for alternating between minimum-time VERSE gradient waveform reshaping and RF pulse redesign is presented. The algorithm can produce much sharper spatial slabs while preserving the spectral profile. In addition, there is flexibility in the specification of the SPSP excitation pattern and tip-down time within the pulse.

Michael Carl¹, Jing-Tzyh Alan Chiang², and Mark Bydder^2
1Global Applied Science Laboratory, GE Healthcare, San Diego, CA, United States, ²University of California, San Diego, United States

We have derived an analytic expression for the steady state transverse magnetization resulting from VERSE corrected 2D UTE excitation RF pulses, which we used to predict the optimum flip angles (generalized Ernst angles) for short T2 tissues. Simulations and experimental verifications support the validity of the derived results.

Yushan Chen¹, Congbo Cai¹, Fenglian Gao¹, Shuhui Cai¹, and Zhong Chen^1
1Communication Engineering and Physics, Fujian Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, Fujian, China, People's Republic of

Intermolecular multiple quantum coherences (iMQCs) caused by long-range dipolar interactions possess some appealing unique properties for high-resoultion NMR spectroscopy in inhomogeneous fields. In this abstract, a new iMQC method involving Hadamard encoding is proposed to obtain high- resolution 1D NMR spectrum in inhomogeneous fields within a short acquisition time. Its application on simple and complex systems was tested. Experimental results indicate that the spectral line-width can be considerably reduced and the signal to noise ratio can be improved.