Robert Wayne Stobbe¹, and Christian Beaulieu^1
1Biomedical Engineering, University of Alberta, Edmonton, Alberta, Canada

A looping trajectory, as one may wind a ball of yarn, is introduced for single-shot 3D image acquisition. When acceleration is constrained for a constant rate of gradient change, a hardware limiting constraint, this rate of change is similar to that of a ‘back-and-forth’ approach sampling the same k-space volume in the same time. Sampling density and efficiency are addressed and preliminary phantom images created at 4.7 Tesla. Theoretical advantages of the yarn-ball trajectory include improved off-resonant performance, given early sampling of central k-space, and the potential for designed ‘random’ undersampling to increase the volume of k-space sampled.

James Grant Pipe¹, and Nicholas Ryan Zwart^1
1Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ, United States

FLORET is a 3D k-space trajectory previously described, which samples all of k-space twice, in orthogonal directions. This work shows the performance of variable density FLORET, which is very efficient and has very incoherent aliasing, which is desirable for many types of reconstruction algorithms.

Payal Sharad Bhavsar¹, and Jim Pipe^1
1Neuroimaging, Barrow Neurological Institute, Phoenix, AZ, United States

A new method is proposed to estimate time varying gradient delays in spiral MRI. Gradient delays cause blurring and distortions in spiral images. Several methods have been proposed to estimate these delays. Continuous, independent delays are estimated for all three gradient channels as a function of the ADC time. This method includes gradient coupling effects and requires minimal modification of the pulse sequence design.A 3D spiral trajectory FLORET was used to estimate the gradient delays. To validate the results, a constant delay correction method was used for analysis. In-vivo and phantom experiments were reconstructed using constant and variable gradient delay correction. For in-vivo data, variable correction method showed significant reduction in the artifacts. Comparable results were obtained for constant and variable correction method for phantom simulations. The method can be applied to any stack of spirals based trajectory.

Walter RT Witschey¹, Christian A. Cocosco¹, Daniel Gallichan¹, Gerrit Schultz¹, Hans Weber¹, Anna Masako Welz¹, Jürgen Hennig¹, and Maxim Zaitsev^1
1Medical Physics, University Medical Center Freiburg, Freiburg i. Breisgau, Germany

A technique is described for localization of MR signals from a target volume using nonlinear pulsed magnetic fields and spatial encoding trajectories designed using local k-space theory. In vivo human brain images were obtained using a custom quadrupolar gradient coil integrated with a whole-body 3 T MRI scanner to demonstrate localization in 3D T2*-weighted imaging.

Hans Weber¹, Daniel Gallichan¹, Gerrit Schultz¹, Walter R Witschey¹, Anna Masako Welz¹, Christian A Cocosco¹, Jürgen Hennig¹, and Maxim Zaitsev^1
1Department of Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany

Adaptation of the slice shape to the volume of interest would be beneficial for many MRI applications, as it allows coverage with fewer excited slices. In this work, we theoretically design and demonstrate experimentally a method (ExLoc) for excitation of customized curved slices, based on a combination of linear and nonlinear gradients. As signal is encoded along the curved surface, a local rectangular voxel shape is maintained. In contrast to linear spatial encoding, partial volume effects and through-slice dephasing is avoided.

Noam Ben-Eliezer¹, and Lucio Frydman^1
1Chemical-Physics, Weizmann Institute of Science, Rehovot, Israel

The generation of 2D NMR images in a single-scan is a common ingredient in a variety of applications. Real life applications, however, often require 3D volumetric coverage, while keeping the overall scanning on a sub-second timescale. Current ultrafast 3D methods are based on EPI, which notwithstanding its proven success is still challenged by B0 heterogeneities and/or multiple chemical sites. We explore here a new approach, based on combining spatiotemporal-encoding with super-resolution reconstruction algorithms, to achieve higher immunity to these artifacts. A number of 3D sequence schemes are presented and demonstrated on in-vivo mouse models.

Gilles Puy^1,2, José Marques^2,3, Rolf Gruetter^2,3, Jean-Philippe Thiran¹, Dimitri Van De Ville^4,5, Pierre Vandergheynst¹, and Yves Wiaux^1,5
1Institute of Electrical Engineering, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ²Institute of the Physics of Biological Systems, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ³Department of Radiology, University of Lausanne (UNIL), Lausanne, Switzerland, ⁴Institute of Bioengineering, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁵Department of Radiology and Medical Informatics, University of Geneva (UniGE), Geneva, Switzerland

We advocate a spread spectrum technique for accelerated MRI. The method, justified by the compressed sensing theory, resides in pre modulating the signal by a linear chirp before uniform random k-space undersampling. Numerical simulations confirm the theoretical prediction according to which the pre-modulation is essential in enhancing the signal reconstruction quality. The chirp modulation was implemented on a 7T scanner with the use of a second order shim coil. Results from 3D single coil phantom and in vivo data acquisitions already exhibit slightly better reconstruction qualities than state of the art variable density sampling methods.

Jonathan C Sharp¹, and Scott B King^2
1Institute for Biodiagnostics (West), National Research Council of Canada, Calgary, AB, Canada, ²Institute for Biodiagnostics, National Research Council of Canada, Winnipeg, MB, Canada

Standard MRI methods achieve a scanning of k-space by 2 methods: B0-gradients (providing continuous motion) and refocusing (point reflection about the k-space origin). Here we analyze a third option: point reflection about other k-space locations. The method is achieved by refocusing with B1-phase-gradients and is the basis of the TRASE (Transmit Array Spatial Encoding) imaging method. These novel k-space operations offer the pulse sequence designer an expanded toolkit. Our aim here is to formalize the rules governing the construction of k-space trajectories for this type of sequence.

Hirad Karimi¹, and Charles H Cunningham^1,2
1Medical Biophysics, University of Toronto, Toronto, ON, Canada, ²Imaging Research, Sunnybrook Health Sciences Centre, Toronto, ON, Canada

In conventional MRI systems, the gradient fields that encode the spatial information are stationary relative to any physiologic movements. In this abstract we propose a novel MR endoscope system based on spatial encoding using the field perturbations around susceptibility markers instead of conventional gradients.

Alexey A. Samsonov¹, Julia V. Velikina², and Walter F. Block^2
1Radiology, University of Wisconsin, Madison, WI, United States, ²Medical Physics, University of Wisconsin, Madison, WI, United States

Sliding window (SW) gridding reconstruction has been an efficient way to reconstruct undersampled interleaved radial MRI data. Temporal SW filters reduce streaking artifacts, but also reduce temporal resolution at higher spatial frequencies. Parallel MRI (pMRI) may mitigate the artifacts and improve temporal footprint. However, the actual undersampling factors in existing radial MRI applications may result in significant noise/resolution loss. The combination of pMRI and SW may result in improved image quality than with each individual method alone. In this research, we propose an optimized method to combine k-space based parallel MRI with SW reconstruction for radial trajectories.

Yanqiu Feng¹, Yanli Song¹, Cong Wang¹, Taigang He², Xuegang Xin¹, and Wufan Chen^1
1School of Biomedical Engineering, Southern Medical University, Guangzhou, China, People's Republic of, ²Royal Brompton Hospital and Imperial College, London, United Kingdom

Direct Fourier transform (DFT) could reconstruct MR image from non-Cartesian data with high precision. However, the high computation complexity makes DFT impractical for clinical application. Up to now, the published "FFT" algorithms for non-equispaced data do not strictly compute the DFT of nonequispaced data, but rather some approximation. In this work, an efficient algorithm for DFT using the Chirp Transform Algorithm (CTA) was proposed to reconstruct MR image from non-Cartesian trajectories consisting of lines with equispaced points such as radial or PROPERLLER sampling. The proposed CTA-DFT algorithm is demonstrated to be significantly faster than DFT while keeping the same accuracy.

Johannes Tran-Gia¹, Daniel Stäb¹, Christian Oliver Ritter¹, Dietbert Hahn¹, and Herbert Köstler^1
1Institute of Radiology, University of Würzburg, Würzburg, Bavaria, Germany

A model-based image reconstruction algorithm is presented using radially sampled data to reconstruct one fully sampled image for each acquired radial projection after magnetization preparation. For radial trajectories, every acquired projection contains information about the image contrast. By incorporating a signal model into the image reconstruction, it is possible to use this information to resolve the signal evolution with a high temporal resolution. The functionality of the algorithm is demonstrated in phantom studies as well as in in-vivo measurements on a healthy volunteer.

Simon Konstandin¹, Armin Michael Nagel², Patrick Michael Heiler¹, and Lothar Rudi Schad^1
1Computer Assisted Clinical Medicine, Heidelberg University, Mannheim, Germany, ²Medical Physics in Radiology, German Cancer Research Center, Heidelberg, Germany

Large voxel sizes (>30mm³) are required for sodium imaging due to its low sensitivity that can result in Gibbs ringing if no filter is used. In this work, a Hamming filter was implemented into a 2D radial trajectory and compared to imaging with a post-filter and without filtering. The absence of Gibbs ringing and an SNR increase of 60% for sodium heart imaging are clearly visible for the filtered images. In conclusion, the use of an implemented filter in the trajectory will improve the SNR and the shape of the PSF.

David Manuel Grodzki^1,2, Peter M Jakob¹, and Bjoern Heismann^2
1Department of Physics EP5, University of Wuerzburg, Wuerzburg, Bavaria, Germany, ²Magnetic Resonance, Siemens AG, Erlangen, Bavaria, Germany

Sequences with ultra short echo time (TE) enable new applications of MRI, including bone, tendon, ligament and dent imaging. In this work a sequence is presented that achieves the shortest possible TE given by TX/RX switching time and gradient performance of the MR Scanner. In PETRA (Pointwise Encoding Time reduction with Radial Acquisition) outer k-space is filled with radial half-projections whereas the centre is measured single pointwise on a Cartesian trajectory. This hybrid sequence combines the features of single point imaging with radial projection imaging. No hardware changes are required.

Kai Tobias Block¹, and Martin Uecker^2
1MR Application and Workflow Development, Healthcare Sector, Siemens AG, Erlangen, Germany, ²Biomedizinische NMR Forschungs GmbH, Göttingen, Germany

This work describes a novel approach for the correction of system-dependent gradient delays, which pose a major problem for radial sampling in routine applications. The method performs a cross-correlation analysis of anti-parallel calibration spokes to assess the angle-dependent k-space shift introduced by the delay. The delay is compensated during the gridding procedure by realigning the data according to the estimated shift distance. The method has been implemented for a radial 2D and 3D sequence. It has been evaluated on a large number of commercial MR systems and proved to deliver consistent image quality at a wide range of acquisition bandwidths.

MinOh Ghim¹, Sang-Young Zho¹, Eunhae Joe¹, and Dong-Hyun Kim^1
1Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of

Conventional 3D MR imaging typically uses 3D Fourier encoding acquisitions and/or 3D Fourier transform based reconstruction. However, for high resolution 3D imaging using these methods, SNR can be a limiting factor along with side effects such as ringing with Fourier based reconstruction. Here, we propose an alternative 3D high resolution imaging method where multiple 2D acquisitions are performed using a variable oblique-view pulse scheme. An image-based reconstruction is performed using an iterative back projection process thereby enabling flexible tradeoffs between SNR and resolution.

Saiprasad Ravishankar¹, and Yoram Bresler^1
1Department of Electrical and Computer Engineering and the Coordinated Science Laboratory, University of Illinois, Urbana, IL, United States

Compressed Sensing (CS) MRI with non-adaptive sparsifying transforms such as wavelets and finite differences can perform poorly at high undersampling factors. In this work, we introduce an adaptive framework for MR image reconstruction employing multiscale sparse representations. The multiscale patch-based sparsifying transform (dictionary) is learnt directly using the undersampled k-space data and is thus adapted to the current image. An alternating reconstruction algorithm learns the sparsifying dictionary at different scales, and uses it to remove aliasing and noise in one step, and subsequently restores and fills-in the sampled k-space data in the other step. Experimental results demonstrate the superior performance of such an image reconstruction formulation that exploits image patch sparsity at several scales. The multiscale framework provides highly accurate reconstructions at high undersampling factors. We also show the reconstructions with previous CSMRI methods employing nonadaptive dictionaries and demonstrate significant improvements with our approach over the former.

Bo Kou¹, Guoxi xie², Bensheng Qiu², and Xin Liu^2
1Shenzhen Institutes of Advanced Technology, Chinese Academy of Science, Shenzhen, China, People's Republic of, ²Shenzhen Institutes of Advanced Technology, Chinese Academy of Science, China, People's Republic of

The main idea of Compressed Sensing is to exploit the fact that there is some structure and redundancy in most signals of interest. Clearly, the more we known about the signal and the more the information we encode into the signal processing algorithm, the better performance we can achieve. In this paper, we propose an adaptive compressed MRI sensing scheme that combined wavelet encoding with compressed sensing which originated from the optic image data acquisition . Our approach exploits not only the fact that most of the wavelet coefficients of MR images are small but also the fact that values and locations of the large coefficients have a particular structure. Exploiting the structure of the wavelet coefficients of MR images is achieved by replacing the pseudo-random measurements with a direct and fast method of adaptive wavelet coefficient acquisition.

Changwei Hu¹, Xiaobo Qu², Di Guo², Lijun Bao¹, Shuhui Cai¹, and Zhong Chen^1
1Department of Physics, Xiamen University, Xiamen, Fujian, China, People's Republic of, ²Department of Communication Engineering, Xiamen University, Xiamen, Fujian, China, People's Republic of

Undersampling k-space is an effective way to reduce acquisition time for MRI. However, aliasing artifacts introduced by undersampling may blur the edges of MR images, which often contain important information for clinical diagnosis. In this work, we propose an edge-weighted model by pluging two weighting matrix into the objective function of constrained l1 norm minimization problem. Reconstructions with more precise edge recovery are then obtained by the proposed EWIT algorithm.

Sairam Geethanath¹, Steen Moeller², Curtis A Corum², Matthew A Lewis^1,3, and Vikram D Kodibagkar^1,3
1Joint graduate program in biomedical engineering, UT Arlington and UT Southwestern Medical Center, Dallas, Texas, United States, ²Center for Magnetic Resonance Research, University of Minnesota, ³Radiology, UT Southwestern Medical Center

Sweep imaging with Fourier transformation (SWIFT) is a novel MRI technique which facilitates imaging of short T2 nuclei. Currently, exquisite 3D images can be acquired within minutes but imaging dynamic changes in short T2 species, such as T2 exchange contrast agents, would require an even faster imaging scheme. The application of compressed sensing to accelerate SWIFT MR imaging has been demonstrated on a phantom for an acceleration factor ~5. Compressed sensing based reconstruction of the MR volume shows high fidelity with reduced artifacts and lower noise. Future work involves the reconstruction of in vivo data and applications to dynamic imaging.

Jun Miao¹, Feng Huang², and David L Wilson^3,4
1Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States, ²Invivo Corporation, Gainesville, Florida, United States, ³Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, ⁴Radiology, University Hospitals of Cleveland, Cleveland, Ohio, United States

Case-PDM is applied to investigate TV weight regularization parameter of Compressed Sensing (CS). Experimental results show that optimal TV weight varies across different imaged subject/ pulse sequence/ scanner, and may change slightly for the same data set with different sampling ratio. To achieve a high image quality CS reconstruction, a pre-calibration is necessary to find the optimal TV weight.

Satoshi Ito¹, Koji Miyabayashi², and Yoshifumi Yamada^2
1Research Division of Intelligence and Information Sciencs, Utsunomiya University, Utsunomiya, Tochigi, Japan, ²Utsunomiya University, Utsunomiya, Tochigi, Japan

Compressed sensing (CS) aims to reconstruct signals and images from significantly fewer measurements than were traditionally thought necessary. MRI is a medical imaging tool burdened by an inherently slow data acquisition process. The application of CS to MRI has the potential for significant scan time reductions. In this paper we present a new CS method based on the FREBAS transform which we have proposed as a new kind of multi-resolution image analysis. The algorithm and the performances of proposed method were demonstrated and it was shown that proposed CS method can achieve a reduction factor higher than the standard CS method.

Joonsung Choi¹, Yeji Han¹, Jinyoung Hwang¹, Jun-Young Chung², Zang-Hee Cho², and HyunWook Park^1
1Department of Electrical Engineering, KAIST, Daejeon, Korea, Republic of, ²Neuroscience Research Institute, Gachon University of Medicine and Science

Compressed sensing (CS) is a newly emerging technique to reconstruct undersampled signal. If certain conditions, such as sparsity and incoherent sampling, are satisfied, the signal can be accurately reconstructed from undersampled data by the CS technique. Therefore, a sparsifying transform is required in the CS technique for the target signal. Conventionally, the CS uses a single sparsifying transform. However, the target signal can be more sparsely represented in some cases by adopting multiple sparsifying transforms. In this work, an adaptive CS algorithm using two transforms is proposed to reconstruct MR angiography images with high accuracy and quality.

Merry P Mani¹, Tong Zhu², Jianhui Zhong², and Mathews Jacob^3
1Rochester Center for Brain Imaging, Electrical and Computer Engineering, University of Rochester, Rochester, NY, United States, ²Imaging Sciences, University of Rochester, Rochester, NY, United States, ³Biomedical Engineering, University of Rochester, Rochester, NY, United States

The aim of the study is to test the feasibility of using compressed sensing to accelerate the high angular sampling schemes for DTI. A realistic but simulated k-space data was randomly under-sampled in the frequency-diffusion domain with 256 diffusion directions. By making use of the sparsity of the orientation information in each voxel and choosing an appropriate set of basis functions, the diffusion-weighted images were reconstructed using compressed sensing without aliasing. Up to 8 fold acceleration could be achieved within reasonable reconstruction errors. The technique allows to fit parametric models to high angular resolution DW data due to the chosen set of basis functions.

Xiaobo Qu¹, Changwei Hu², Di Guo¹, Lijun Bao², and Zhong Chen^2
1Department of Communication Engineering, Xiamen University, Xiamen, Fujian, China, People's Republic of, ²Department of Physics, Xiamen University, Xiamen, Fujian, China, People's Republic of

Undersampling the measurement can reduce the acquisition time in magnetic resonance imaging (MRI) at the cost of introducing the aliasing artifacts. The sparsity of magnetic resonance (MR) images in wavelet transforms shows promising results to suppress these artifacts [1]. Recent developments demonstrate the Gaussian Scale Mixture (GSM) for modeling dependency of wavelet coefficients for single image can incorporate more prior information and improve the traditional wavelet-based reconstruction. In this paper, we consider the cases that MR study is comprised by many different types of images of the same patient (e.g. T1, T2, Proton density-PD, etc). By modeling the dependency of wavelet coefficients of these multi-component images as multicomponent GSM (mGSM), we propose an iterative algorithm to jointly reconstruct these MR images from undersampled k-space measurements. Simulation demonstrates that this model can improve the reconstructed MR images than the wavelet-based hard iterative thresholding separately does for each image.

Joshua Trzasko¹, and Armando Manduca^1
1Mayo Clinic, Rochester, MN, United States

In this work we investigate a generalization of sparsity-driven undersampled image reconstruction strategies for image series which do not have a readily identifiable temporal or parametric sparse transformation but for which strong yet unknown spatiotemporal correlations are anticipated.

Yue Li¹, Manisha Aggarwal¹, Jiangyang Zhang², and Susumu Mori^2
1Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²Radiology, Johns Hopkins University School of Medicine, Baltimore, MD, United States

Compressed sensing method and its variations have been developed for reconstruction of MRI using undersampled k space data. Here we propose a constraint dedicated for diffusion tensor imaging (DTI) as well as a phase constraint that has been used in partial Fourier reconstruction to improve compressed sensing reconstruction accuracy. The testing result on high field mouse embryo DTI shows improvement of reconstruction accuracy in both diffusion weighted (DW) and tensor images.

Feng Zhao¹, Jeffrey A. Fessler², Jon-Fredrik Nielsen¹, and Douglas C. Noll^1
1Biomedical Engineering, University of Michigan, Ann Arbor, Michigan, United States, ²Electrical Engineering, University of Michigan

It is an image reconstruction method which combines Compressed Sensing with separate magnitude and phase regularization. It is proposed to speed up the acquisition and improve performance in reconstructing images with rapid phase map variation.

Zhipeng Cao¹, Christopher T. Sica², Philipp Ehses³, Sukhoon Oh², Yeun C. Ryu², Christopher M. Collins^1,2, and Mark A. Griswold^4
1Bioengineering, The Pennsylvania State University, Hershey, PA, United States, ²Radiology, The Pennsylvania State University, Hershey, PA, United States, ³Research Center for Magnetic Resonance Bavaria (MRB), Würzburg, Germany, ⁴Radiology, Case Western Reserve University, Cleveland, OH, United States

A phase-constrained compressed sensing reconstruction method is proposed. It is validated through simulation and experiments to be effective on smooth or focal, mild or dramatic temperature changes for rapid MRI temperature mapping both on phantom and in vivo based on proton resonance frequency shift. The method can also be combined with partially parallel acquisition techniques for additional acceleration.

Fan Lam^1,2, Raman Subramanian³, Dan Xu³, and Kevin F. King^3
1Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ²Beckman Institute, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ³GE Healthcare, Waukesha, WI, United States

We present a novel reconstruction scheme incorporating not only the prior information that the MR image is sparse in certain transformation domain but also the support information for the target image to be reconstructed. Support can be detected either from low resolution estimate or from certain transformation domain of a high resolution reference image. A mix weighted L1-L2 regularization formulation is established for reconstruction. Data from a noncontrast MRA and a brain imaging experiment are used to demonstrate the advantageous performance of the proposed method compared to conventional compressed sensing based reconstruction from sparsely sampled data.

Hisamoto Moriguchi^1,2, Shin-ichi Urayama³, Yutaka Imai¹, Manabu Honda⁴, and Takashi Hanakawa^4,5
1Radiology, Tokai University, Isehara, Kanagawa, Japan, ²Radiology, Hiratsuka municipal hospital, Hiratsuka, Kanagawa, Japan, ³Human Brain Research Center, Kyoto University, Kyoto, Kyoto, Japan, ⁴Functional Brain Research, National Center of Neurology and Psychiatry, Kodaira, Tokyo, Japan, ⁵Precursory Research for Embryonic Science and Technology, Japan Science and Technology Agency, Japan

Partial Fourier (PF) imaging is one of the most widely used fast data acquisition methods in MRI. In existing PF reconstruction algorithms, image quality depends on estimated phase information. However, it is often difficult to estimate accurate phase. In this study, a novel PF reconstruction technique that does not require phase estimation has been demonstrated. Since the recently proposed focal underdetermined system solver (FOCUSS) has been taken advantage of, the new method is referred to as �ePF-FOCUSS�f. With PF-FOCUSS, over 50% reduction in scan time can be achieved. Furthermore, images reconstructed using PF-FOCUSS are generally of quite high quality.

David S Smith¹, and Edward Brian Welch^1
1Radiology and Radiological Sciences and Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States

We have designed a phantom that, unlike the Shepp-Logan phantom, is difficult to reconstruct with compressed sensing MRI techniques. This phantom is not sparse under a gradient transformation and contains many features susceptible to corruption under a partial Fourier measurement operator. And unlike real images, the phantom has no noise, so it may be used for exact measurements of noise-free reconstruction accuracy.

Hua Wang¹, Adnan Trakic¹, Bing Keong Li¹, Yeyang Yu¹, Feng Liu¹, and Stuart Crozier^1
1The University of Queensland, Brisbane, QLD, Australia

This abstract presents the application of compressed sensing to rotating RF coil concept to exploit the data acquisition redundancy and better imaging quality, based on the fact that compressible signals can be reconstructed from randomly under-sampled frequency information, thus, the imaging acceleration can be achieved without sacrificing much image quality. The proposed method applies the recently developed Time-Division-Multiplexed Sensitivity Encoding scheme with compressed sensing framework to improve and accelerate the image reconstruction using a physically rotating RF coil. RF excitation pulse is randomized to generate a measurement matrix which is incoherent to the sparsity basis. The reconstruction performance is evaluated and shows improved reconstruction performance as compared with conventional case.

Daniel Holland¹, Careesa Liu², Chris V. Bowen², Andy Sederman¹, Lynn Gladden¹, and Steven D. Beyea^2
1Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge, United Kingdom, ²Institute for Biodiagnostics (Atlantic), National Research Council Canada, Halifax, Nova Scotia, Canada

We demonstrate that the use of Compressed Sensing reconstruction of variable density spiral fMRI data minimizes the aliasing artifact inherent to sparse k-space acquisitions that can result in additional signal fluctuations in the time course data. Our data demonstrate an improvement in the apparent fMRI sensitivity relative to the same images reconstructed without CS. CS reconstruction of a 50% undersampled VD spiral acquisition resulted in a 40% increase in whole brain activation volume relative to the same data reconstructed with conventional (non-CS) techniques, demonstrating that fMRI data obtained using VD trajectories should be reconstructed using Compressed Sensing.

Bertram Jakob Wilm¹, Christoph Barmet¹, Matthieu Guerquin-Kern^1,2, Max Haeberlin³, and Klaas Paul Pruessmann^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Zurich, Switzerland, ²Biomedical Imaging Group, EPFL Lausanne, Lausanne, Vaude, Switzerland, ³Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

We show the application of Partial Fourier to spiral imaging. The method is demonstrated for phantom and in-vivo single-shot acquisition allowing for an increase in k-space undersampling. The images with an in-plane resolution of 1.4 mm do not show aliasing related artifacts and are virtually free from B0 off-resonance effects.

Henry Szu-Meng Chen¹, Angshul Majumdar², Rabab Kreidieh Ward², and Piotr Kozlowski^1,3
1UBC MRI Research Centre, Vancouver, BC, Canada, ²Electrical and Computer Engineering, University of British Columbia, Vancouver, BC, Canada, ³ICORD, Vancouver, BC, Canada

Myelin water imaging based on multi-echo CPMG sequence is inherently slow, especially for high resolution in vivo animal studies. We hypothesize that using compressed sensing with group-sparse reconstruction will significantly increase the acquisition efficiency of myelin water images. CS accelerated myelin water fraction (MWF) maps were simulated using fully sampled k-space data acquired from excised rat spinal cords at 7T. MWF map quality was assessed and found to be minimally impacted up to an acceleration factor of 2, making the technique a promising approach at increasing acquisition efficiency in myelin water mapping.

Yi-Yu Shih¹, Jia-Shuo Hsu², Yi-Ru Lin³, Shang-Yueh Tsai⁴, and Hsiao-Wen Chung^1,2
1Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan, ²Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan,³Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, ⁴Department of Electrical Engineering, Chang Gung University, Taoyuan, Taiwan

k-t PCA was used to reconstruct the dynamic 3D lung images under various accelerating conditions. The effects of the number of principal components (PC) were investigated, where the number of PC required was found to larger than in cardiac imaging due to the higher complexity of the branching pulmonary vasculature. The time-intensity curves showed very good consistency with the full-sampled data-set, and the overshoot resulting from temporal discontinuity at the beginning and the end of the curves was mildened, suggesting feasibility of accelerated 3D lung perfusion imaging with better slice coverage and improved temporal resolution.

Maria Magnusson^1,2, Olof Dahlqvist Leinhard^2,3, and Peter Lundberg^2,3
1Dept. of Electrical Engineering, Linköping University, Linköping, Sweden, ²Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, Sweden, ³Radiation Physics, Linköping University, Linköping, Sweden

The novel method PRESTO-CAN for 3D-plus-time resolved MRI includes a radial-Cartesian hybrid sampling. Golden ratio angular sampling and hourglass filtering provides high temporal resolution. When the MRI-data is used for fMRI, the echo times are long, TE ≈ 37-40 ms, and result in field inhomogeneities and phase variations in the reconstructed images. Therefore, PRESTO-CAN also includes its own calibration and correction procedure. Reconstruction is performed using gridding. The image quality is almost identical to what can be obtained with conventional Cartesian sampling. However, it will be improved further by using a recently proposed optimal pre-weighting function during gridding.

Peng Cao^1,2, and Ed X. Wu^1,2
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong SAR, China, People's Republic of, ²Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong SAR, China, People's Republic of

This study aims to employ compressed sensing (CS) in phase-encoded 3D magnetic resonance imaging. In this study a CS recover algorithm is developed and implemented to 2 fold accelerate 3D imaging while maintaining image quality. This work reports the first demonstration of CS to phase-encoded, 3D, MR image.

Ramin Eslami¹, and Mathews Jacob^2
1Biomedical Engineering, University of Rochester, Rochester, NY, United States, ²University of Rochester

Recently, we proposed an efficient MRSI sparse reconstruction technique where we modeled the system using priors such as inhomogeneity, and brain and lipid masks estimated from a companion MR scan for the EPSI sequence using a single channel coil. Here, we propose a fast parallel MRSI acquisition scheme designed on a spiral trajectory. Using a 12-channel head coil, we acquire the in vivo MRSI data at spatial resolution of 44×44 with a single average. We extend our sparse reconstruction scheme to parallel MRSI data on the spiral trajectory. This way, we efficiently reduce measurement noise and other artifacts such as field inhomogeneity and spectral leakage in our proposed reconstruction while we have a fast MRSI acquisition (~1"min" for a slice). We show that the proposed scheme could recover the spectral data and outperforms Tikhonov-regularized SENSE reconstruction. We also demonstrate a two-fold acceleration of the acquisition that leads to a comparable reconstruction.

Dany Merhej^1,2, Helene Ratiney¹, Chaouki Diab³, Mohamad Khalil², and Rémy Prost^1
1CREATIS, CNRS UMR 5220, Inserm U1044, INSA-Lyon, Université Lyon 1, Université de Lyon, Lyon, France, ²EDST, Azm research center, Lebanese University, Tripoli, Lebanon, ³ISAE – Cnam Liban, Beirut, Lebanon

A major drawback in application of magnetic resonance spectroscopic imaging is the long acquisition time required to gather necessary data to achieve satisfactory resolution. When the chemical shift spectrum is inherently sparse, i.e. with a limited support, it is possible to reconstruct this spectrum from a subset of the k-space samples, thus reducing the number of phase encoding steps and subsequently reducing acquisition time. In this case, our approach outperforms compressive sensing using L1-minimization. The proposed reconstruction technique is validated on simulated, in vitro and in vivo data.

Behzad Sharif¹, Debiao Li^1,2, and Shawn Wagner^1
1Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, ²Northwestern University, Chicago, IL, United States

In hyperpolarization imaging, the magnetization is limited from the start of an imaging sequence and should be utilized while relaxation to thermal equilibrium occurs. Compressed sensing has been used to overcome the rapid loss of magnetization from molecules with short T1. With the introduction of molecules with longer T1, there is potential for high resolution imaging with compressed sensing used in all aspects of data acquisition. We show that high-quality images can be reconstructed from 2-fold undersampled data by exploiting the image-domain sparsity in C-13 imaging. The resulting temporal acceleration will enable higher resolution full-body imaging of hyperpolarized biomarkers for cancer detection.

Peter J Shin¹, Michael A Ohliger², Simon Hu², Peder E. Z. Larson², Cornelius Von Morze², Michael Lustig³, and Daniel B Vigneron^2
1Joint Graduate Group in Bioengineering, University of California at San Francisco & Berkeley, San Francisco, California, United States, ²Department of Radiology and Biomedical Imaging, University of California at San Francisco, San Francisco, California, United States, ³Department of Electrical Engineering and Computer Science, University of California at Berkeley, Berkeley, California, United States

Hyperpolarized ¹³C MRSI is a powerful tool for studying metabolic processes in vivo, enabling monitoring of ¹³C substrates and downstream metabolic products. However, spatial coverage and resolution of such metabolic images are fundamentally limited by the rapid metabolism and T₁ relaxation, necessitating the development of fast data acquisition schemes. In this work, we investigated accelerating hyperpolarized ¹³C spectroscopic imaging with L₁-SPIRiT compressed sensing autocalibration parallel imaging and showed that application of SPIRiT on simulated hyperpolarized ¹³C parallel imaging provided excellent noise performance and reduced artifacts in highly accelerated imaging schemes.