Parallel Imaging: Methods & Applications
Exhibit Hall 2-3                    Thursday 13:30-15:30

                  2710.     GRAPPA-Accelerated Short Axis BLADE EPI for Multi-Shot Diffusion Weighted Imaging

Alto Stemmer1, Berthold Kiefer1

1Siemens AG, Healthcare Sector, Erlangen, Germany

In this work the traverse speed along the phase encoding direction of DW EPI is maximized by the combination of a short-axis BLADE k-space trajectory with parallel imaging. Sensitivity to motion and flow is reduced by acquiring the coil calibration data for each blade by a second EPI echo train between excitation and diffusion sensitization. The traverse speed during the acquisition of the sufficiently sampled calibration data is adjusted to the undersampled imaging echo train by reducing the echo spacing and the number of samples per readout.

                  2711.     Comprehensive GRAPPA

Jun Miao1, Wilbur C. K. Wong1,2, Donglai Huo3, David L. Wilson1,4

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA; 2Computer Science and Engineering, The Hong Kong University of Science and Technology, Kowloon, Hong Kong; 3Keller Center for Imaging Innovation, St. Joseph Hospital and Medical Center, Phoenix, AZ, USA; 4Radiology, University Hospitals of Cleveland, Cleveland, OH, USA

In Parallel MR k-space reconstruction, most algorithms like GRAPPA make a common assumption that unsampled signals can be explained linearly and globally by sampled one. However, this global assumption is invalid theoretically. To better model the relationship between sampled and unsampled signals, we locally fit the linear function within clusters on a mathematically sound framework, Geographically Weighted Regression. Simulated and acquired MR data with different image contents and acquisition schemes including MR tagging data were tested. Results showed that comprehensive GRAPPA can significantly and robustly improve the image quality in parallel MR imaging.

                  2712.     Quantitative G-Factors for Radial GRAPPA

Felix A. Breuer1, Nicole Seiberlich2, Martin Blaimer1, Peter M. Jakob1,3, Mark A. Griswold2

1Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Germany; 2Dept. of Radiology, Case Western Reserve University, Cleveland, OH, USA; 3Dept. of Experimental Physics 5, University of Würzburg, Würzburg, Germany

In this abstract,a framework for quantitative estimation of the non-uniform noise enhancement in GRAPPA reconstructions (the so-called GRAPPA g-factor) where multiple kernels are employed is introduced. This abstract includes g-factor calculations for radial GRAPPA reconstructions where multiple kernels are required due to the lack of Cartesian symmetry in the undersampled k-space. This concept allows one to quantitatively estimate the noise characteristics of non Cartesian GRAPPA prior to the actual GRAPPA reconstruction and therefore can help to identify the optimal GRAPPA reconstruction parameters

                  2713.     Simplified Iterative GRAPPA for Fast and Robust Parallel MRI with Arbitrary Trajectories

Martin Blaimer1, André Fischer1, Philipp Ehses2, Nicole Seiberlich3, Mark A. Griswold3, Peter M. Jakob1,2, Felix A. Breuer1

1Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Bavaria, Germany; 2Department of Experimental Physics 5, University of Würzburg, Würzburg, Bavaria, Germany; 3Department of Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, USA

Generalized solutions for parallel MRI using arbitrary k-space trajectories are mainly based on iterative algorithms, such as conjugate gradient SENSE. In addition, an iterative GRAPPA approach for arbitrary k-space sampling has been presented. However, the reconstruction process includes several steps, including gridding, GRAPPA convolution and re-sampling operations during each iteration. Here, we present a simplified iterative GRAPPA algorithm. The key element is a GRAPPA-operator gridding (GROG) step prior to the iteration loop which removes the need for gridding and resampling during each iteration. This allows a robust and computationally efficient reconstruction of accelerated non-Cartesian MRI experiments.

                  2714.     Creation of Arbitrary Spatial Harmonics Though the Combination of Orthogonal Weights (CASHCOW):  a Generalized Direct GRAPPA Approach for Non-Cartesian Data

Nicole Seiberlich1, Felix A. Breuer2, Stephen R. Yutzy3, Martin Blaimer2, Mark A. Griswold1

1Radiology, Case Western Reserve University, Cleveland, OH, USA; 2Research Center Magnetic Resonance Bavaria (MRB), Wuerzburg, Germany; 3Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA

Non-Cartesian parallel imaging is complicated by the varying locations of acquired points with respect to each Cartesian k-space location. The method proposed here uses the GRAPPA Operator formalism to generate GRAPPA weights which perform a shift directly from the arbitrarily located non-Cartesian points to a Cartesian location, without necessitating either an iterative reconstruction or a coil map. This method provides a link between GRAPPA and GROG, providing GRAPPA-like stability over large shift distances along with flexibility and limited weight set requirements of GROG. CASHCOW is demonstrated in vivo with a radial acquisition for acceleration factors of up to R=4.

                  2715.     Anisotropic Kernel Support for Improved Fast GRAPPA Imaging

Jun Miao1, Wilbur C. K. Wong2, David L. Wilson1,3

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA; 2Computer Science and Engineering, The Hong Kong University of Science and Technology, Kowloon, Hong Kong; 3Radiology, University Hospitals of CLeveland, Cleveland, OH, USA

In Parallel MR k-space reconstruction, very commonly an isotropically shaped kernel is used. However, there is anisotropy in k-space due to inhomogeneous coil sensitivities or anisotropic imaged objects. We propose a method to determine anisotropy in k-space, and use it for GRAPPA kernel design. Experimental results show that GRAPPA with anisotropic kernel support can significantly improve reconstruction image quality, particularly in highly accelerated data sets.

                  2716.     Optimizing the Net Acceleration of GRAPPA and PEAK-GRAPPA

Simon Bauer1, Michael Markl1, Bernd Andre Jung1

1Department of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

Many parallel imaging methods measure additional phase encoding lines for autocalibration of the reconstruction parameters to the coil sensitivities. This leads to a reduction of the net acceleration of the measurement. The aim of this work was to systematically evaluate trade-offs between undersampling factor and number of calibration lines for a given net acceleration for GRAPPA and PEAK-GRAPPA. Analysis was performed for in-vivo cardiac imaging and using a specially designed moving phantom. Optimal results regarding RMSE and noise were provided by the reconstructions using the smallest possible undersampling factor and number of calibration lines.

                  2717.     Field Strength and B1 Influence on Kernel Support Selection for Highly Accelerated GRAPPA Reconstruction

Jun Miao1, Wilbur C. K. Wong2

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA; 2Computer Science and Engineering, The Hong Kong University of Science and Technology, Kowloon, Hong Kong

High magnetic field can provide an improvement of signal-to-noise ratio (SNR) in the MR images but creates larger B1 inhomogeneity, and its influence on k-space signals structure that affects kernel support selection remains unclear. Very commonly, an isotropic shape kernel is used in diffuse applications and field strengths. In this paper, we study k-space signals anisotropy in 3T and 7T MR phantom data sets, and compared our proposed anisotropic kernel to isotropic kernel for highly accelerated GRAPPA reconstruction. Experimental results showed that k-space signals anisotropy is correlated to B1 inhomogeneity, and anisotropic kernel is preferred over isotropic kernel for highly accelerated GRAPPA reconstruction, particularly in 7T MR data set.

                  2718.     A Fast Jump Move Graph Cut Reconstruction Algorithm for MR Parallel Imaging

Ashish Raj1, Gurmeet Singh2, Yi Wang, Ramin Zabih3

1Radiology, Weill Medical College of Cornell University, New York, NY, USA; 2Electrical Engineering, Cornell U; 3Computer Science, Cornell U, Ithaca, NY

Among recent parallel MR imaging reconstruction methods, a Bayesian method called Edge-preserving Parallel Imaging with GRAph cut Minimization (EPIGRAM) demonstrated significantly improved SNR, edge preservation and visual quality over conventional regularized SENSE method. Unfortunately, EPIGRAM requires a large number of steps in proportion to the number of intensity labels in the image, making it computationally expensive for images with high dynamic range. Here a new jump-move based graph-cut algorithm is presented that provides a logarithmic reduction in reconstruction time (typically 25-50 times) while maintaining SNR reported by EPIGRAM. Preliminary in-vivo validation for coronary angiography and short axis cine at acceleration factors of 3 and 4 are reported. Our proposal constitutes a critical step towards real-time reconstruction of cardiac MRI using the EPIGRAM approach.

                  2719.     IRES: Self-Calibrated Parallel Imaging with No Loss of Net Acceleration

Ek Tsoon Tan1, Roger C. Grimm1, Stephen J. Riederer1

1Radiology, Mayo Clinic, Rochester, MN, USA

In self-calibrated parallel imaging, the center of k-space is densely sampled, which reduces net acceleration. This reduction worsens at high acceleration factors. The concept of inversion recovery with embedded self-calibration (IRES) is to perform calibration acquisitions within delay intervals of inversion-prepared gradient echo with minimal perturbation to magnetization recovery. This results in no reduction in net acceleration due to self-calibration. It is shown in in-vivo brain imaging and in a phantom that at 2D accelerations of four, IRES provides (i) time savings at comparable image quality compared to standard self-calibration; or (ii) higher image quality at the same scan time.

                  2720.     Improved Self-Referenced Parallel MRI Imaging in EPI by Using UNFOLD to Remove Nyquist Ghosts

W Scott Hoge1, Huan Tan2, Robert A. Kraft2

1Radiology, Brigham and Women's Hospital, Boston, MA, USA; 2Wake Forest University School of Medicine, Winston-Salem, NC, USA

Self-referenced parallel MR imaging techniques have advantages in imaging scenarios that monitor changes. Their use with EPI is limited, however, due Nyquist or N/2 ghosting artifacts. These artifacts occur when data acquired on alternate positive and negative readout gradients is inconsistent. Recently, Nyquist ghost removal approaches that do not require a reference scan have been proposed. These methods interleave data from multiple frames---which reduces the effective temporal resolution. We propose using UNFOLD with a modified acquisition trajectory to eliminate Nyquist ghosts in multi-image EPI acquisitions, enabling the use of self-referenced pMRI reconstruction techniques while maintaining a high EPI imaging frame-rate.

                  2721.     Improved Reconstruction in Non-Cartesian Parallel Imaging by Regularized Nonlinear Inversion

Florian Knoll1, Christian Clason2, Martin Uecker3, Rudolf Stollberger1

1Institute of Medical Engineering, TU Graz, Graz, Austria; 2Institute for Mathematics and Scientific Computing, University of Graz, Graz, Austria; 3Biomedizinische NMR Forschungs GmbH, MPI für biophysikalische Chemie, Goettingen, Germany

It has been shown recently that it is possible to reformulate image reconstruction for Cartesian autocalibrated parallel imaging as a non-linear inversion problem. This approach allows joint optimization of the estimated coil sensitivities and image content, which improves image quality. We extend this method to arbitrary non-Cartesian sampling patterns. This eliminates the need to collect additional reference lines for the estimation of the coil sensitivities, and enables the use of even higher acceleration factors. Excellent removal of undersampling artifacts was achieved for human brain images (matrix size: 256x256, 32 radial projections, 4 channel head coil).

                  2722.     Rapid Parallel Imaging Reconstruction of NonCartesian Data

James G.  Pipe1

1Imaging Research, Barrow Neurological Institute, Phoenix, AZ, USA

A general non-Cartesian parallel imaging method is presented that iteratively enforces B1 map consistency and collected data consistency. Because it is linear and does not require gridding or degridding in each iteration, it is rapid and extendible to 3D in a computationally reasonable time.

                  2723.     Parallel Reconstruction of Projection MRI

Alan B. McMillan1

1University of Maryland School of Medicine, Baltimore, MD, USA

In Cartesian parallel imaging, sub-sampling of phase encoding lines induces aliasing in the phase encoding direction of the reconstructed image. Conveniently, this aliasing exists in only the phase encoding dimension, and provided adequate sensitivity information of receiver coils, the SENSE parallel imaging technique removes this aliasing through multiplication by an unfolding matrix. For non-Cartesian acquisitions, such as projection acquisitions, there is no simple implementation of the SENSE algorithm because the aliased energy is dispersed throughout the image. Presented here is a non-iterative parallel imaging method for undersampled projection acquisitions.

                  2724.     Robust SENSE Reconstruction Using Non-Local Regularization

Sheng Fang1, Li Zhao1, Xinlu Xu1, Kui Ying1, Jiangping Cheng1

1Engineering Physics, Tsinghua University, Beijing, China

A new regularization technique named non-local regularization is proposed for robust SENSE reconstruction. Unlike current regularization methods, the proposed method does not rely on any specific image model or prior image acquisition. It utilizes the information redundancy of an image and maximizes the consistence and similarity of pixel values within the image. The regularizing functional can automatically adapt to different local structures within an image. The phantom simulation and MR experiments results demonstrate that this method can effectively suppress noises in SENSE reconstruction with well-preserved image details. The image quality is better than the popular Total Variation regularization.

                  2725.     Prewhitening and Maximum Likelihood Estimation for Root-Sum-Of-Squares Images

Ryan Fobel1, Greg J. Stanisz1

1Sunnybrook Research Institute, Toronto, ON, Canada

Magnitude bias presents a serious challenge for quantitative imaging techniques that require data points close to the noise floor (e.g. DTI, qT2, MT, etc.). This bias becomes increasingly worse with the number of coils in a sum-of-squares reconstruction. This work presents a fitting strategy based on Maximum Likelihood theory to significantly reduce this bias. It relies on prior knowledge of the noise distribution from a reference scan and can account for correlation between coils with a prewhitening approach. Results from Monte Carlo simulations and a T2-decay experiment are presented.

                  2726.     A Point Spread Function Based Measure of Artefact in Parallel MRI Reconstruction

Linghe Yang1, Phil Robson2, Charles McKenzie1

1Medical Biophysics, University of Western Ontario, London, Ontario, Canada; 2Department of Radiology, Beth Israel Deaconess Medical Center, Boston, MA, USA

This paper illustrates the use of Point Spread Function (PSF) to assess the image reconstruction quality of Parallel MRI, specifically GRAPPA. Image quality in parallel imaging is most often quantified by the SNR and the noise amplification g-factor. However, this neglects residual artefact in the reconstruction, which also has an important impact on image quality. Artefact in the final image after parallel imaging reconstruction may be quantified by calculating a PSF. We demonstrate a metric of the PSF that includes the spin density of the imaged object to better quantify the residual artefact in a parallel imaging reconstruction.

                  2727.     Feasibility of Direct Virtual Coil (DVC) Reconstruction for 3-D Imaging

Philip James Beatty1, Joseph Y. Cheng2, Ajit Shankaranarayanan1, Ananth Madhuranthakam3, Huanzhou Yu1, Ersin Bayram4, Shaorong Chang4, Jean H. Brittain5

1Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 2Electrical Engineering, Stanford University, Stanford, CA, USA; 3Applied Science Laboratory, GE Healthcare, Boston, MA, USA; 4GE Healthcare, Waukesha, WI, USA; 5Applied Science Laboratory, GE Healthcare, Madison, WI, USA

The Direct Virtual Coil (DVC) reconstruction method improves the computational efficiency of data-driven parallel imaging reconstruction, while maintaining good image quality, by forgoing the computationally expensive ‘coil-by-coil’ approach in favor of directly synthesizing ‘virtual coil’ data. For this work, the DVC method is implemented to support 3-D data sets with 2-D acceleration. Results indicate that image quality comparable to coil-by-coil reconstructions can be obtained for 3-D imaging.

                  2728.     Efficient Large-Array K-Domain Parallel MRI Using Channel-By-Channel Array Reduction

Jim Ji1, Shuo Feng1

1Texas A&M University, College Station, TX, USA

Large coil arrays with up to 64 or 128 elements have enabled parallel MRI at unprecedented speed. However, applying k-domain parallel MRI such as the GRAPPA methods in these systems has been rare due to the computational challenge. In this work, we develop an efficient reconstruction method for k-space based parallel MRI with large arrays using array reduction. Experimental results show the new method can dramatically reduce processing time with minimal degradation of reconstruction quality.

                  2729.     Prospective SNR Optimization in k-T-Based Sensitivity-Encoded Dynamic Imaging Using a Fast Geometric Algorithm

Behzad Sharif1, John Andrew Derbyshire2, Yoram Bresler1

1Coordinated Science Lab, Department of Electrical & Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, USA; 2Cardiovascular Branch, NHLBI, National Institutes of Health, DHHS, Bethesda, MD, USA

The acceleration in k-t-based dynamic parallel imaging results from multi-fold undersampling (relative to the Nyquist rate) provided by the k-t sampling scheme. Such sampling results in aliasing of the signal spectrum in the reciprocal domain (x-f space). We propose a novel algorithm for prospective design of SNR-optimal k-t sampling patterns and study its performance in-vivo for real-time cardiac parallel imaging. All computations are based on geometry of overlap patterns in x-f space and are independent of coil sensitivities. This

                  2730.     Dynamically Regularized TSENSE Improves Image Quality in Parallel MRI

Martin Blaimer1, Felix A. Breuer1, Peter M. Jakob1, Mark A. Griswold2, Peter Kellman3

1Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Bavaria, Germany; 2Department of Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, USA; 3Laboratory of Cardiac Energetics, National Institutes of Health, National Heart, Lung and Blood Institute, Bethesda, MD, USA

The proposed dynamic regularized TSENSE parallel MRI method improves image quality by utilizing temporal signal correlations. Parallel MRI is applied only to the dynamic portion of the signal as determined by changes from the temporal average, also using the signal change for regularizing the solution. A first pass estimate with full temporal and spatial resolution is derived using the TGRAPPA method without the requirement for separate training data. The full resolution estimate is used for dynamic regularized SENSE reconstruction after removal of the temporal average (DC term) and SNR based regularization uses the difference between the current and the temporal average image.

                  2731.     Time Efficient Parallel Imaging for 2D Multi-Slice MRI with Continuously Moving Table

Matthias Honal1, Ute Ludwig1, Jochen Leupold1

1Dept. of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

The acquisition time is an important issue in MRI with continuously moving table, especially for acquisitions performed during multiple breath holding phases. In this study the application of the recently introduced z-GRAPPA method to MRI with continuously moving table is evaluated. This method increases the efficiency of parallel imaging for 2D multi-slice acquisitions by avoiding the measurement of additional calibration data. Instead k-space lines from adjacent slices are used for calibration. For a clinically relevant protocol, a speedup of 18% compared to conventional GRAPPA was achieved.

                  2732.     Temporal Filtering Effects in Auto-Calibrating TSENSE

Peter Kellman1, Martin Blaimer2, Felix Breuer2, Mark Griswold3

1National Institutes of Health, Bethesda, MD, USA; 2Research Center Magnetic Resonance Bavaria, Wurzburg, Germany; 3Department of Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, USA

The auto-calibrating TSENSE method may lead to temporal filtering in the form of signal nulls in the temporal spectra of the dynamic SENSE images due to errors in the coil sensitivity estimates. The TGRAPPA reconstruction using the same data has no temporal nulls but does not meet the pixel-wise optimum SNR that the SENSE method does. A hybrid auto-calibrating TSENSE method is proposed which combines TSENSE and TGRAPPA to achieve improved SNR without undesired temporal filtering.

                  2733.     3101

Ke Liu1, Jingxin Zhang1, Ran Yang2, Cishen Zhang3

1Department of Electrical and Computer Systems Engineering, Monash University, Melbourne, VIC, Australia; 2School of Infomation Science and Technology, Sun Yat-Sen University, Guangzhou, Guangdong, China; 3School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore

The incorporation of dynamic imaging and parallel imaging techniques is highly desirable and intensively researched. We propose a 3D IIR-GRAPPA method which significantly reduces the number of ACS lines required and improves the quality of reconstructed images with less visual artefact. This method fits an IIR kernel model to reconstruct the unacquired k-t space data points for acceleration purpose. Experimental results demonstrate the advantages of the proposed method over the existing ones.

                  2734.     High Temporal and Spatial Resolution 3D CE-MRA of the Peripheral Vascualture Using 12-Fold 2D SENSE

Clifton R. Haider1, James F. Glockner1, Anthony W. Stanson1, Casey P. Johnson1, Stephen J. Riederer1

1Radiology, Mayo Clinic, Rochester, MN, USA

In imaging a transient phenomenon such as the passage of a contrast bolus, MR image formation benefits from being temporally compact. Improvements in parallel imaging can therefore improve the efficiency with which contrast is mapped to k-space and decrease the susceptibility of the sequence to temporal blurring and venous contamination. We hypothesize that improved coil configurations can allow for robust 12-fold SENSE-accelerated time-resolved 3D CE-MRA in the peripheral vasculature. We demonstrate high quality 1 mm isotropic resolution time-resolved MR angiograms of the lower legs generated with frame times under 4 sec and image acquisition times under 16 sec.

                  2735.     Efficient Parallel Imaging Strategies for Time-Of-Flight Venography with Continuously Moving Table

Matthias Honal1, Sandra Huff1

1Dept. of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

Subtraction TOF venography with continuously moving table is an efficient method to image the veins of the whole legs with one measurement. The acquisition time is an important issue for that method. In this study two variants of the GRAPPA algorithm which do not require extra calibration data for coil weight computation are applied to accelerate the substraction TOF venography method. Without a significant loss in image quality a speedup of 9% compared to conventional GRAPPA could be achieved.

                  2736.     High Resolution Single-Shot Diffusion Weighted Imaging with a Combination of Zoomed EPI and Parallel Imaging

Robin Martin Heidemann1, Thorsten Feiweier2, Alfred Anwander1, Fabrizio Fasano3,4, Josef Pfeuffer5, Robert Turner1

1Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany; 2Siemens Healthcare Sector, Erlangen, Germany; 3Fondazione Santa Lucia, Rome, Italy; 4Siemens Medical Solutions, Milano, Italy; 5Siemens Medical Solutions, Charlestown, MA, USA

High resolution diffusion weighted imaging is affected by blurring due to T2* relaxation and distortions due to off-resonance effects. The use of parallel imaging can reduce those effects resulting in significantly improved image quality. However, parallel imaging is not without limitations and even with large phased arrays it is still a challenge to obtain high acceleration factors greater than four in a single dimension. The combination of a zoomed approach and parallel imaging can overcome this limitation enabling high resolution DW imaging with single-shot EPI.

                  2737.     Parallel Line-Scan Echo-Planar Spectroscopic Imaging

Yoshitaka Bito1, Toru Shirai1, Satoshi Hirata1, Yo Taniguchi1, Koji Hirata1, Yosuke Otake1, Yoshihisa Soutome1, Hisaaki Ochi1, Toshihiko Ebisu2, Yuko Kawai3, Masahiro Umeda3, Toshihiro Higuchi4, Chuzo Tanaka4

1Central Research Laboratory, Hitachi, Ltd., Kokubunji-shi, Tokyo, Japan; 2Neurosurgery, Nantan General Hospital, Kyoto, Japan; 3Medical Informatics, Meiji University of Integrative Medicine, Kyoto, Japan; 4Neurosurgery, Meiji University of Integrative Medicine, Kyoto, Japan

A parallel line-scan echo-planar spectroscopic imaging (PLSEPSI) technique has been developed to increase the number of lines, spectral resolution, and scan speed.@The technique uses simultaneous acquisition of multiple lines by a pair of parallel slice excitations and discrimination of the lines at crossing slices by using sensitivity maps of multiple receive coils. Interestingly, the signal-to-noise ratio (SNR) of PLSEPSI improves under certain conditions, and this SNR improvement never occur in the case of conventional parallel imaging with reduced phase encoding. Acquisition of spectroscopic images in half the measurement time is demonstrated using a 2-channel receive array coil by 2-line acquisition of PLSEPSI.

                  2738.     Parallel Imaging with RASER Using Multiband Frequency-Modulated Excitation Pulses

Tram Nguyen1, Ute Goerke2, Steen Moeller2, Kamil Ugurbil2, Michael Garwood2

1High-Field Magnetic Resonance Center, Max-Planck Institute for Biological Cybernetics, Tübingen, Germany; 2Center for Magnetic Resonance Research, Minneapolis, MN, USA

The many advantages of the recently proposed RASER sequence have been demonstrated. Hence, RASER holds great promises for functional MRI (fMRI), particularly for studies of the orbital-frontal cortex and other brain regions near air cavities, which cause distortion and signal loss in conventional EPI methods. However, the single-shot RASER sequence implemented so far inherently presents a set of temporal and spatial limitations that hinders it feasibility and full potential for fMRI applications. It is believed that parallel imaging will help overcome such restrictions. In this work, the RASER acquisition and reconstruction scheme is extended for parallel imaging using tailored pulses for simultaneous multi-band excitation.

Quantitative Imaging
Exhibit Hall 2-3                    Monday 14:00-16:00

                  2739.     Correlation of Tissue Proteomic Profiles with MR Measures in a Rat Brain

Tuhin K. Sinha1, Zhengyu K. Yang2, Erin H. Seeley3, Mary Loveless4, Daniel C. Colvin4, Richard M. Caprioli3, John C. Gore1

1Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, USA; 2Chemical and Physical Biology, Vanderbilt University, Nashville, TN, USA; 3Mass Spectrometry Research Center, Vanderbilt University, Nashville, TN, USA; 4Biomedical Engineering, Vanderbilt University, Nashville, TN, USA

We demonstrate a method to interrogate and correlate proteomic information from matrix-assisted laser desorption/ionization imaging mass spectrometry (MALDI IMS) in light of corresponding in vivo MR measures. MALDI IMS is capable of generating an image where each pixel represents a spectrum of constituent tissue proteins. We examine these spectra, in a co-registered, voxel-by-voxel basis to investigate contrast variations in MR. We hypothesize that the intrinsic variation in MR that is related to heterogeneous macromolecular distribution in tissues can be elucidated using MALDI IMS. We demonstrate quantitative correlation of MALDI IMS data in a rat brain with T1 measures, and show the varying proteomic profiles which correspond with these correlations.

                  2740.     T2 Estimation for Small Lesions Using a Model-Based Reconstruction with Sparsifying Penalty Functions and Highly Under-Sampled Radial FSE Data

Chuan Huang1, Ali Bilgin2, Christian Graff3, Maria I. Altbach4

1Dept of Mathematics, University of Arizona, Tucson, AZ, USA; 2Department of Electrical and Computer Engineering, University of Arizona; 3Program in Applied Mathematics, University of Arizona; 4Dept of Radiology, University of Arizona

T2 weighted MRI is used clinically for characterization of various pathologies. It has been demonstrated that quantitative methods for measuring T2 values are superior to visual evaluation. However, many of the proposed methods to measure T2 values suffer from long acquisition times, motion-induced errors, low spatial resolution and/or low number of measured points on the T2 relaxation curve, etc. To overcome the problems we proposed a model-based approach with sparsifying penalty functions for estimating T2’s from radial FSE data which yields accurate T2 estimates of small lesions from highly undersampled data. This novel method has great potential for characterization of small neoplasms in the body where the acquisition time is restricted to a breath hold.

                  2741.     Optimisation of the Combined Gradient Echo/spin Echo (GESE) Sequence for the Measurement of T2 at 7.0 T

Eleanor F. Cox1, Penny A. Gowland1

1SPMMRC, School of Physics & Astronomy, University of Nottingham, Nottingham, Nottinghamshire, UK

The GESE sequence can be used to simultaneously measure T2 and T2Œ in the brain and is particularly useful in conditions of B1 inhomogeneity such as ultrahigh field. In order to maximize the potential of GESE for tissue characterization, it is necessary to optimize the sequence. Here we have investigated two different fitting algorithms for this sequence to identify which one gives the lowest error in fitted values of T2 and optimized sequence timings in terms of the number and spacing of gradient echoes and echo times used for measuring T2 and T2Œ in the brain at 7T.

                  2742.     Non-Regularized Multivoxel NNLS Is a Robust Analysis Approach to Quantitative T2

Thorarin A. Bjarnason1, Cheryl R. McCreary1, Jeff F. Dunn1, J Ross Mitchell1

1University of Calgary, Calgary, AB, Canada

Quantitative T2 (qT2) enables scientists to discern tissue microcompartments by measuring multiple T2 decays using a multiecho acquisition. qT2 is sensitive to myelin content. We show that traditional qT2 analysis underestimates the myelin water fraction (MWF) with decreasing SNR. We present a robust approach to qT2 analysis that does not underestimate the MWF with decreasing SNR, provides variance estimates for a single ROI, and has confidence intervals on the T2 distribution.

                  2743.     Localized In Vivo Fast Field-Cycling Relaxometry

Kerrin J. Pine1, Gareth R. Davies1, David J. Lurie1

1Bio-Medical Physics and Bio-Engineering, University of Aberdeen, Aberdeen, Scotland, UK

Fast field-cycling (FFC) can provide access to endogenous information not available from conventional fixed-field imagers. One example is the T1 dispersion plot, where T1 is measured over a range of evolution field strengths. We present a new pulse sequence using field-cycling saturation-recovery/inversion-recovery with an implementation of point-resolved spectroscopy (PRESS), allowing dispersion plots to be produced for a voxel selected from a pilot image. On a whole-body FFC-MRI system consisting of a 59 mT permanent magnet and coaxial resistive magnet, the sensitivity of the experiment is sufficient to obtain distinctive “quadrupole dips” in dispersion plots of protein-rich human tissue in vivo.

                  2744.     Voxel-Based Relaxometry for Cases of an Unresolved Epilepsy Diagnosis

Robert Karl Kosior1,2, Rachel Sharkey2,3, Paolo Federico2,4, Richard Frayne2,4

1Biomedical Engineering, University of Calgary, Calgary, Alberta, Canada; 2Seaman Family MR Centre, Foothills Medical Centre, Calgary Health Region, Calgary, Alberta, Canada; 3Radiology and Clinical Neurosciences, University of Calgary, Calgary, Alberta, Canada; 4Radiology and Clinical Neurosciences, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada

A patient may be diagnosed with epilepsy, but without a firm understanding of the seizure focus, or without confidence in the diagnosis. Voxel-based relaxometry (VBR) is a T2 relaxometry technique that may provide information to corroborate, or refute indeterminate information from the other sources. Our objective was to assess the performance of single-subject VBR at 3 T for cases of uncertain epilepsy diagnoses. We found that more VBR abnormalities occur in patients with a suspected focus. Our results show that single subject VBR can help identify seizure foci in patients in whom conventional investigations have failed to yield a diagnosis.

                  2745.     Comparison of Analysis of Brain Relaxation Times in Standard Space with Analysis in Individuals’ Real Space

Benjamin Segun Aribisala1, Andrew M. Blamire1

1Newcastle Magnetic Resonance Centre, Newcastle University, Newcastle upon Tyne, Tyne and Wear, UK

One commonly used image analysis technique is registration of images to a standard space. This has the major benefit that target regions are defined only once in standard space and can then be applied to any dataset. We hypothesize that the re-sampling and spatial smoothing used in the registration processes introduces partial volume effects (PVE) into the data under analysis biasing results. Here we compare this standard approach with an alternative method which transforms standard space target ROIs into real space where quantitative analysis is performed. We show that real space analysis suffers from less PVE than standard space analysis

                  2746.     Detection of Changes in T1 Values in Normal Brain During Normobaric Hyperoxia

Samantha Jane Mills1, Gerard Thompson1, Giovanni Alessandro Buonaccorsi1, Geoff J. Parker1, Alan Jackson1

1Imaging Science and Biomedical Engineering, School of Cancer and Imaging Sciences, University of Manchester, Manchester, UK

Molecular oxygen is paramagnetic and can therefore reduce observed T1 values when dissolved in blood and in tissue. Presented here are novel data from a pilot study at 3Tesla in four healthy human volunteers showing a significant decrease in the T1 values in cortical grey matter(CGM), deep white matter(DWM), thalamus(BG), and skeletal muscle(SkM) during administration of 100% oxygen versus air. Baseline mean T1 values were 832+/-53ms, 1175+/-120ms, 1553+/-57ms, and 962+/-82ms for DWM, CGM, BG, and SkM respectively. Following correction for baseline drift , mean hyperoxic δT1 was -8.5+/-2.3ms, -33.8+/-4.1ms, -16.4+/-6.5ms, and -10.1+/-3.3 in DWM, CGM, BG, and SkM respectively (p<0.05).

                  2747.     Off-Resonance Magnetisation Transfer Contrast MRI Using Fast Field-Cycling Technique

Chang-Hoon Choi1, Gareth R. Davies1, David J. Lurie1

1Bio-Medical Physics and Bio-Engineering, University of Aberdeen, Aberdeen, Scotland, UK

Magnetisation transfer contrast (MTC) imaging normally employs off-resonant irradiation with varying RF offset frequency but maintaining a desired constant RF field strength (B1). This presents the main technical difficulty of MTC at low field, because the larger offset frequencies are likely to be outside the bandwidth of the RF transmit system, causing B1 to vary with the frequency offset. In this work, we demonstrate a novel off-resonance irradiation method using the fast field-cycling technique which permits to counter this problem. The results obtained by the new technique agreed well with those obtained by the conventional technique.

                  2748.     Optimization and Validation of a Constrained Reconstruction Algorithm for Rapid Whole-Brain Cross-Relaxation Imaging at 3.0 Tesla

Hunter R. Underhill1, Chun Yuan1, Vasily L. Yarnykh1

1University of Washington, Seattle, WA, USA

Cross-relaxation imaging quantitatively maps the parameters describing magnetization transfer between free and bound protons in tissues. Whole-brain parametric maps of the fraction of macromolecular protons (f) and the rate constant (k) have been enabled at 1.5T by constraining the transverse relaxation time of both the free (T2F) and bound (T2B ) pools. We determine the optimized constraints for T2F and T2B to enable rapid, whole-brain parametric maps of f and k at 3.0T. In addition, we report the average values of f and k across various gray and white matter structures from a group of volunteers imaged at 3.0T.

                  2749.     Direct Quantitative Comparison Between Cross-Relaxation Imaging and Diffusion Tensor Imaging of the Human Brain

Hunter R. Underhill1, Chun Yuan1, Vasily L. Yarnykh1

1University of Washington, Seattle, WA, USA

Diffusion tensor imaging (DTI) has been proven to be an effective tool for the visualization of white matter fiber tracts in the human brain based on the anisotropic diffusion effect. Parameters acquired during cross-relaxation imaging (CRI) that describe magnetization transfer have also been shown to have a close association with major fiber tracts in the human brain. In this 3.0T study, we report the results of quantitative comparisons between the CRI parameters that describe the fraction of macromolecular protons and the magnetization transfer rate constant and the principle scalar metric of DTI, fractional anisotropy, in the human brain in vivo.

                  2750.     19F Signal Amplification by Heteronuclear Polarization Transfer

Florian Schmid1, Manuel Tsotsalas2, Michael Kuhlmann3, Carsten Höltke1, Christoph Bremer1, Walter Heindel1, Andres Guerrero-Martinez2, Jochen Keupp4, Michael Schäfers3, Luisa De Cola2, Cornelius Faber1

1Institute for Clinical Radiology, University Hospital Münster, Münster, Germany; 2Physikalisches Institut and NRW Graduate School of Chemistry, Westfälische Wilhelms-Universität Münster, Münster, Germany; 3European Institute for Molecular Imaging, Westfälische Wilhelms-Universität Münster, Münster, Germany; 4Philips Research Europe, Hamburg, Germany

Signal enhancement by NOE was investigated in 19F-MRI using 2-Fluoro-2-deoxy-D-glucose (FDG) and 2,2,2-Trifluoroethanol (TFE). The 19F signal was considerably increased by steady-state NOE preparation in spectroscopy and imaging sequences. TFE was loaded in novel potential carriers of contrast agents, zeolite L crystal nanocontainers. FDG was applied to mice and 19F images of the fixed mouse heart could be aquired at a clinical MRI scanner at 3T.

                  2751.     Incorporation of ADC Information Into SMM-Based MREIT for Small Animal Conductivity Imaging

Mark Jason Hamamura1, Orhan Nalcioglu1, Lufti Tugan Muftuler1

1Tu & Yuen Center for Functional Onco-Imaging, University of California, Irvine, CA, USA

In this study, we investigated the incorporation of apparent diffusion coefficient (ADC) data into the sensitivity matrix method (SMM) for magnetic resonance electrical impedance tomography (MREIT) of a tumor-bearing rat. The results indicate that such an incorporation allows us to reconstruct sensible conductivity images, something which has remained problematic for in vivo MREIT.

                  2752.     SENSE Accelerated MREIT with Optimized RF Coil Design

Lutfi Tugan Muftuler1, Gang Chen1, Mark J. Hamamura1, Orhan Nalcioglu1

1Center for Functional Onco-Imaging, University of California, Irvine, CA, USA

SENSE imaging technique is adapted to MR Electrical Impedance Tomography (MREIT) to shorten acquisition times. A phased array coil was designed and optimized for this application and MREIT data was reconstructed from the SENSE accelerated images. A quasi-intrinsic decoupling scheme was used to minimize couplings between coil elements and the geometry was optimized to achieve high and uniform SNR. Data was acquired with and without SENSE acceleration and the results were compared. Phantom studies demonstrated that conductivity maps could be reconstructed from MREIT data acquired with 2.7 times SENSE acceleration with no discernible artifacts.

                  2753.     Bright-Iron Imaging Applying an Off-Resonance Modulating Pre-Pulse (OMPP)

Volker Rasche1, Vinzenz Hombach1, Sonu Sharma1, Gerlinde Schmidtke-Schrezenmeier2, Axel Bornstedt1

1Internal Medicine II, University Ulm, Ulm, Germany; 2Transfusion Medicine, University Ulm, Ulm, Germany

A new off-resonance sensitive pre-pulse is presented. The pre-pulse enables direct visualization of off-resonant spins. The pre-pulse was applied to imaging of local iron oxide particle distributions at different concentrations and to direct visualization of stents.

                  2754.     Positive Iron Contrast by Flow Enhanced Off-Resonance-Saturation for Remote Detection of Iron Based Contrast Agents

Philipp Krämer1, Xavier Helluy1, Esra Lang1, Peter Michael Jakob1

1Department of Experimental Physics V, University of Würzburg, Würzburg, Bavaria, Germany

Hypointense contrast originating from iron-loaded atherosclerotic plaques within a vessel wall have been observed in contrast-enhanced gradient echo based MRA images of vessel lumen, allowing a fast remote detection of iron-loaded plaques. We demonstrate here on a flow phantom how a more specific positive iron contrast mechanism - flow enhanced off-resonance saturation (feORS) -, combined with MRA might facilitate the detection of iron based contrast agents confined within a small vessel wall volume.

                  2755.     Switching of MRI Contrast Agents with Ultrasound

Nouri Elmiladi1, Christian Hoehl1, Karl Maier1

1Helmholtz-Institut für Strahlen- und Kernphysik (HISKP), Rheinische Friedrich-Wilhelms-Universität, Bonn, Germany

A method involving the application of ultrasound (US) while performing proton nuclear magnetic resonance (1HNMR) spectroscopy in the presence of antibody coated magnetic nanoparticles has been developed. Especially prepared magnetic nanoparticles act as an US driven radio frequency antenna, emitting photons of US frequency. Influence of resonant US on 1HNMR using the asymmetric magnetic nanoparticles is examined through the determination of the relaxation rate by an inversion recovery sequence. Measurements are obtained for two different sizes of magnetic particles. A significant increase of the relaxation rate is observed when using smaller sized particles.

                  2756.     Numerical Study of the Distant Dipolar Field

Stefan Kirsch1, Peter Bachert2

1Computer Assisted Clinical Medicine, Faculty of Medicine Mannheim, University of Heidelberg, Mannheim, Germany; 2Dept. of Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany

In this numerical study we present 2D presentations of the distant dipolar field (DDF) for various sample geometries and different experimental conditions (e.g., gradient settings). The calculations show that the DDF depends on the geometry and/or spatial distribution of the magnetization within the sample. The spatial structure of the DDF shows the known dependence on distance (“correlation distance” d). The simulations suggest that d should be considered as an average value rather than an exact distance. Our results are consistent with previously published 1D representations of the DDF and can be used to find optimized spatial magnetization distributions which result in an enhanced DDF.

                  2757.     CRAZED Signal Dependence on Correlation Distance and Sample Orientation in Rat Sciatic Nerve

Arman S. Kussainov1, Mark D. Does2, Nellie E. Byun3, John C. Gore1,3, Daniel F. Gochberg1,3

1Physics, Vanderbilt, Nashville, TN, USA; 2Biomedical Engineering, Vanderbilt, Nashville, TN, USA; 3Radiology, Vanderbilt, Nashville, TN, USA

We measured the CRAZED signal from ex vivo rat sciatic nerve while varying the sample orientation and the gradient induced correlation distance, and we compared these results to a doped water cylindrical sample of the same size but lacking any internal structure. The results for the sciatic nerve have characteristics indicative of cylinders at both the ten and hundreds of micrometers distance scales, re&#64258;ecting both axon and the tibial/peroneal fascicle structures that compose the nerve. These results support the view that CRAZED methods are able to probe a range of distance scales not available in other magnetic resonance methods.

                  2758.     Observation of Signal Dips in IMQC Study

Zhong Chen1, Shengchun Zhang1, Shuhui Cai1

1Physics, Xiamen University, Xiamen, Fujian, China

The CRAZED-like sequences were qualitatively investigated with the strength of coherence selection gradients (CSGs) varying from a very small value to a large value. Obvious signal dip was observed at a specific CSG value, even when the influence of the inhomogeneous fields was eliminated. A detailed study shows that the position of the dip is directly related to the size of sample tube. Results demonstrate the appearance of the dip is a complex effect of CSG, inhomogeneous field and sample geometry£¬which can be utilized to explore the structural feature of the sample in MRI applications.

High Field Imaging
Exhibit Hall 2-3                    Tuesday 13:30-15:30

                  2759.     T1 Contrast in the Human Brain at 7 Tesla

Michael Wyss1, David Otto Brunner1, Anne Morel2, Klaas Paul Pruessmann1

1Institute for Biomedical Engineering, University and ETH Zürich, Zurich, Switzerland; 2Laboratory for Functional Neurosurgery, University Hospital Zurich, Zurich, Switzerland

MRI in humans at 7Tesla offers not only high baseline sensitivity but also interesting contrast behavior. In the present work we report sequence considerations for robust T1-weighted brain imaging at 7T. An inversion recovery (IR) approach with an adiabatic inversion pulse was chosen. The inversion pulse was optimized between the competing goals of robustness against B1 inhomogeneity, bandwidth and SAR. It was found that the high field yielded not only high SNR and resolution but also surprisingly strong contrast throughout the brain. It permitted the distinction of anatomical structures which have been hard to recognize at lower field strengths.

                  2760.     Magnetization Prepared 3D FLAIR Imaging at 7.0 Tesla

Fredy Visser1,2, Jaco Zwanenburg3, Hans Hoogduin3, Peter Luijten3

1University Medical Centre Utrecht, Utrecht, Netherlands; 2Philips Healthcare, Best, Netherlands; 3UMC

This study shows that with dedicated magnetization preparation pre-pulses it is possible to make High Resolution 3D TSE FLAIR images at 7Tesla. As FLAIR imaging should be routinely included in many clinical studies, this finding will have a positive impact on the applicability of 7T MRI in clinical studies.

                  2761.     3D FLAIR at 7 Tesla Highlights Peripheral Layers of the Cortex

Jaco J.M. Zwanenburg1,2, Fredy Visser1,3, Taro Takahara1, Wim G.M. Spliet4, Peter R. Luijten1

1Dept. of Radiology, University Medical Center Utrecht, Utrecht, Netherlands; 2Image Sciences Institute, University Medical Center Utrecht; 3Philips Healthcare, Best, Netherlands; 4Dept. of Pathology, University Medical Center Utrecht

We observed that 3D FLAIR at 7T yields high signal in a layer different from the line of Gennari. This bright line is visible at the periphery of the cortex, and also around the ventricles, and is therefore consistent with the demyelinated layers I and II in the cortex and with the ependyma around the ventricles. The purpose of this work to describe this contrast in the FLAIR images and explore its anatomical origin.

                  2762.     Reconstruction of Multi-Channel MR Magnitude and Phase Images Using Dual-Echo Sequence at 7 Tesla

Chan Hong Moon1, Kyongtae Ty Bae1,2

1Radiology, University of Pittsburgh, Pittsburgh, PA, USA; 2Bioengineering, University of Pittsburgh, Pittsburgh, PA, USA

MR magnitude and phase images were reconstructed with improved intensity homogeneity and phase SNR by the highly efficient algorithm combined with dual-echo acquisition.

                  2763.     Phase Contrast in the Human Brain and Its Field Dependence

Ana Maria Oros-Peusquens1, Miriam Rabea Kubach1, N. Jon Shah1

1Institute of Neurosciences and Biophysics, Research Centre Juelich, Juelich, Germany

Phase information, though not fully understood, is intrinsically contained in MR data and is being increasingly used to enhance data quality. In addition to specific ROIs, in this study we have investigated the field dependence of the phase contrast of the entire brain at three fields (1.5T, 3T and 4T). We have used field maps and not single-TE phase images which are more difficult to compare between fields. The field dependence of the phase shift of the individual ROIs is approximately linear, suggesting a dominant susceptibility effect. However, the changes observed in the whole-brain distribution of the phase require further investigation.

                  2764.     Triple-Layer Appearance of Human Cerebral Cortices on Phase-Difference Enhanced Imaging Using 3D Principle of Echo Shifting with a Train of Observations (PRESTO) Sequence

Tetsuya Yoneda1, Toshinori Hirai2, Yasuhiro Hiai1, Hiroki Arimura1, Yoko Kitajima3, Tomoyo Ideta4, Mika Kitajima2, Yasuyuki Yamashita2

1School of Health Sciences, Kumamoto University, Kumamoto, Japan; 2Department of Diagnostic Radiology, Graduate School of Medical Sciences, Kumamoto University, Kumamoto, Japan; 3Kumamoto Neurosurgery Hospital, Kumamoto, Japan; 4Saiseikai Kumamoto Hospital, Kumamoto, Japan

The laminar patterns in the human cerebral cortices have not been investigated at 3T MRI in living humans. We developed a novel technique of phase-difference enhanced (PADRE) imaging to enhance a phase difference of tissue structures. We evaluated axial 1-mm-thick PADRE images obtained using a 3T MR scanner in 6 healthy volunteers. We adjusted parameters of PADRE imaging using high-pass filter and phase enhancement techniques in order to enhance the laminar patterns of the cortices. In all 6 cases a triple-layer appearance was frequently seen in all bilateral primary motor, sensory, and visual cortices.

Exhibit Hall 2-3                    Wednesday 13:30-15:30

                  2765.     Robust Fast Clinical Neurological Examination Using Golden Angle Ordered Radial IR-TrueFISP

Stephen R. Yutzy1,2, Nicole Seiberlich2, Vikas Gulani2, Jeffrey L. Duerk1,2, Mark A. Griswold1,2

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA; 2Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, USA

A rapid steady state radial method is presented which allows estimation of “synthetic” images with multiple contrasts in an acquisition time of <5s per slice. A golden angle echo sharing reconstruction is used to reconstruct images with high temporal resolution which are used to fit tissue relaxation parameters and proton density. Perfectly registered synthetic images can be calculated with any arbitrary contrast retrospectively after the acquisition. This technique has the potential to dramatically decrease examination time per patient, thereby increasing patient comfort and compliance, decreasing interscan motion, and increasing patient throughput.

                  2766.     The Effect of B0 Inhomogeneity on the SSFP Dixon Technique: A Comaprison of Variants

Yin-Cheng Kris Huang1,2, Teng-Yi Huang3, Hing-Chiu Chang4, Hsiao-Wen Chung1,2

1Department of Electrical Engineering, National Taiwan University, Taipei City, Taiwan; 2Department of Radiology, Tri-Service General Hospital, Taipei City, Taiwan; 3Department of Electrical Engineering, National Taiwan University of Scienece and Technology, Taipei City, Taiwan; 4Applied Science Laboratory, GE Healthcare Taiwan, Taipei City, Taiwan

Images of fat-water separation can be obtained by combining the Dixon method and the balanced steady-state free precession (bSSFP) sequence, which is termed as SSFP Dixon method. The application of this method was successful at regions of relatively homogeneous B0. Another variant of this method was the dual-TR method, which was anticipated to have a better immunity to the shimming condition. In this study, we validated this concept with experiments of both methods at regions with significant B0 inhomogeneity.

                  2767.     Chimera Steady State Free Precession (Chimera SSFP)

Oliver Bieri1, Markus Klarhöfer1, Klaus Scheffler1

1Division of Radiological Physics, University of Basel Hospital, Basel, Switzerland

A new type of steady-state free precession (SSFP) sequence is introduced, termed chimera SSFP. The sequence consists of two alternating SSFP kernels (heads): odd TR-intervals feature a balanced SSFP (bSSFP) type of protocol, whereas even TR-intervals undergo gradient dephasing (unbalanced SSFP) and hence the name. Chimera SSFP features a peculiar frequency response profile with respect to the bSSFP interval of triangular shape. Potential applications of this new type of SSFP sequence are, among others, functional MRI (fMRI) and temperature mapping.

                  2768.     Multiple Acquisition Fat-Saturated Balanced Steady State Free Precession Imaging

Marcus T. Alley1, Shreyas S. Vasanawala1, Robert J. Herfkens1, Brian A. Hargreaves1

1Radiology, Stanford University, Stanford, CA, USA

While bSSFP imaging is advantageous in its ability to produce high SNR in short imaging times, its clinical application has been limited. The fat signal is typically unwanted, and the short TRs required to avoid banding artifacts lead to resolution limitations. This is especially true at higher field strengths where SAR considerations begin to force longer TRs. Previous methods have combined multiple bSSFP acquisitions with different linearly-increasing excitation phases to either reduce signal variations with frequency or suppress fat. We propose the addition of fat suppression to a multiple phase-cycled, 3D bSSFP acquisition to address both issues simultaneously.

Exhibit Hall 2-3                    Thursday 13:30-15:30

                  2769.     Investigation of  SPIO T1-Signature: Positive Contrast Using Ultrashort TE Imaging

Olivier Maciej Girard1, Jiang Du1, Robert Frederick Mattrey1

1Department of Radiology, University of California, San Diego, CA, USA

Superparamagnetic iron oxide (SPIO) nanoparticles are efficient contrast agents for molecular and cellular MR imaging. They are usually considered to be negative contrast agents due to their strong T2* effect, but they also have intrinsic T1 shortening properties that can produce positive contrast using appropriate MR pulse sequences. In the present study Utrashort TE (UTE) sequences were used to demonstrate the positive contrast even at high iron concentrations. Saturation Recovery prepared UTE sequences revealed an efficient contrast pattern. In addition to the usual T2* contrast it is possible to use the T1-signature of SPIO agents to improve detection specificity and localization.

                  2770.     Imaging of Short T2 and Ultra-Short T2 Tissues Using R2* Map

Weitian Chen1, Huanzhou Yu1, Atsushi Takahashi1, Anja Brau1, Jeffrey Stainsby2, Ann Shimakawa1, Brian A. Hargreaves3, Garry E. Gold3

1Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 2Applied Science Laboratory, GE Healthcare, Toronto, Canada; 3Radiology, Stanford University, Stanford, CA, USA

Ultrashort echo time (UTE) imaging techniques have potential clinical applications for imaging of short T2 species. To increase contrast of short T2 species, it is usually necessary to suppress signal from long T2 tissues. Two common methods used to suppress long T2 tissues are RF pulse preparation and image subtraction. In this work, we investigated imaging of short T2 species using a R2* map, which creates image contrast between long and short T2 species different to that provided by current long T2 suppression methods. We demonstrated our methods with both Cartesian and radial acquisition.

                  2771.     Application of 3Dcones Sequence for Ultra-Short Echo Time Imaging

Jian-Xiong Wang1, Brian K. Rutt2, Jeffrey A. Stainsby3

1Applied Science Laboratory, GE HEALTHCARE, London, Ontario, Canada; 2Imaging Research Laboratories, Robarts Research Institute, London, Ontario, Canada; 3Applied Science Laboratory, GE HEALTHCARE, Toronto, Ontario, Canada

3D Cones sequence utilizes a central-out k-space trajectory that does not require field gradients prior to the sampling. As well, the ultra-short TE (<0.1ms) capability of the 3D Cones provides an approach for measurement of ultra-short T2 images. This work presents a preliminary application work of 3D Cones sequence in human brain short T2 imaging.

                  2772.     MP-SWIFT with Adiabatic Inversion Preparation for Quiet, B1 Insensitive T1 Weighted Imaging.

Curtis Andrew Corum1, Djaudat Idiyatullin1, Steen Moeller1, Michael Garwood1

1CMRR, University of Minnesota, Minneapolis, MN, USA

In this work we add a magnetization preparation (MP) segment interleaved

                  2773.     Enhancing Lipid Signals with SWIFT

Lauri Juhani Lehto1,2, Curt Corum2, Djaudat Idiyatullin2, Michael Garwood2, Olli Gröhn1

1A. I. Virtanen Institute for Molecular Sciences, University of Kuopio, Kuopio, Pohjois-Savo, Finland; 2Center for Magnetic Resonance Research, Minneapolis, MN, USA

We show in a proof of principle experiment that the imaging sequence SWIFT is able to enhance lipid signals (having very short T1 and T2) due to its extremely short dead time. Images of a lipid phantom taken with SWIFT and FLASH are compared qualitatively and quantitatively. With increasing flip angle the lipid signals imaged with SWIFT increase considerably, whereas for FLASH the difference is small. This shows promise that SWIFT could image short T2 signals from low concentration lipid signals in pathological cells.

                  2774.     Estimation of Attenuation Maps from UTE Derived R2 Images

Vincent Keereman1, Yves De Deene1,2, Tom Broux3, Ignace Lemahieu1, Stefaan Vandenberghe1

1MEDISIP, Ghent University - IBBT - IBiTech, Ghent, Belgium; 2Laboratory for Quantitative Nuclear Magnetic Resonance in Medicine and Biology, ECNURAD, Ghent University, Ghent, Belgium; 3Radiology/BEFY, UZ Brussel, Brussels, Belgium

The very low signal intensity of cortical bone in conventional MRI sequences makes it difficult to distinguish this tissue type from air and derive a PET attenuation map from MR images. We investigate the use of a quantitative parameter (R2) that correlates well with the density images obtained with CT. This parameter is derived from UTE images acquired at different echo times. We demonstrate the feasibility of using the R2-map derived from MR UTE images for making the distinction between bone and soft tissue. Segmentation of human MR images into bone, soft tissue and air is also shown.

Fat Water Imaging
Exhibit Hall 2-3                    Thursday 13:30-15:30

                  2775.     Multi-Echo IDEAL Cardiac Water-Fat Imaging

Karl Kristopher Vigen1, Christopher J. Francois1, Huanzhou Yu2, Ann Shimakawa2, Jean H. Brittain3, Scott B. Reeder1

1Radiology, University of Wisconsin-Madison, Madison, WI, USA; 2Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 3Applied Science Laboratory, GE Healthcare, Madison, WI, USA

Separation of water and fat signals is of great importance in cardiac imaging, both to improve visualization of pathology through fat suppression and for direct visualization of fat-containing pathology. Separation of fat and water was achieved using a cardiac gated segmented k-space multi-echo chemical shift based water fat separation method (IDEAL). The use of a multi-echo sequence and the ability to add an inversion recovery pulse and/or T2-preparation allows rapid acquisition of separated fat and water images for important clinical applications, including morphological imaging, delayed enhancement imaging, and black-blood imaging.

                  2776.     Spatially Variable Fat-Water Separation Using Short 2DRF Pulses for Fast Imaging

Jing Yuan1, Bruno Madore1, Lawrence P. Panych1

1Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA

Conventional spatial-spectral pulses can only produce uniform fat-water contrast in the whole FOV. Their usefulness may be limited by the relatively long duration, especially for short TR sequences like SSFP. In this study, we demonstrate that more flexible spatially varying fat-water contrast can be achieved within the FOV with a novel-designed shorter 2DRF pulse. Up to 3.7X reduction factors in pulse duration has be achieved. The proposed approach may also lend itself to reduced-FOV imaging and accelerated acquisitions. The proposed shorter pulses should benefit applications in MR-guided therapy where ROIs for separation may be pre-specified.

Artifacts & Correction
Exhibit Hall 2-3                    Monday 14:00-16:00

                  2777.     Coil Combination Method for Multiple-Echo Sequences and PSF Mapping

Kun Zhou1,2, Shanglian Bao1, Juergen Hennig2, Maxim Zaitsev2

1Beijing Key Lab of Medical Physics and Engineering, Peking University, Beijing, China; 2Department of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

In this work a new method for coil combination is proposed, which is particularly suited for multi-echo, multi-reference or spectroscopic acquisitions, which intrinsically contain information on the temporal phase evolution. In this method, receiver coil phase can be estimated for each pixel and coil. Thereafter, data from individual coils are divided by the resulting phase correction maps and can be added to generate a composite complex-valued data set. The method has been tested using point spread function mapping data and it can be extended to other multi-echo MRI sequences and spectroscopic applications.

                  2778.     Iterative Tuning of System Delay and Phase Correction for Echo-Planar Imaging

Holger Eggers1

1Philips Research Europe, Hamburg, Germany

Different phase corrections are commonly applied in echo-planar imaging to reduce ghosting artifacts arising from differences in the sample positions for echoes acquired with positive and negative readout gradient polarity. In this work, it is demonstrated that especially the nonlinear phase corrections are susceptible to a badly adjusted system delay. A new, iterative procedure for a joint tuning of the system delay and the phase correction is proposed to improve tolerance in this respect and is shown to substantially reduce ghosting artifacts in adverse conditions.

                  2779.     Self-Correction of Image Distortion in High-Field EPI Using EPI Phase

Kathryn E. Hammond1,2, Duan Xu1, Doug AC Kelley3, Sarah J. Nelson1,2

1Radiology and Biomedical Imaging, University of California, San Francisco (UCSF), CA, USA; 2UCSF/UCB Joint Graduate Group in Bioengineering; 3GE Global Healthcare

At high fields echo-planar imaging (EPI) becomes increasingly distorted due to imperfections in the magnetic field. This study implemented a technique for using the phase of the EPI to self-correct image distortion and demonstrated its application to phantom and in-vivo human data acquired at 3T and 7T.

                  2780.     Fieldmap-Free Retrospective Registration and Distortion Correction for EPI-Based Diffusion-Weighted Imaging

Clare Poynton1, Mark Jenkinson2, Carlo Pierpaoli3, William Wells III4

1MIT, Boston, MA, USA; 2Oxford; 3NIH; 4Harvard, BWH

We evaluate a recent method for EPI B0 distortion correction in a test framework that was previously developed for evaluating B0 distortion correction methods for DWI. The correction method uses structural MRI, segmentation, registration, and physics-based estimation of a fieldmap, and accounts for unknown shim parameters. The predicted fieldmap is subsequently used to unwarp the EPI. The test framework uses a specially collected set of DWI data and associated methodology that analyzes the variability in FA and trace following correction. The distortion correction method produces results that agree well with current fieldmap-based methodology.

                  2781.     Noise Considerations in Slice Encoding for Metal Artifact Correction

Wenmiao Lu1, Kim Butts Pauly2, Garry Evan Gold2, John Mark Pauly3, Brian Andrew Hargreaves2

1Electrical & Electronic Engr., Nanyang Technological University, Singapore, Singapore; 2Radiology, Stanford University, Stanford, CA, USA; 3Electrical Engineering, Stanford University, Stanford, CA, USA

Slice encoding for metal artifact correction (SEMAC) is recently proposed to obtain artifact-free MR images near metallic implants. SEMAC corrects through-plane distortions by summing the resolved signals from multiple slices. This can lead to the degradation in SNR without careful noise considerations. This work describes two approaches to avoid the SNR degradation, namely eliminating the relative phases between the slices and excluding the slices containing only background noise. Maintaining high SNR of the correction results is particularly important for incorporating SEMAC with various acceleration techniques.

                  2782.     Phase Angle Tilting (PAT) for Distortion Correction Caused by Susceptibility in EPI

Lirong Yan1, Deheng Weng1,2, Rong Xue1, Jiongjiong Wang3, Yan Zhuo1

1State Key Laboratory of Brain and Cognitive Science, Institute of Biophysics, CAS, Beijing, China; 2Siemens Mindit Magnetic Resonance Ltd., Shenzhen, China; 3Radiology, University of Pennsylvania, Philadelphia, PA, USA

A modified EPI sequence using Phase Angle Tilting (PAT) to correct distortion caused by susceptibility effects was introduced. Using this method, the local distortions were well corrected on both phantom and human. Compared with other correction methods, our method does not cost additional scanning time, and can easily be implemented in EPI sequences.

                  2783.     Parallel MRI Near Metallic Implants

Weitian Chen1, Philip Beatty1, Kevin Koch2, Anja Brau1

1Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 2Applied Science Laboratory, GE Healthcare, Waukesha, WI, USA

MRI has potential advantages for diagnostic imaging near metallic implants compared to other modalities, but can suffer from severe artifacts caused by implant-induced inhomogeneities. MAVRIC is one of the recently developed MRI techniques for this application. A main challenge to MAVRIC, however, is prolonged scan time. In this work, we investigated the performance of parallel imaging methods in the presence of significant off-resonance and signal distortion caused by metal implants. We demonstrate that parallel imaging can effectively be combined with MAVRIC for imaging near metallic implants in reduced scan times.

                  2784.     A Novel Algorithm for the Correction of Eddy Current Distortion in Diffusion Weighted Imaging Data

Wei Liu1,2, Guang Yang1, Zhenyu Zhou2, Yongdi Zhou1, Gengying Li1, Bradley S. Peterson2, Dongrong Xu1,2

1Physics, FMRI Unit, East China Normal University, Shanghai, China; 2Psychiatry, MRI Unit, Columbia University, New York, USA

A novel approach is proposed for the correction of eddy-current distortion in diffusion-weighted imaging (DWI) data acquired using an echo-planar imaging sequence. Our method offers improvement over the commonly used Iterative Cross-Correlation (ICC) algorithm, in that our method excludes computational biases caused by variations of signal contrast in the cerebral ventricles, and it incorporates a mutual information-based coregistration procedure into ICC. Moreover, our method does not need to collect an additional imaging dataset, nor does it work well only with DWI data that are acquired at low b-values, which are the major disadvantages of conventional ICC. The effectiveness of the proposed method is demonstrated in two experiments using both simulated and real-world DWI data.

                  2785.     Correcting Phase Errors from B0 Eddy Currents in Non-Cartesian Imaging

Ethan K. Brodsky1,2, Jessica L. Klaers2, Walter F. Block2,3

1Radiology, University of Wisconsin, Madison, WI, USA; 2Medical Physics, University of Wisconsin, Madison, WI, USA; 3Biomedical Engineering, University of Wisconsin, Madison, WI, USA

Overall consistency in image quality from non-Cartesian sequences is typically still not equivalent to that possible with Cartesian methods. One cause for this is phase errors from B0 eddy currents. While scanners actively maintain a stable field, this compensation is limited to longer time constants, so variation is still seen during gradient activity. We enhance our per-scan gradient calibration to additionally measure phase errors due to B0 eddy currents and measure and correct the phase errors for several multi-echo 3DPR acquisition. Phase variations of up to 45° are seen across a readout, but the resulting effects are typically negligible.

                  2786.     Quantifying Changes in Geometric Distortion Across a Gradient Coil Replacement

Jonathan Samuel Jackson1, Daniel Tozer1, Mark Symms2, Claudia A. M. Wheeler-Kingshott1

1Department of Neuroinflammation, UCL Institute of Neurology, London, UK; 2Department of Clinical and Experimental Epilepsy, UCL Institute of Neurology, London, UK

To quantify the potential impact of a scheduled gradient coil replacement on longitudinal quantitative neuro-imaging studies, a 3D phantom with high spatial definition was scanned before and after the event. The change in geometric distortion was shown to be less than 0.3% over 98.5% of a typical brain volume and the impact on gray and white matter volumes less than 0.2%. This generic method uses a detailed 3D phantom and off-the-shelf non-linear registration.

                  2787.     A Generic Method of Quantifying Geometric Distortion Using Non-Linear Registration and a 3D Phantom

Jonathan Samuel Jackson1, Daniel Tozer1, Jerome M. Goffin2, John S. Thornton3, Paul S. Tofts4, Claudia A. M. Wheeler-Kingshott1, Mark Symms5

1Department of Neuroinflammation, UCL Institute of Neurology, London, UK; 2University of Edinburgh, Edinburgh, UK; 3National Hospital for Neurology and Neurosurgery, London, UK; 4Brighton and Sussex Medical School, Brighton, UK; 5Department of Clinical and Experimental Epilepsy, UCL Institute of Neurology, London, UK

Measuring geometric distortion due to gradient non-uniformity has previously required image processing tools specific to the phantom design. Here we demonstrate geometric distortion measurement achieving a precision of 0.3mm over 95% of the useful volume (6.2 litres) of a phantom using off-the-shelf non-linear registration. This method can be used with any sufficiently detailed phantom. Geometric distortion is measured in a short-bore interventional MRI system, finding residual errors of 2mm at a distance of 9cm from the isocenter after the built-in gradient non-uniformity correction.

                  2788.     Sample-Specific Passive Shimming Using Optimization Algorithm

Sejung Yang1, Hahnsung Kim2, Sang-Young Zho2, Byung-Uk Lee1, Dong-Hyun Kim2,3

1Electronics Engineering, Ewha W. University, Seoul, Korea; 2School of Electrical & Electronic Engineering, Yonsei University, Seoul, Korea; 3Radiology, Yonsei University, Seoul, Korea

Shimming using passive shims can enhance the homogeneity. The position of the passive shims have been typically placed at fixed regions or manually adjusted. Further improvements are possible if the positions can be optimized for individual subjects. We present a sample-specific passive shimming method. A shim structure capable of adjusting the position of the passive shims is built and optimal shim positions are computed using an optimization algorithm. Compared to previous in-vivo passive shim approaches, the proposed method avoids trial and error when finding the best shim position and can be a robust method for improving B0 homogeneity in vivo.

                  2789.     Dynamic Magnetic Field Corrections Improve Phase-Only FMRI Activations

Andrew David Hahn1, Andrew S. Nencka1, Daniel B. Rowe1

1Biophysics, Medical College of Wisconsin, Milwaukee, WI, USA

The contribution of various “not-of-interest” global physiologic phenomena to phase time series in MRI is well documented and poses experimental challenges for complex data analysis techniques. Recently published results demonstrated that correction of global temporal variations in the main magnetic field, which directly impact phase, significantly reduces temporal phase variability, ultimately restoring statistical power of local phenomena in complex-valued functional MRI (fMRI) regression analysis. When this correction is applied in conjunction with an fMRI analysis of phase-only activation, local task related phase changes that are statistically invisible in the raw signal, become detectable with high levels of significance.

                  2790.     Application of K-Space Energy Spectrum Analysis for Inherent and Dynamic B0 Mapping and Deblurring in Spiral Imaging

Trong-Kha Truong1, Nan-kuei Chen1, Allen W. Song1

1Brain Imaging and Analysis Center, Duke University, Durham, NC, USA

Spiral imaging is vulnerable to susceptibility-induced B0 inhomogeneities as well as temporal B0 variations due to subject motion, physiological noise, and system instabilities, resulting in image blurring. Here, we further develop and combine 1) a novel method termed k-space energy spectrum analysis, which can inherently and dynamically generate a full-resolution B0 map from the k-space data at each time point without requiring any additional data acquisition, and 2) a multi-channel modulation deblurring method, to effectively and efficiently correct for these artifacts and achieve a high spatial fidelity and temporal stability in spiral imaging.

                  2791.     B0 Field Monitoring by Air-Matched Phantoms

Frederik Testud1, Maxim Zaitsev1

1Dept. of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

Fast magnetic resonance imaging sequences such as GRE EPI are sensitive to spatial and temporal changes of B0 arising from susceptibility effects, temperature changes of the shims and motion of the object. This leads to time variant common image artefacts such as distortions and intensity losses. A field map is typically acquired at the beginning of the scan whereas neither temporal nor spatial changes of B0 are monitored during the experiment. As a possible remedy, we propose to use phantoms susceptibility-matched to air which are used to monitor the B0 evolution and determine the changes of the shim coefficients.

                  2792.     On Dephasing Effects in Complex Projection Data: Implications for Rapid B0 Estimation

Daniel Nicolas Splitthoff1, Maxim Zaitsev1

1Dept. of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, BW, Germany

Often complex projection data is used for fast assessment of B0 field inhomogeneities. We here present considerations concerning dephasing effects to be taken into account when dealing with such complex projection data, and present a possible solution for the example of an Echo Planar Imaging sequence, comparing field inhomogeneities detected by projections with values from field maps.

                  2793.     The Impact of Finite RF Excitation on Steady State Free Precession

Oliver Bieri1, Klaus Scheffler1

1Division of Radiological Physics, University of Basel Hospital, Basel, Switzerland

For quantitative imaging a detailed understanding of the pulse sequence’s signal behavior is of fundamental importance. Although SSFP signal theory is based on the concept of instantaneous radio-frequency (RF) rotation, and thus being in contradiction to any real sequence implementation, SSFP theory is generally well-accepted and unquestioned. To clear things up, SSFP signal behavior for finite RF pulses is analyzed.

                  2794.     Evaluation of a Template-Based B1 Field Correction Approach for 3T MRI Brain Images

Marcelo Adrian Castro1, Jianhua Yao1, Christabel Lee1, Yuxi Pang1,2, Eva Baker1, John Butman1, David Thomasson1

1Radiology and Imaging Sciences, National Institutes of Health, Bethesda, MD, USA; 2Philips Healthcare, Cleveland, OH, USA

A method to correct B1 field inhomogeneity in brain MRI images from a template B1 map is presented and applied to five volunteers. T1 maps are computed from T1-weighted images without B1 correction, with the proper B1 correction (reference T1 map), and using B1 maps from other subjects. The quality of a given correction is characterized by the percentage of voxels having a difference less than 10% with respect to the reference map. The T1 maps significantly improved when compared to the non-corrected ones. For each case the best correction was comparable to the reference T1 map.

                  2795.     Inherent Insensitivity to B1 Field Inhomogeneity Using Regularized Nonlinear Inversion Reconstruction

Aiming Lu1, Keith R. Thulborn1

1Center for MR Research, Univ. of Illinois at Chicago, Chicago, IL, USA

Signal intensity variation that arises from inhomogeneous transmit and receive fields not only complicates the qualitative appearance of MR images but also presents the major source of error in quantitative MRI/S. To date no approach can correct for both fields in vivo effectively. An iterative nonlinear image reconstruction method was proposed recently for parallel imaging, which can jointly reconstruct the images and estimate the coil sensitivity. Here we show that this algorithm may inherently compensate for both field effects, therefore it may play a vital role in MRI/S beyond its application for parallel imaging.

                  2796.     Optimization and Group Comparison of RF (B1) Mapping Methods at 3T

Antoine Lutti1, Chloe Hutton1, Gunther Helms2, Jurgen Finsterbusch3, Nikolaus Weiskopf1

1Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, London, UK; 2MR-Research in Neurology and Psychiatry, Faculty of Medicine, University of Göttingen, Göttingen, Germany; 3Department of Systems Neuroscience, University Medical Center Hamburg–Eppendorf, Hamburg, Germany

Inhomogeneities in the RF field (B1) lead to bias in quantitative imaging. We optimize three B1 mapping methods (dual-TR FLASH/AFI, SE/STE 3D EPI, 2D STEAM) in order to reduce the level of artefacts arising from inhomogeneities of the main static field (B0) and physiology in the resulting B1 maps. Our group study shows that flip angle maps obtained using the AFI and EPI methods are within 5% agreement and their stability is ~2% over repetitions across subjects. These optimized methods are therefore ideally suited for correction of B1 inhomogeneities in quantitative imaging at 3T.

                  2797.     Optimization of 3D EPI Technique for Radio Frequency (B1) Field Mapping at 3T

Antoine Lutti1, Chloe Hutton1, Nikolaus Weiskopf1

1Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, London, UK

Spatial inhomogeneities in the radio-frequency field (B1) pose various problems such as spurious signal and contrast changes in MRI, leading to bias in quantitative imaging and difficulties in tissue segmentation. We optimize a 3D EPI B1 mapping method using carefully chosen RF pulses and a post-processing procedure that corrects biased flip angles and image distortions along the phase direction. We obtain reliable flip angle maps at 3T over the entire brain volume with a sufficient resolution and in an achievable experimental time.

                  2798.     UTE Excitation Pulses Followed by Spin Lock to Preserve Magnetization

Michael Carl1, Mark Bydder2, Atsushi Takahashi1, Eric Han1, Graeme Bydder2

1GE Healthcare, Applied Science Lab, Milwaukee, WI, USA; 2Radiology, University of California, San Diego, CA, USA

UTE excitation usually operates near the maximum gradient slew-rates and can cause transient short-term eddy currents and gradient amplifier non-linearities that can last from tens to hundreds of microseconds and thus may produce image artifacts. We propose a sequence in which the original UTE excitation pulse is immediately followed by a spin lock pulse, applied 90° out of phase with the excitation pulse to allow the transients to decay away. We have investigated this approach by studying the results of Bloch simulations.

                  2799.     Characterizing and Correcting Gradient Errors in Non-Cartesian Imaging: Are Gradient Errors Linear Time-Invariant?

Ethan Brodsky1,2, Walter F. Block2,3

1Radiology, University of Wisconsin - Madison, Madison, WI, USA; 2Medical Physics, University of Wisconsin, Madison, WI, USA; 3Biomedical Engineering, University of Wisconsin, Madison, WI, USA

Non-Cartesian sequences are sensitive to gradient delays and eddy currents, imperfections that vary between scanners and over time. Uncorrected trajectory errors lead to misregistration of k-space data and reductions in image quality. We apply a per-scan rapid gradient calibration and correction to multi-echo 3DPR acquisitions. Trajectories for oblique projections are synthesized by linearly combining measurements made on the three orthogonal gradient axes. This rapid calibration depends on the assumption that gradient errors are linear (with respect to gradient amplitude) and time-invariant (within a scan). We show that the assumption of linearity is reasonable, but that gradient errors can change rapidly over time, highlighting the importance of per-scan calibration.

                  2800.     k-Space Water-Fat Decomposition with Chemical Shift Correction and T2* Decay Compensation for Ultrashort TE Imaging

Kang Wang1, Jiang Du2, Huanzhou Yu3, Jean H. Brittain4, Scott B. Reeder1,5

1Medical Physics, University of Wisconsin-Madison, Madison, WI, USA; 2Radiology, University of California, San Diego, San Diego, CA, USA; 3Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 4Applied Science Laboratory, GE Healthcare, Madison, WI, USA; 5Radiology, University of Wisconsin-Madison, Madison, WI, USA

Tissues with very short T2*, such as bone and tendon, can be imaged with ultrashort TE (UTE) techniques. However, like all radial methods, UTE methods can be degraded by chemical shift artifacts and T2* decay. Here we proposed a k-space decomposition method that incorporates chemical shift correction and accounts for the T2* decay between echo images. More importantly, it also compensates for the blurring artifacts caused by the intrinsic T2* decay during readout, which can be pronounced for tissue with very short T2*.

                  2801.     Computer Simulation for Evaluating MR Image Artifacts Caused by Magnetic Field Fluctuations Induced by Gradient Pulses

Yo Taniguchi1, Suguru Yokosawa1, Yoshitaka Bito1

1Hitachi, Ltd., Central Research Laboratory, Kokubunji-shi, Tokyo, Japan

A computer simulator for evaluating MR image quality is extended for easily evaluating MR image quality in the presence of temporal fluctuations of the magnetic field caused by gradient switching. The temporal fluctuations are defined by feature parameters, which are inputs to the simulator. The feature parameters of the fluctuations are spatial distributions of amplitude, decay time constant, frequency, and phase. We confirmed that geometrical distortions in the images obtained by DWEPI (diffusion weighted echo-planar imaging) simulations are correspond well with those from the experiments with intentionally increased eddy currents.

                  2802.     Reduction of FSE Cusp Artifact Using Slice Tilting

Novena Rangwala1,2, Xiaohong Joe Zhou1,3

1Center for Magnetic Resonance Research, University of Illinois Medical Center, Chicago, IL, USA; 2Dept. of Bioengineering, University of Illinois at Chicago, Chicago, IL, USA; 3Depts. of Radiology, Neurosurgery and Bioengineering, University of Illinois Medical Center, Chicago, IL, USA

In fast spin echo (FSE) imaging, a “feather-like” artifact, termed as cusp artifact, is commonly seen along the phase-encoding direction in sagittal and coronal images. This artifact occurs due to specific combinations of B0-field, B1-field, and gradient non-linearity at regions beyond the field of view. To reduce this artifact, an FSE pulse sequence was modified to slightly tilt the slice selected by the RF excitation pulse. This modification, followed by simple post-processing, has been shown to noticeably decrease the cusp artifact by more than 80% in both phantom and human subjects.

                  2803.     A Method to Measure Gradient Delay Due to Gradient Coupling

Ajit Devaraj1, James G. Pipe1

1Keller Center for Imaging Innovation, Barrow Neurological Institute, Phoenix, AZ, USA

Gradient delays can potentially be considerably influenced by gradient interactions. This work presents a simple approach to measure changes in gradient delays due to gradient coupling for spiral waveforms.

                  2804.     Navigator-Stabilized GRASE Imaging for High Magnetic Fields

Daniel Gembris1, Anita Siebert1, Franzisek Hennel1

1Bruker BioSpin MRI GmbH, Ettlingen, Germany

The GRASE imaging technique allows a flexible combination of gradient- and spin echoes to gather a maximum of spatial information within a single T2 decay. A source of ghosting in this technique is the combined modulation of the k-space data by the relaxation decay and resonance offsets. The original interleaved scheme of Feinberg and Oshio that converts resonance offsets to minimal pixel shifts is the most promising at high fields. We demonstrate that the ghosting due to the T2 decay (the main drawback of interleaved GRASE) can be alleviated using navigator signals.

                  2805.     k-Space Trajectory Mapping for Ultra-Short, Single-Shot, Non-Cartesian Imaging

Florian Wiesinger1, Pekka Tapani Sipilae1,2, Yi-Fen Yen3, Dirk Mayer4, Eric Fiveland5, Sebastian A. Greding1, Daniel M. Spielman4, Adolf Pfefferbaum4, Rolf F. Schulte1

1Imaging Technologies, GE Global Research, Munich, Germany; 2Institute for Physics of Electrotechnology, Technical University Munich, Munich, Germany; 3Global Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 4Department of Radiology, Stanford University School of Medicine, Stanford, CA, USA; 5MRI Laboratory, GE Global Research, Niskayuna, NY, USA

In-vivo MR image encoding speed is physiologically limited by gradient-induced peripheral nerve stimulations. This makes high performance gradient insert coils an attractive choice for performing small-animal MR imaging and microscopy studies in a whole-body, clinical MR scanner. Such enhanced performance also amplifies gradient imperfections due to Eddy currents, coupling effects, mechanical vibrations, etc. Recently, magnetic field sensors in the form of small NMR probes have been described as a highly accurate tool for spatiotemporal magnetic field mapping. In this work such magnetic field sensors were used in combination with a high-performance gradient insert coil for ultra-fast, single-shot, high-resolution, non-Cartesian imaging.

                  2806.     Robust Compensation of Magnetic Field Perturbations for Localized Spectroscopy Applications

caixia Fu1, GuoBin Li1

1Siemens Mindit Magnetic Resonance Ltd, ShenZhen, GuangDong, China

In this abstract, factors affecting the results of the sideband cancellation in opposite-gradient acquisition scheme are analyzed. Based on this scheme, a robust and simple compensation method is presented, in which optimized cancellation is achieved by carrying out eddy-current phase correction for positive-gradient and negative-gradient acquisitions respectively.

                  2807.     Robust Volume Segmentation Using Noise Statistics of Phase and Magnitude

Yiping P. Du1, Zhaoyang Jin2,3, Yuzheng Hu4

1Psychiatry, Radiology, University of Colorado Denver, Aurora, CO, USA; 2Institute of Information and Control, Hangzhou Dianzi University, Hangzhou, Zhejiang, China; 3Biomedical Engineering, Zhejiang University, Hangzhou, Zhejiang, China; 4Physics, Zhejiang University, Hangzhou, Zhejiang, China

A robust volume segmentation algorithm using the noise statistics of both phase and magnitude is proposed. First-order phase difference is used to calculate local phase distribution to circumvent the need for phase unwrapping. A correction scheme is presented to correct the effect of linear background phase introduced by local field gradient in regions with severe field inhomogeneity. Robust volume segmentation is obtained in 3D susceptibility weighted images acquired with a TE=16ms at 3T. By applying the segmented brain volume, veins in the peripheral regions of the brain are well depicted in the minimum-intensity projection of the 3D data.

                  2808.     Removing Phase Artifacts Using Fourier Transform Based Field Estimation Can Lead to Significant Improvement in Phase Unwrapping Based SWI Processing

Jaladhar Neelavalli1,2, Yu-Chung Norman Cheng3, Jing Jiang2, Ewart Mark Haacke3

1Biomedical Engineering, Wayne State University, Detroit, MI, USA; 2The MRI Institute for Biomedical Engineering, Detroit, MI, USA; 3Academic Radiology, Wayne State University, Detroit, MI, USA

Recently an novel method for removing background field effects from SWI phase images was proposed which was based on predicting the air-tissue geometry induced phase and removing its contribution form the collected phase. The resultant images are referred to as Geometry Dependent Artifact Corrected phase images (GDAC phase). In this abstract, results of using GDAC phase versus original phase for the conventional SWI processing steps of, (1) phase unwrapping and (2) subsequent high pass filtering, were compared and it is shown that using GDAC phase improves the results of this conventional SWI processing.

                  2809.     A Multi-Mask Multi-Seed Free Growing Field Map Estimation Algorithm for Iterative Multi-Point Water-Fat Decomposition

Chuan Huang1, Maria I. Altbach2

1Dept of Mathematics, University of Arizona, Tucson, AZ, USA; 2Dept of Radiology, University of Arizona

In MRI, fat/water separation in the presence of B0 field inhomogeneity is an important research area. In this work, we introduce a new method called Multi-mask Multi-seed Free Growing as an alternative to the Region Growing (RG) algorithm to provide the initial field map estimate for IDEAL fat water separation. Our results show that this algorithm is more robust than RG algorithm, especially when a large region of low signal intensity is present or strong field inhomogeneity occurs.

                  2810.     Challenges for MR-Based Attenuation Correction in PET Imaging of the Head

Andre Jan Willem van der Kouwe1,2, Ciprian Catana1,2, Thomas Benner1,2, Christian J. Michel3, Michael Hamm4, Sonia Nielles-Vallespin5, Stephan Kannengiesser5, Matthias Fenchel5, Bruce R. Rosen1,2, A. Gregory Sorensen1,2

1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, USA; 2Department of Radiology, Harvard Medical School, Brookline, MA, USA; 3Siemens Medical Solutions USA Inc., Knoxville, TN, USA; 4Siemens Medical Solutions USA Inc., Charlestown, MA, USA; 5Siemens Healthcare, Erlangen, Germany

Attenuation correction (AC) is essential for accurate reconstruction of PET images. With combined MR-PET scanners such as the prototype Siemens BrainPET/3T TIM Trio combination, CT and transmission source data are unavailable for AC purposes. We demonstrate that with this head imaging system, the skull and the location of the MR RF coil can be detected with radially encoded ultrashort TE imaging and that inclusion of the RF coil in the AC map is critical for precise PET image reconstruction.

                  2811.     Gradient Echo Assisted 3D Look-Locker  - A Method for Improved Volume T1-Quantification Accuracy Applied to DGEMRIC

Carl Siversson1, Carl-Johan Tiderius2, Leif Dahlberg2, Jonas Svensson1

1Department of Radiation Physics, Lund University, Malmo, Sweden; 2Department of Orthopaedics, Lund University, Malmo, Sweden

3D Look-Locker (LL) can be used for various T1 quantification applications, such as delayed Gadolinium enhanced MRI of cartilage (dGEMRIC). The LL method is however usually dependent on both the excitation- and inversion pulse performing well in order to generate reliable T1 values. By introducing an additional very quick (45 seconds) spoiled Gradient Echo sequence it is however possible to map both the B1 field and the inversion quality and compensate for any fluctuations. With this method T1 is measured very accurately, throughout all slices, both in gel phantoms and in in vivo dGEMRIC measurements.

Reconstruction Methods
Exhibit Hall 2-3                    Tuesday 13:30-15:30

                  2812.     Accelerated MR Parameter Mapping Using Compressed Sensing with Model-Based Sparsifying Transform

Mariya Doneva1, Julien Sénégas2, Peter Börnert2, Holger Eggers2, Alfred Mertins1

1University of Luebeck, Luebeck, Germany; 2Philips Research Europe, Hamburg, Germany

The estimation of MR parameters, such as the relaxation times T1, T2 and diffusion coefficients D, requires the acquisition of multiple images at different sequence parameters, which is often associated with long acquisition times. These data show a high temporal correlation, which could be described by a model facilitating accelerated image acquisition by data undersampling. In this work we show that the prior knowledge about the data could be used to define a model-based sparsity transform for improved compressed sensing reconstruction for MR parameter estimation.

                  2813.     A Combination of Nonconvex Compressed Sensing and GRAPPA (CS-GRAPPA)

Andre Fischer1,2, Nicole Seiberlich3, Martin Blaimer1, Peter Jakob1,2, Felix Breuer1, Mark Griswold3

1Research Center Magnetic Resonance Bavaria e.V., Würzburg, Germany; 2Department for Experimental Physics 5, University of Würzburg, Würzburg, Germany; 3Department of Radiology, Case Western Reserve University, Cleveland, OH, USA

An extension of the nonconvex Compressed Sensing (CS) algorithm is presented. We propose to add a GRAPPA reconstruction step to take advantage of the inherent coil sensitivity profiles of multi-coil datasets. This parallel imaging step introduces a new constraint to the CS algorithm which ensures that the reconstructed data are in accordance with the inherent coil sensitivity profiles. By adding the GRAPPA step, one can reach higher acceleration factors with the CS-GRAPPA algorithm than with each individual technique due to synergy effects of both methods. We were able to recover high quality reconstructions of the sparse dynamic differences between a temporal average and a single timeframe from only 16 interleaved radial projections out of 224 of a dynamic cardiac radial dataset. The results show a significantly improved reconstruction quality compared to nonconvex CS reconstructions without the GRAPPA step. Thus, the proposed method shows potential for all applications where high spatial and high temporal resolution is required.

                  2814.     Compressed Sensing Using Dual-Tree Complex Wavelet Transform

Yookyung Kim1, Maria I. Altbach2, Theodore P. Trouard2,3, Ali Bilgin1,2

1Electrical and Computer Engineering, University of Arizona, Tucson, AZ, USA; 2Radiology, University of Arizona, Tucson, AZ, USA; 3Biomedical Engineering, University of Arizona, Tucson, AZ, USA

Compressed Sensing (CS) holds great promise in accelerating magnetic resonance imaging (MRI). In this work, we explore the use of dual-tree complex wavelet transform (DT-CWT) as a sparsifying transform in CS MRI. Our results indicate that DT-CWT can alleviate some of the shortcomings of traditional discrete wavelet transform.

                  2815.     Compressed Sensing in Parallel Imaging: Towards Optimal Sampling

Panu Tapani Vesanen1,2, Fa-Hsuan Lin2, Risto Juhani Ilmoniemi1

1Department of Biomedical Engineering and Computational Science, Helsinki University of Technology, Espoo, Finland; 2Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan

Compressed sensing (CS) is a method to measure and reconstruct signals from highly incomplete information. CS can be applied to parallel MRI to achieve great reductions in the scanning time. The performance of CS reconstruction depends on the randomness of k-space sampling. Here we consider a true random 2D under-sampling of k-space. By simulations based on real brain MRI data using 2-12 fold imaging time reduction, we show an 8-12% reduction in the reconstruction error compared to the constricted sampling patterns suggested earlier. In practice, the method can be implemented with 3D sequences by randomizing the phase encoding gradients.

                  2816.     Region of Interest Compressed Sensing

Samir D. Sharma1, Krishna S. Nayak1

1Electrical Engineering, University of Southern California, Los Angeles, CA, USA

Compressed sensing (CS) MRI has been shown to provide satisfactory image reconstruction from fewer k-space samples than in traditional methods. CS assumes transform sparsity of the image. However, the use of this assumption may introduce image artifacts that could lead to misdiagnoses in clinical applications. In this work, we exploit the benefits of CS by only imposing the sparsity constraint outside of a region of primary clinical interest (ROI), where small errors are tolerable. With this approach, we are able to increase the ROI reconstruction quality without sacrificing the acceleration gain.

                  2817.     Compressed Sensing with Vascular Phase Contrast Acquisition

Kevin F. King1, Wei Sun2

1Applied Science Lab, GE Healthcare, Waukesha, WI, USA; 2MR Science and Technology, GE Healthcare, Waukesha, WI, USA

Vascular phase contrast imaging has recently become more popular due to the emergence of nephrogenic systemic fibrosis. Several echoes with different flow moments are processed to give flow velocity weighted images. Since the images are sparse, compressed sensing can be combined with parallel imaging to accelerate phase contrast acquisition with better results than if parallel imaging were used alone. Most of the benefit is obtained by applying compressed sensing to each echo separately. A small additional improvement results from using a sparsifying transform of the final flow image as well. This produces a jointly optimized solution for all echoes.

                  2818.     Case-PDM Optimized Random Acquisition in High Quality Compressed Sensing MR Image Reconstruction

Jun Miao1, Feng Huang2, David L. Wilson1,3

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA; 2Invivo Corporation, Gainesville, FL, USA; 3Radiology, University Hospitals of Cleveland, Cleveland, OH, USA

Case-PDM is applied to optimize acquisition trajectory of compressed sensing (CS). Based on the experimental results, it can be concluded 1) CS can generate significantly better images with the optimized trajectory than the experience-based trajectory when all other parameters are same; 2) The optimized trajectory trained with one data set can be successfully applied to other data sets of different objects acquired by different MR systems, different coil, and different pulse sequences if only the anatomical structure is similar.

                  2819.     Accelerated High-Resolution Imaging of Trabecular Bone Using Total Variation Constrained Reconstruction

Michael J. Wald1, Ganesh Adluru1, Hee Kwon Song1, Felix W. Wehrli1

1Laboratory for Structural NMR Imaging, University of Pennsylvania Medical Center, Philadelphia, PA, USA

High-resolution micro-MR imaging of trabecular bone is hampered by long scan times. Parallel imaging techniques offer a means to reduce scan times but the associated losses in SNR limit the achievable acceleration rates as structural parameters are sensitive to noise. Compressed sensing techniques take advantage of image sparsity to reconstruct alias-free images from undersampled data. Here we explore the application of Total Variation Constrained Reconstruction (TVCR) to trabecular bone imaging and compare the method in terms the derived topological parameters with a generalized partially-parallel acquisition-based reconstruction.

                  2820.     Nullspace Compressed Sensing for Accelerated Imaging

Leo J. Grady1, Jonathan R. Polimeni2

1Imaging and Visualization, Siemens Corporate Research, Princeton, NJ, USA; 2NMR Center, Radiology Department, Massachusetts General Hospital, Charlestown, MA, USA

Compressed sensing techniques have recently become very popular for image reconstruction given sparsely sampled data. However, previous methods only approximately enforce the constraint that the reconstructed image has Fourier coefficients at the sampled locations, which require manual parameter tuning and result in slower speeds for convergence. We present a fast, parallelizable method that it is capable of quickly reconstructing a sparsely sampled image that exactly satisfies the acquisition. We demonstrate that our technique is able to produce an exact reconstruction of the Shepp-Logan phantom with very sparsely sampled k-space data for realizable accelerated acquisitions using two common sampling trajectories.

                  2821.     Accelerated Free Breathing MRI with Continously Moving Table Using Compressed Sensing

Matthias Honal1, Jochen Leupold1, Ute Ludwig1

1Dept. of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

Breathing motion is an important issue in MRI with continuously moving table (CMT). In this study the compressed sensing method is applied to an axial multi-slice CMT technique to acquire artifact free snapshots of the breathing motion. For a reduction factor of R = 2 images without significant motion or reconstruction artifacts are obtained.

                  2822.     Adaptive Regularization in Compressed Sensing Using the Discrepancy Principle

Kevin F. King1, Luca Marinelli2, Christopher J. Hardy2

1Applied Science Lab, GE Healthcare, Waukesha, WI, USA; 2GE Global Research, Niskayuna, NY, USA

Compressed sensing images are usually reconstructed by finding the minimum of an objective function with two terms: one that measures the difference between the k-space data of the reconstructed image and the measured k-space data (discrepancy term), and a term that measures the L1-norm of the image in a sparsifying transform domain. A weighting factor (regularization parameter) that must be properly chosen for good image quality, balances the contribution of the two terms. A method called the discrepancy principle automatically chooses the regularization parameter based on the size of the discrepancy term and the measured noise in the k-space data.

                  2823.     Compressed Sensing with Phase Constrained Partial Radial K-Space

Ali Bilgin1,2, Yookyung Kim1, Vineeth Abraham3, Theodore P. Trouard2,4, Maria I. Altbach2

1Electrical and Computer Engineering, University of Arizona, Tucson, AZ, USA; 2Dept. of Radiology, University of Arizona, Tucson, AZ, USA; 3Optical Sciences Center, University of Arizona, Tucson, AZ, USA; 4Biomedical Engineering, University of Arizona, Tucson, AZ, USA

Compressed sensing has been successfully used with radial k-space trajectories. In this work, we illustrate that partial Fourier

                  2824.     Sequential Application of Parallel Imaging and Compressed Sensing

Philip James Beatty1, Kevin F. King2, Luca Marinelli3, Chris J. Hardy3, Michael Lustig4

1Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 2Applied Science Laboratory, GE Healthcare, Waukesha, WI, USA; 3Global Research Center, General Electric, Niskayuna, NY, USA; 4Electrical Engineering, Stanford University, Stanford, CA, USA

An alternative approach to combining parallel imaging (PI) and compressed sensing (CS) is proposed. The k-space sampling pattern is modified to allow parallel image and compressed sensing to be applied in separate sequential stages during image reconstruction. The proposed approach allows compressed sensing to be used with many proven and pre-existing parallel imaging techniques and has the potential to reduce the amount of computation required for PI+CS reconstructions.

                  2825.     Reconstruction of Undersampled MR Data Using Lp (0<p<1) Spatial Constraints

Ganesh Adluru1, Mark Rosen1, Hee Kwon Song1

1Radiology, University of Pennsylvania, Philadelphia, PA, USA

Compressed sensing reconstruction using an L1 norm constraint offers high quality reconstructions from undersampled k-space data. The method is based on exploiting the implicit sparsity of the image. More recently it has been shown mathematically and using simulated numerical experiments that faithful reconstructions can be obtained from even fewer Fourier samples when an Lp constraint, where 0<p<1, is used. Here we test the feasibility of using the Lp spatial constrained reconstruction on in-vivo Cartesian MR images. Even though the cost function to be minimized is non-convex when p<1, results demonstrate that improved reconstructions can be consistently obtained.

                  2826.     Catheter Visualization for Endovascular MR Using Compressive Sampling: Comparison Against POCS

Jerome Yerly1,2, M Louis Lauzon2,3, Henry S. Chen2,4, Richard Frayne2,3

1Electrical and Computer Engineering, University of Calgary, Calgary, AB, Canada; 2Seaman Family MR Research Centre, Foothills Medical Centre, Calgary Health Region, Calgary, AB, Canada; 3Radiology and Clinical Neurosciences, University of Calgary, Calgary, AB, Canada; 4Physics and Astronomy, University of Calgary, Calgary, AB, Canada

Passive catheter MR imaging is a promising visualization technique applicable to endovascular procedures. It provides several clinical advantages compared to the current x-ray fluoroscopy gold standard. Regrettably, conventional MR fluoroscopy offers limited refresh rates at moderate-to-high spatial resolution, so k-space undersampling is usually required. Compressive sampling (CS) and projection-onto-convex sets (POCS) are two nonlinear reconstruction algorithms that can accurately reconstruct undersampled k-space datasets. We demonstrate that passive catheter tracking using multi-cycle projection dephasers (a background suppression technique) coupled with CS generates higher quality images compared to POCS, and is fast enough to permit real-time monitoring.

                  2827.     Accelerated 2D Fourier-Velocity Encoded MRI Using Compressed Sensing

Luca Marinelli1, Kedar Khare1, Kevin F. King2, Robert Darrow1, Christopher J. Hardy1

1Global Research Center, General Electric, Niskayuna, NY, USA; 2GE Healthcare, Waukesha, WI, USA

We develop a new pulse sequence for accelerated 2D Fourier-velocity-encoded (FVE) MRI to measure blood-velocity in disturbed flows. 2D Fourier-velocity-encoded M-mode MRI provides a non-invasive probe of 2D velocity distributions that cannot be measured by other modalities such as Doppler ultrasound but can be exceedingly time consuming. Unlike conventional imaging, parallel imaging cannot be utilized to reduce number of velocity encoding steps. We apply compressed sensing to this newly developed pulse sequence to increase velocity resolution without a scan time penalty and show that an accurate reconstruction with reduction factors up to 5 is possible.

                  2828.     Auto-Calibration Method for K-T-SENSE: Using TGRAPPA to Calculate the Training Data

Paola Irene Ponce1, Martin Blaimer2, Felix A. Breuer2, Peter M. Jakob1, Mark A. Griswold3, Peter Kellman4

1Department of Experimental Physics 5, University of Würzburg, Würzburg, Bavaria, Germany; 2Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Bavaria, Germany; 3Department of Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, USA; 4Laboratory of Cardiac Energetics, National Institutes of Health, National Heart, Lung and Blood Institute, Bethesda, MD, USA

In accelerated dynamic MRI, several methods have been proposed to reconstruct missing data based on prior knowledge about the motion. The k-t SENSE method utilizes coil sensitivity variations and correlations in k-space and time to reconstruct missing data. However, k-t SENSE requires additional training data as prior information about the dynamics of the object. This work removes the requirement of an additional training data acquisition by employing additional TGRAPPA reconstructions on the undersampled dataset. TGRAPPA produces images with high spatial and temporal resolution without temporal filtering effects and is therefore fully applicable for producing high-quality training data for k-t SENSE.

                  2829.     Optimal Method for Obtaining Dynamic Phase Images from Multi-Channel EPI Data

Erik B. Beall1, Ken E. Sakaie1, Kecheng Liu2, Pallab K. Bhattacharyya1, Mark J. Lowe1

1Imaging Institute, Cleveland Clinic, Cleveland, OH, USA; 2Siemens Medical Solutions, USA Inc.

Dynamic MR phase information is important for many applications, and current multi-channel methods are not optimal. Optimal combination and phase referencing are two separable steps that can be used to improve results. The implementation of an improved method for obtaining MR phase information over time in rapid acquisitions was used to obtain phase images and the benefits are demonstrated.

                  2830.     Constrained Reconstruction of Dynamic Contrast Enhanced Tumor MR Data with Respiratory Self-Gating

Ganesh Adluru1, Wei Lin1, Mark Rosen1, Hee Kwon Song1

1Radiology, University of Pennsylvania, Philadelphia, PA, USA

Compressed sensing/constrained reconstruction methods have recently been proposed and applied to accelerate MR acquisitions. Here we apply the spatio-temporal constrained reconstruction method in conjunction with the recently proposed respiratory self-gated DCE-MRI strategy to improve the reconstructions of radially undersampled DCE tumor data. Respiratory gating effectively reduces the temporal signal variations due to breathing and leads to enhanced sparsity in the temporal gradient. Improved image quality is clearly demonstrated in undersampled lung and liver tumor DCE-MRI datasets. The approach offers a promising framework to accelerate DCE-MRI of lesions located in regions severely affected by respiratory motion.

                  2831.     Fast and Motion Robust R2* Reconstruction for Functional MRI

Valur Olafsson1, Jeffrey A. Fessler1, Douglas C. Noll2

1EECS, The University of Michigan, Ann Arbor, MI, USA; 2BME, The University of Michigan, Ann Arbor, MI, USA

Quantitative images such as R2* maps are beneficial in understanding the BOLD contrast commonly used for functional MRI. However, due to its non-linear relationship to the MR signal equation it is problematic to efficiently estimate it directly from k-space data. We propose a new fast reconstruction that uses a linear approximation facilitated by a reference image to reconstruct R2* along with field map and M0 to make it robust to motion. Results from an fMRI experiment shows increased functional sensitivity for the R2* images compared to conventional T2*-weighted images and a reconstruction time just under a minute per slice.

                  2832.     Analytical Analysis of the Accuracy of HYPR LR in CE MRA

Huimin Wu1, Alexey A. Samsonov2, Julia Velikina2, Walter F. Block3

1Medical Physics, University of Wisconsin-Madison, Madison, WI, USA; 2Radiology, University of Wisconsin-Madison, USA; 3Biomedical Engineering, University of Wisconsin-Madison, USA

Image acceleration methods have recently been of significant interest due to their capabilities for capturing dynamic events with MRI. However, few accuracy characterizations with analytical methods have been developed for these non-linear methods. An analytical prediction of the temporal and spatial accuracy for HYPR Local Reconstruction (HYPR LR) is presented with a model for CE-MRA including arterial and venous vessels and enhancing background signal. We demonstrate the theory with a digital phantom whose results agree with the prediction of the analytical expression.

                  2833.     Motion Compensated Reconstruction for Free Breathing Dynamic Constrast-Enhanced MRI

Marina Filipovic1,2, Pierre-André Vuissoz1,2, Michel Claudon1,3, Jacques Felblinger1,2

1IADI, Nancy-Université, Nancy, France; 2INSERM-U947, Nancy, France; 3CHU-Nancy, Nancy, France

The presented method generalizes GRICS, a motion compensated reconstruction algorithm, in order to apply it to contrast-enhanced MRI. The mathematical representation of the true image has been replaced by a new model taking into account possible contrast changes induced by contrast agent concentration. Tested in a two-step contrast-enhancing experiment on healthy subjects, the method both corrects for motion artefacts and performs a good detection of local contrast change. This first step opens a path to auto-registered motion compensated reconstruction for DCE-MRI and for substraction imaging.

                  2834.     Time-Resolved Imaging with Multiplicative Algebraic Reconstruction Technique (MART): An Application of HYPR Principles for Variable Density Cartesian Acquisitions

Reed F. Busse1, Angel R. Pineda2, Kang Wang3, James H. Holmes1, Jean H. Brittain1, Frank R. Korosec4

1Applied Science Laboratory, GE Healthcare, Madison, WI, USA; 2Mathematics, California State University, Fullerton, CA, USA; 3Medical Physics, University of Wisconsin, Madison, WI, USA; 4Radiology, University of Wisconsin, Madison, WI, USA

MART reconstruction is applied to time-resolved MRI with an interleaved variable k-t density Cartesian acquisition. As introduced by HYPR, a time-averaged image is used as an initial estimate, conditioning the problem. Several multiplicative update iterations enforce consistency with the data. Additionally, a median filter in the time dimension is applied to suppress artifacts not consistent with physiological assumptions. Modeling experiments, simulating instantaneous contrast arrival in vessels of various sizes, are used to compare the proposed technique to conventional view sharing. MART was found to reproduce the enhancing objects more faithfully with less blurring in the spatial and temporal dimensions.

                  2835.     Iterative Soft-Thresholding Reconstruction for Time-Of-Flight MR Angiography

Kedar Khare1, Luca Marinelli1, Christopher J. Hardy1, Kevin F. King2

1Global Research Center, General Electric, Niskayuna, NY, USA; 2GE Healthcare, Waukesha, WI, USA

The implementation of Compressed Sensing strategies for accelerated MRI applications typically involve iterative methods to minimize nonlinear functionals consisting of terms representing L1-norm, total variation, etc. Iterative approaches (e.g. conjugate gradient) involve computation of complex cost functions and the solution is also sensitive to the choice of free parameters associated with the various non-linear priors. In this paper we present an iterative soft-thresholding method with slowly decreasing thresholding parameter and apply it to reconstruction and analysis of Time-of-Flight MRA. This method is not sensitive to the choice of free parameters and the computations involve only Fourier and wavelet transforms, both of which can be implemented efficiently. This method may be used either for reducing the scan time or for improving the spatial resolution of TOF-MRA images for fixed scan time.

                  2836.     Improved Temporal Resolution in Radial K-Space Sampling Using an Hourglass Filter

Maria Magnusson1,2, Olof Dahlqvist Leinhard2,3, Patrik Brynolfsson2,4, Peter Lundberg2,3

1Department of Electrical Engineering, Linköping University, Linköping, Sweden; 2Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, Sweden; 3Department of Medical and Health sciences, Linköpings University, Linköping, Sweden; 4Radiation Sciences, Umeå Universitet, Umeå, Sweden

Using a radial sampling of k-space where consecutive profiles are separated by an angle δγ as close as possible to 180°/(golden ratio), dynamic scans with high temporal resolution can be achieved. By using N profiles where N is a prime number, all profiles are updated before any profile is revisited, ensuring a completely filled k-space before any profile is re-sampled. Using a modified hourglass-shaped sliding window filter during reconstruction, oversampling near the origin can be used to significantly increase temporal resolution without compromising image quality.

                  2837.     2D Graphical Assessment for Fast Imaging Strategies

Tzu-Cheng Chao1,2, Hsiao-Wen Chung2, William Scott Hoge3, Jing Yuan4, Bruno Madore4

1Department of Radiology, Brigham and Women's Hospital, Harvard Medical School,, Boston, MA, USA; 2Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan; 3Department of Radiology,, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA; 4Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA

As the number and complexity of partially sampled dynamic imaging methods continues to increase, a reliable method to evaluate performance would be important. In the present work, an analytical formulation to evaluate the reconstruction efficiency of given fast imaging strategies was proposed. Assessments in terms of MTF, signal to artifact ratio and noise enhancement are visualized graphically, for all combinations of temporal and spatial frequencies. Three fast imaging strategies, kt-SENSE, UNFOLD-SENSE and a combination of the two (hybrid SENSE) were tested to illustrate their respective abilities, using in-vivo datasets.

                  2838.     Real-Time Elimination of Undersampling Artifacts Using 3D Total Variation on Graphics Hardware

Florian Knoll1, Markus Unger2, Franz Ebner3, Rudolf Stollberger1

1Institute of Medical Engineering, TU Graz, Graz, Austria; 2Institute for Computer Graphics and Vision, TU Graz, Graz, Austria; 3Department of Radiology, Medical University Graz, Graz, Austria

Undersampled radial acquisition strategies have great potential to acquire CE MRA data sets with high spatial and temporal resolution. It has been shown recently that image reconstruction with a total variation constraint efficiently eliminates undersampling artifacts. We show that 3D TV optimization problems can be solved on modern graphic processing units (GPUs), with computation times that allow real-time image reconstruction. Our results show excellent removal of streaking artifacts for highly undersampled CE MRA data sets. With the GPU implementation we achieve a reconstruction time of 0.24s for the whole 3D data set.

                  2839.     Online Real-Time Reconstruction of Adaptive TSENSE with Commodity CPU / GPU Hardware

Sébastien Roujol1, Baudouin Dennis De Senneville1, Erkki Vahala2, Thomas Sangild Sørensen3, Chrit Moonen1, Mario Ries1

1UMR 5231, CNRS/Université Bordeaux 2, Laboratory for Molecular and Functional Imaging, Bordeaux, France; 2Philips Medical Systems, Vantaa, Finland; 3University of Aarhus, Department of Computer Science and Institute of Clinical Medicine, Aarhus, Denmark

A real-time reconstruction for adaptive TSENSE is presented that is optimized for MR-guidance of interventional procedures. The proposed method allows high frame-rate imaging with low image latencies, even when large coil arrays are employed and can be implemented on affordable commodity hardware.

                  2840.     Segmentation of Multiple Sclerosis Lesions Using Quantitative MRI

Dagmar R. Lang1, Veronika Ermer1, N. Jon Shah1,2, Heiko Neeb3

1Institute of Neurosciences and Biophysics, Research Centre Juelich, Juelich, Germany; 2Faculty of Medicine, Department of Neurology, RWTH Aachen University, JARA, Aachen, Germany; 3RheinAhrCampus Remagen, University of Applied Sciences Koblenz, Remagen, Germany

Quantitative MRI provides the possibility to accurately monitor the course of neurological diseases, such as multiple sclerosis. Quantitative maps of the absolute water content of the brain of MS patients exhibit an approximately 8 % higher water content in lesions compared to that of the normal appearing white matter (WM). In order to investigate different characteristics of lesions in pure WM or close to the ventricles, different automatic segmentation algorithms were developed. Based on an improved tissue segmentation approach, lesions were segmented using the combination of a quantitative water content brain atlas and typical lesion characteristics such as the quantitative MR parameter distribution or the spatial localisation.

                  2841.     Artifact-Free and Multi-Contrast SSFP Imaging Achieved in a Single Acquisition

Ming-Long Wu1, Nan-Kuei Chen1

1Brain Imaging and Analysis Center, Duke University, Durham, NC, USA

Balanced SSFP (bSSFP) is a pulse sequence with great clinical potential because of its short scan time, high SNR and lower SAR (specific absorption rate). However, due to its sensitivity to field inhomogeneities, SSFP imaging is always degraded by banding artifacts. Recently, partially-dephased SSFP (PD-SSFP) was developed to reduce banding artifact. Here we report novel echo separation and multi-scheme reconstruction methods to significantly improve PD-SSFP, enabling artifact-free, higher-SNR, and multi-contrast imaging with a single PD-SSFP acquisition.

                  2842.     Improved Image Reconstruction for Multi-Shot Multi-Coil Diffusion Imaging

Martin Uecker1, Alexander Karaus1, Jens Frahm1

1Biomedizinische NMR Forschungs GmbH, Max-Planck-Institut fuer biophysikalische Chemie, Goettingen, Germany

In multi-shot diffusion imaging each segment is affected differently by phase variations caused by brain pulsation during the diffusion gradients. For this reason image reconstruction requires phase maps for each segment. Since the resolution of the maps obtained with navigator based approaches is limited by its size, images reconstructed from strongly disturbed data suffer from residual phase cancellation. Reconstructing each segment independently with parallel imaging allows to calculate phase maps with high resolution avoiding phase cancellation problems during the reconstruction. The proposed technique was applied to data from a segmented STEAM sequence where large 2D-navigators are not feasible.

                  2843.     Tradeoffs in Supersampling of DTI Metrics

Jadrian Miles1, Ron A. Cohen2, David H. Laidlaw1

1Dept. of Computer Science, Brown University, Providence, RI, USA; 2Dept. of Psychiatry, Brown Medical School, Providence, RI, USA

Values derived from diffusion tensors (such as FA and eigenvectors) are often needed at higher spatial resolution than provided by a DW-MRI scan. One may magnify, or "supersample", the DWIs, DTI, and fields of derived values in turn, but what are the tradeoffs? In this work we examine the 15 different combinations of power-of-two supersamplings that achieve 16x magnification of the derived value fields. We find that interpolating derived values is as good as interpolating diffusion tensors, provided the DWIs have been sufficiently supersampled.

                  2844.     Fast Concomitant Gradient Field Correction for Spiral Scans

Joseph Yitan Cheng1, Juan M. Santos1, John M. Pauly1

1Department of Electrical Engineering, Stanford University, Stanford, CA, USA

For non-Cartesian scans, the off-resonance distortion greatly degrades the image quality. In this work, two post-gridding correction algorithms are presented: one performed in the frequency domain and the other in the image domain. Both of these computationally fast methods yield noticeable improvements. However, the image domain approach is preferred due to its flexibility in increasing accuracy with little cost. These phase correction algorithms can be applied prior to the correction for B0 inhomogeneities and gradient nonlinearities with accurate results.

                  2845.     Improved Gradient Field Distortion Correction in Continously Moving Table Acquisitions Using GIRAFFE

Rita Gouveia Nunes1, Joseph V. Hajnal1, David J. Larkman1

1Robert Steiner MRI Unit, Imaging Sciences Department, MRC Clinical Sciences Centre, Hammersmith Hospital, Imperial College London, London, UK

Continuously-Moving table allows efficient acquisition of whole-body images. Unfortunately, B0, gradient and B1 field imperfections result in data inconsistency leading to artifacts. Commonly, to reduce these, smaller excitation volumes are used and linear k-space traversal imposed, limiting the range of contrasts attainable. Polzin et al. developed a reconstruction method which incorporates distortion correction due to gradient non-linearities. We show that although this approach is reasonably accurate for linear schemes, its performance is poorer for segmented acquisitions. Improved reconstructions can be obtained for all schemes using the Generalised Image Reconstruction Accounting For Field Effects method, increasing design flexibility and scanning efficiency.

                  2846.     Water-Fat Decomposition with MR Data Based Regularized Estimation in MRI

Won Seok Huh1, Jeffrey A. Fessler1

1EE:system, University of Michigan, Ann Arbor, MI, USA

The ability of MRI to separate chemical components is important in clinical use. We describe a new regularized iterative algorithm, termed k-space method, which reconstructs water and fat and field map. The k-space method has been implemented with MR data to consider different data acquisition time. In the method, the cost function includes a regularized term due to prior knowledge that off-resonance is usually smooth. We use iterative approach with field map estimation, and water and fat estimation. By providing readout information to the cost function, the method yields improved estimates of the chemical components.

                  2847.     Independent Estimation of T2* for Water and Fat for Improved Accuracy of Fat Quantification

Venkata Veerendranadh Chebrolu1, Catherine D G Hines1, Huanzhou Yu2, Angel R. Pineda3, Ann Shimakawa2, Charles McKenzie4, Jean H. Brittain5, Scott B. Reeder1,6

1Biomedical Engineering, University of Wisconsin Madison, Madison, WI, USA; 2Applied Science Laboratory, GE Healthcare, Menlo Park, CA, USA; 3Mathematics, California State University, Fullerton, Fullerton, CA, USA; 4Medical Biophysics, The University of Western Ontario, London, London, Ontario, Canada; 5MR Global Applied Science Lab, GE Healthcare, Madison, WI, USA; 6Radiology, University of Wisconsin Madison, Madison, WI, USA

Non-invasive quantification of intracellular fatty accumulation of the liver (steatosis) is needed for early detection and grading of fatty liver disease, the leading cause of chronic liver disease. T2* values of water and fat are independently estimated using a modified Gauss-Newton method for multiple variables, relaxing the assumptions made by single exponential T2* correction methods, that assume a common value of T2* for both water and fat. Accurate quantification of fat is demonstrated with independent estimation of T2* values for water and fat, in phantom experiments.

                  2848.     Robust Multicoil Reconstruction for Phase-Sensitive Fat-Water Separation

Tolga Çukur1, Brain A. Hargreaves2, Dwight G. Nishimura1

1Electrical Engineering, Stanford University, Stanford, CA, USA; 2Radiology, Stanford University, Stanford, CA, USA

Phase-sensitive (PS) SSFP efficiently separates the fat and water resonances by exploiting the 180o phase-difference between adjacent pass-bands in balanced SSFP. This difference can be detected if all other sources introduce slow phase variations in space. However, for surface coils, the abrupt variations at the edge of the sensitivity regions can lead to fat-water misclassification in multicoil reconstructions. In this work, we propose to first obtain a reliable phase estimate by combining individual coil images. The PS-reconstruction is then performed for successful fat-water separation.

                  2849.     Water-Fat Imaging by Direct Phase Encoding with Solution Along Isophasic Lines (SAIL)

Qing-San Xiang1

1Department of Radiology, University of British Columbia, Vancouver, BC, Canada

Direct Phase Encoding (DPE) is a simple and efficient technique for water-fat imaging, allowing even pixel level analytical water-fat separation and identification. However, for pixels containing only a single chemical component, the relative phase relationship between water and fat is unavailable. As a result, region growing is often needed to extract phase information from neighboring pixels. This work describes a novel approach to address the single-peak issue. Pixels along isophasic lines should experience the same magnetic field inhomogeneity, and thus satisfy the same equations. Water and fat can then be found by DPE with Solutions Along Isophasic Lines (SAIL).

                  2850.     Phase Optimized Sensitivity Estimation Method for SENSE

Zhaolin Chen1,2, Leigh Johnston1,2, Zang-Hee Cho3, Gary Egan1,4

1Howard Florey Institute, Carlton South, VIC, Australia; 2Department of Electrical and Electronic Engineering, University of Melbourne, Melbourne, VIC, Australia; 3Neuroscience Research Institute, Incheon, Korea; 4Centre for Neuroscience, University of Melbourne, Australia

MR phase images provide additional information beyond magnitude images. It has been noted that the SENSE reconstruction using traditional square-root-sum-of-squares sensitivity estimation approach cannot provide accurate phase reconstruction for single echo although it can preserve dynamic phase updates among echoes. This paper presents a phase optimized sensitivity estimation method for the self-calibrated SENSE methods. The new method can provide improved phase contrasts in structural imaging applications. Demonstrated in both the 3T and the 7T datasets, the proposed method produces dramatically improved phase image contrasts when compared with the current sensitivity estimation method. The new method is potentially useful for both structural and dynamic phase sensitive imaging applications.

                  2851.     Accelerating Sparse Object Imaging by Simplified Skipped Phase Encoding and Edge Deghosting with Array Coil Enhancement (SPEED-ACE)

Zheng Chang1, Qing-San Xiang2,3, Fang-Fang Yin1

1Department of Radiation Oncology, Duke University, Durham, NC, USA; 2Department of Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada; 3Department of Radiology, University of British Columbia, Vancouver, BC, Canada

The parallel imaging method of Skipped Phase Encoding and Edge Deghosting with Array Coil Enhancement (SPEED-ACE) is simplified with enhanced acceleration for imaging sparse objects; the method is termed S-SPEED-ACE that partially samples k-space with multiple coils in parallel and yields a deghosted image based on a least-square-error solution. Without differential filtering and full central k-space sampling, S-SPEED-ACE uses more straightforward reconstruction and achieves further scan time reduction as compared with SPEED-ACE. As demonstrated by a simulated scan using experimental head PC-MRA data, S-SPEED-ACE can achieve 8 times acceleration using 4 coils with reasonable image quality.

                  2852.     Performance of Three Parallel MRI Reconstruction Methods in the Presence of Coil Sensitivity Map Errors

Lotfi Chaâri1,2, Philippe Ciuciu2, Amel Benazza-Benyahia3, Jean-Christophe Pesquet1

1IGM Lab. Info., Université Paris Est, Marne-la-Vallée, France; 2NeuroSpin, CEA, Saclay, France; 3URISA, SUP'COM, Tunis, Tunisia

A study of parallel MRI reconstruction techniques performance in the presence of coil sensitivity map errors, using artificially simulated parallel MRI images.

                  2853.     Boosting Image Quality by Automatic Low SNR Channel Detection and Its Removal in K-Space

Christian Ros1, Jan Sedlacik2, Najat C. Daw3, Ralf Loeffler2, Juergen R. Reichenbach1, Claudia M. Hillenbrand2

1Medical Physics Group, Institute for Diagnostic and Interventional Radiology, Friedrich Schiller University, Jena, Germany; 2Department of Radiological Sciences, St. Jude Children's Research Hospital, Memphis, TN, USA; 3Department of Oncology, St. Jude Children's Research Hospital, Memphis, TN, USA

Phased array coils with a large number of elements are used for peripheral and/or body MRI. Activated coil elements that are located far from the measurement volume detect almost no signal and cause the automated system adjustment to improperly increase the gain factor of the respective receiver channels. If such low signal channels are incorporated in the final combination of coil elements, their artificially elevated noise is added to the reconstructed image and may result in degenerated image quality. We propose a novel k-space based algorithm which identifies such channels and discards them from the reconstruction process.

                  2854.     Fast Phase Unwrapping Using Fourier Transform

Amir M. Abduljalil1, Seongjin Choi1, Michael Vincent Knopp1, Petra Schmalbrock1

1Radiology, The Ohio State University, Columbus, OH, USA

Phase image extraction and unwrapping is explored using discrete Fourier transforms. The slow varying phase changes are filtered out using a convolution algorithm to reveal the phase changes due to the anatomy. The applied method is robust and an improvement in the contrast to noise ratio is observed when the noise in the images is masked out prior to starting the calculations.

                  2855.     Coil Array Compression in Coronary MRA

Johannes Frieder Schmidt1, Martin Buehrer1, Peter Boesiger1, Sebastian Kozerke1

1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

SNR limits the spatial resolution necessary for coronary MRA. Higher SNR is provided by large coil arrays, but at the expense of increased data sizes for image reconstruction. This work demonstrates the applicability of coil array compression for coronary imaging which uses a linear combination of channels to compress the data to a reduced set of virtual coils. In 10 healthy volunteers, it is shown for various compression factors that the visible image quality remains comparable. For a compression factor of 8, the relative loss in SNR on the coronaries is 5%.

                  2856.     Hybrid Roemer Reconstruction for Multi-Element TX/RX Body Coil at 3T

Kay Nehrke1, Klaus Michael Uhr2, Peter Börnert1

1Philips Research Europe, Hamburg, Germany; 2Technische Universität, Hamburg-Harburg, Germany

For a TX/RX coil array, the transmit field inhomogeneity at higher field strengths (>3T) is an important problem. However, as a result of the principle of reciprocity, the receive fields are affected by wave propagation effects at high fields as well. Hence, appropriate combination of the data from the individual receive channels has to be performed by image reconstruction. In the present work, a hybrid Roemer reconstruction based on pre-calibrated sensitivity data is proposed. It has been implemented, and its performance has been compared with simple data combination approaches based on sum-of-squares or complex superposition.

                  2857.     Kaczmarz Iterative Reconstruction for Arbitrary Hybrid Encoding Functions

Jason Peter Stockmann1, Robert Todd Constable2

1Biomedical Engineering, Yale University, New Haven, CT, USA; 2Diagnostic Radiology & Neurosurgery, Yale School of Medicine, and Biomedical Engineering, Yale University, New Haven, CT, USA

The Kaczmarz iterative algorithm (also known as Algebraic Reconstruction Technique, or ART) is applied to the problem of reconstructing images using arbitrary combinations of coil and gradient encoding. Methods have recently been proposed to form images using higher-order gradient encoding functions such as spherical harmonics. Since these encoding functions do not project the object into k-space, we propose to solve the MR signal equation directly without Fourier Transforming the echoes. Kaczmarz is used to reconstruct simulated data produced by linear and non-linear gradients, and spatially-varying point spread functions are illustrated.

                  2858.     A Phase Constrained Graph Cut Algorithm for Reconstruction of MR Parallel Imaging Under 3D Spatial Priors

Ashish Raj1, Ramin Zabih2

1Radiology, Weill Medical College of Cornell University, New York, NY, USA; 2Computer Science, Cornell U, Ithaca, NY

Current parallel imaging methods like SENSE and GRAPPA are seriously limited by SNR and g-factors associated with coil geometry. Most MR images have sharp edges and are piecewise smooth. A new edge preserving Markov Random Field (MRF) prior model was proposed in to address this issue, and a new graph cut-based algorithm was developed to solve the computationally challenging non-convex optimization problem. Here we propose two important extensions of the EPIGRAM method by (a) incorporating 3-D spatial priors and by jointly reconstructing multi-slice images, and (b) using low-frequency phase as a constraint during the reconstruction process. Recent reports indicate that phase-constrained reconstruction improves image quality. The proposed method performs better than the conventional SENSE method applied slice-by-slice.

                  2859.     Extended Anti-Aliasing Reconstruction for Phase-Scrambled MRI with Quadratic Phase Modulation

Maxim Zaitsev1, Gerrit Schultz1, Jürgen Hennig1

1Dept. of Diagnostic Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany

As shown recently, phase-scrambled MRI is capable of a scalable anti-aliasing image reconstruction. The purpose of the present work is to extend the concept via a transparent mathematical framework capable of handling scalable alias-suppressing image reconstruction of the phase-modulated data. Our results show that translation of the Fresnel transform concept to k-space representation allows to define limits and boundary conditions for practical implementations and to understand signal behaviour and potential artefacts.

                  2860.     Generalized Reconstruction of Multi-Spectral MR Acquisitions for Imaging Near Metal Implants

Diego Hernando1, Kevin M. Koch2, Kevin F. King2, Zhi-Pei Liang1

1Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, USA; 2Applied Science Laboratory, General Electric Healthcare, Waukesha, WI, USA

Several techniques have been recently proposed for imaging near metal implants (i.e., in the presence of extreme B0 field inhomogeneity), by effectively dividing the large spectral range into smaller bins. However, conventional sum-of-squares reconstructions result in localized image intensity distortions in regions of rapid field variation. This abstract presents a novel method for generalized reconstruction of multi-spectral acquisitions, where the reconstruction is posed as a regularized least squares problem. The proposed method overcomes some of the main limitations of current techniques, and provides increased flexibility for the selection of imaging parameters.

                  2861.     Bayesian Reconstruction of DW-PROPELLER Images Using Joint Entropy

Jinhua Sheng1, Dong Liang1, Jing Tang2, Donglai Huo3, Leslie Ying1

1Dept. of Electrical Engineering and Computer Science, Univ. of Wisconsin-Milwaukee, Milwaukee, WI, USA; 2Dept. of Radiology, the Johns Hopkins University, Baltimore, MD, USA; 3Keller Center for Imaging Inno,Barrow Neurological Institute, St. Joseph Hospital and Medical Center, Phoenix, AZ, USA

Turboprop has shown to reduce the artifacts of single-shot EPI for Diffusion-weighted imaging (DWI) but at the cost of longer acquisition. In this abstract, we propose a novel Bayesian reconstruction method to reduce the number of acquisitions without compromising the SNR. The method incorporates the information from high quality non-weighted images into the DWI reconstruction, with the joint entropy between the weighted and non-weighted image features as the prior. The method is shown to improve the DW image quality with a set of DW-PROPELLER brain data.

                  2862.     Resolution and Noise Properties of Linear Phase-Constrained Partial Fourier Reconstruction

Justin P. Haldar1, Kenneth Sakaie2, Zhi-Pei Liang1

1Beckman Institute, Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, USA; 2Cleveland Clinic, Cleveland, OH, USA

Phase-constrained partial Fourier (PF) reconstruction is a classical technique that leverages prior knowledge of the image phase to reduce k-space sampling requirements. While the technique has seen wide use, the characteristics of PF reconstructions are usually only evaluated empirically. In this work, we show that resolution and noise properties of the class of linear PF reconstruction methods (including homodyne, projection onto convex sets with linear projections, and matrix inversion methods) can be characterized theoretically in terms of spatial response functions and interference response functions. We demonstrate an application of these theoretical characterizations in the context of regularized PF reconstruction.

                  2863.     Data Interpolation in Phase-Scrambling Fourier Transform Technique for Alias-Free Image Reconstruction by Iterative Signal Restoration Algorithm

Yoshifumi Yamada1, Satoshi Ito1

1Department of Information Systems Science, Utsunomiya University, Utsunomiya, Tochigi, Japan

The phase scrambling Fourier transform (PSFT) signal can be converted into the Fresnel transform signal by multiplying by a quadratic phase term. In this paper, a new alias-free image reconstruction technique in the PSFT imaging technique is presented in which the signal is converted into the Fresnel transform signal and residual alias in the image is removed by increasing the data interpolation accuracy with using iterative signal restoration algorithm. Numerical simulation using MR images showed that almost alias-free images can be reconstructed using this technique.

                  2864.     Iterative Reconstruction for Off-Resonance Effects in SWIFT Imaging

Steen Moeller1, Djaudat Idiyatullin1, Curt Corum1, Rong Zhou2, Michael Garwood1

1Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, USA; 2Departments of Radiology, University of Pennsylvania, PA, USA

: We demonstrate the feasibility for off-resonance correction with SWIFT applications where field mapping is not possible. Myocardium injected with SPIO-labeled stem cells is imaged ex-vivo, and the locally dephased signal is reconstructed.

                  2865.     Including Prior Knowledge in the Reconstruction of ASL Perfusion Images to Reduce Noise and Improve Spatial Resolution

Philip M. Robson1, David C. Alsop1

1Radiology, Beth Israel Deaconess Medical Center, Boston, MA, USA

ASL perfusion images are often acquired at low spatial resolution to attain sufficient SNR in the perfusion signal. However, ASL studies often include acquisition of high resolution anatomical reference images. ASL perfusion imaging is, therefore, a good candidate to benefit from the combined functional and anatomical information available from MRI. We explore the use of constrained image reconstructions and inclusion of prior knowledge from anatomical reference images to enhance SNR in perfusion images and improve spatial resolution.

Image Segmentation & Localization
Exhibit Hall 2-3                    Wednesday 13:30-15:30

                  2866.     Estimation of Tissue at Risk of Infarction Using a Support Vector Machine on Multimodal Stroke MRI Data

Susanne Schnell1, Irina Mader2, Dorothee Saur3, Kim Mouridsen4, Roza Umarova3, Dorothee Kümmerer3, Hans Burkhardt5, Valerij G. Kiselev1

1Dept. of Diagnostic Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany; 2Dept. of Neuroradiology, University Medical Center Freiburg, Freiburg, Germany; 3Dept. of Neurology, University Medical Center Freiburg, Freiburg, Germany; 4Dept. of Neuroradiology, Center for Functionally Integrative Neuroscience, Aarhus University Hospital, Aarhus, Denmark; 5Dept. of Computer Science, Chair in Pattern Recognition and Image Processing, Albert-Ludwigs University Freiburg, Freiburg, Germany

Acute ischemic stroke is a frequent cause of neurological disability. A clinical tool for the differentiation of reversable from irreversable damaged tissue would be of valuable use. This requires accurate prediction of the extent of the infarction. Typically, the estimation relies on models derived from perfusion and diffusion data (e.g. perfusion-diffusion mismatch, CBF, CBV). In this successfully performed trial study we introduce a data-driven method using Support Vector Machine (SVM) classification. The input data for the SVM consists of acute diffusion and perfusion information. A labelled training dataset is created using the patient’s follow-up scan in the chronic stage.

                  2867.     Direction of Plaque Progression Is a Significant Predictor in Plaque Vulnerability: An Evaluation by MRI

Dongxiang Xu1, Chun Yuan1

1Radiology, University of Washington, Seattle, WA, USA

This study explores the relationship between the direction of plaque progression and plaque vulnerability. Projection length along the axial and longitudinal directions were analyzed and correlated with subjects’ symptom status. The statistical results show that plaque progression in the axial direction is a better predictor for patient’s symptom than longitudinal progression. This means that the plaque’s morphology change, in addition to plaque size which has been widely used in atherosclerosis research, can potentially be another significant factor to carotid artery disease evaluation and patient’s symptom prediction.

                  2868.     The Optimised Projection Image Method for Targeted Tissue Contrast Enhancement.

Amanda Ching Lih Ng1,2, Nathan Faggian2,3, Zhaolin Chen2,4, Jingxin Zhang1, Gary F. Egan2,3, Leigh A. Johnston2,4

1Department of Electrical and Computer Systems Engineering, Monash University, Melbourne, Vic, Australia; 2Howard Florey Institute, University of Melbourne, Melbourne, Vic, Australia; 3Centre for Neuroscience, University of Melbourne, Melbourne, Vic, Australia; 4Department of Electrical and Electronic Engineering, University of Melbourne, Melbourne, Vic, Australia

In recent years there has been an increased interest in combining phase and magnitude MRI data. We propose an Optimised Projection Image (OPI) technique which removes heuristics involved in combining phase and magnitude to enhance contrast in MRI for a target tissue class. This method fits a bivariate Gaussian mixture model to the magnitude and phase data, and projects the data onto an intensity axis optimised for the model. This algorithm is shown to combine the detail in both magnitude and phase images to create an image with enhanced edge definition and structural detail in comparison to the original images.

                  2869.     Enhancement of Venous Vasculature in the Brain with Multi-Scale Filtering

Zhaoyang Jin1,2, Ling Xia2, Yiping Peter Du3

1Institute of Information and Control, Hangzhou Dianzi University, Hangzhou, Zhejiang, China; 2Biomedical Engineering, Zhejiang University, Hangzhou, Zhejiang, China; 3Psychiatry,Radiology, University of Colorado Denver School of Medicine, Denver, CO, USA

In susceptibility weighted imaging, off-resonance artifacts can appear in regions with severe field inhomogeneity. The use of minimum-intensity projection can also cause signal loss in the peripheral regions of the brain. In this study, we present a multi-scale filter to enhance venous vasculature using magnitude images. Venography with positive venous contrast can be obtained. The maximum-intensity projection display of the filtered venography does not suffer from signal loss. Excellent venous contrast was obtained with a relatively short TE acquisition (TE=16.3ms) at 3T, making it feasible to reduce the scan time while preserving good visibility of small veins.

                  2870.     Semi-Automated Volume Segmentation of the Brain Based on a Level Set Approach

Zhaoyang Jin1, Ling Xia2, Yiping Peter Du3

1Institute of Information and Control, Hangzhou Dianzi University, Hangzhou, Zhejiang, China; 2Biomedical Engineering, Zhejiang University, Hangzhou, Zhejiang, China; 3Psychiatry,Radiology, University of Colorado Denver School of Medicine, Denver, CO, USA

The minimum-intensity projection (mIP) display of 3D susceptibility weighted images can introduce signal loss in peripheral regions of the brain because of the low intensity in air or bone along the path of projection. In this study, we present the results of semi-automated volume segmentation using a level set approach to obtain a brain mask using magnitude images. Signal loss in the mIP display of 3D susceptibility weighted images in the peripheral regions of the brain can be effectively avoided using the brain mask.

                  2871.     Segmentation of Amyloid Plaques in MR Images of the APP Transgenic Mouse Brain Using SVM

George Iordanescu1, Palamadai Venkatasubramanian1, Alice Wyrwicz1

1Northwestern University, Evanston, IL, USA

We present an automated method for the selection of the segmentation parameters used for the quantification of amyloid plaques in APP transgenic mice that are models for Alzheimer’s disease (AD). For this we use support vector machines (SVM) in an unsupervised way. The usual approach for classification is to use the “ground truth”, in our case real examples of amyloid deposits in MR images. Obtaining such examples is however very difficult, due to the reduced size and low contrast of plaques. To address this, we employ a “learning by counter-example” approach, by training one class SVM classifier on control datasets.

                  2872.     Determination of Whole Body Fat and Visceral Adipose Tissue, Combining Three Volume Estimation Methods

Tania Buehler1, Daniel GQ Chong1, Chris Boesch1

1Dept. of Clinical Research, University Bern, Bern, Switzerland

Whole body fat (WBF) and particularly visceral adipose tissue (VAT) are of paramount importance as cardiovascular risk factors related to the metabolic syndrome. At 3 Tesla MR systems an inherent inhomogeneity of the radiofrequency field in the abdominal region jeopardizes the estimation of VAT and WBF and make it necessary to find a more robust and accurate method to measure these volumes. This study suggests and tests a method that combines the strengths of three volume determination methods (threshold, planimetry, and point counting), which makes the determination of the fragmented VAT compartment feasible and quick.

                  2873.     Establishment of a Novel Automated Approach for Quantification of Subcutaneous and Visceral Adipose Tissues: Correlates with Other Anthropometric Measurements and Cardiometabolic Risk Factors

Sarayu Annamalai Parimal1, Karren Hui Min Chen1, Suresh Anand Sadananthan2, Sam Sim1, Vitali Zagorodnov2, Michael WL Chee1

1Cognitive Neuroscience Laboratory, Duke-NUS Graduate Medical School, Singapore, Singapore; 2Computer Engineering, Nanyang Technological University, Singapore, Singapore

We developed a novel automated algorithm for estimation of visceral and subcutaneous adipose tissues. Using it on abdominal MR scans obtained from 119 healthy elderly volunteers showed that women have more subcutaneous abdominal fat while men have more visceral abdominal fat [6], which was consistent with previous studies. Visceral, and not subcutaneous abdominal fat, is associated with the various risk factors for cardiometabolic diseases [7]. We conclude that this automated method of abdominal adiposity quantification and analysis is a sensitive and valid tool to be utilized in future epidemiological studies of abdominal adiposity and cardiovascular and metabolic health.

                  2874.     Automatic Liver Segmentation of  3D Abdominal MR Images of Rodents

Way Cherng Chen1, Ying Min Wang1, Kai-hsiang Chuang1, Xavier Golay2, Anqi Qiu3

1Singapore Bioimaging Consortium, A*STAR, Singapore, Singapore; 2University College London, UK; 3National University of Singapore, Singapore

An automatic segmentation approach that integrates both a probabilistic model and an atlas-based method for utilizing both image intensity and liver shape information was developed for fast and accurate liver segmentation. An average volume difference of 20% and overlap ratio of 70.3% between the automatic segmentation and manual segmentation was obtained from 5 mice images. These variabilities are due to the limited contrast between the liver and surrounding tissues. More subjects are needed to test the robustness of the algorithm. The segmentation accuracy should be further improved by using a probabilistic atlas based on a large number of subjects.

                  2875.     Influence of a Partial Volume Effect in the Segmentation of Brain Tissue Based on Diffusion Tensor Imaging (DTI) Data: A Digital DTI Phantom Study

Seiji Kumazawa1, Takashi Yoshiura2, Hiroshi Honda2, Fukai Toyofuku1, Yoshiharu Higashida1

1Department of Health Sciences, Faculty of Medical Sciences, Kyushu University, Fukuoka, Japan; 2Department of Clinical Radiology, Kyushu University, Fukuoka, Japan

The performance of the brain segmentation based on diffusion tensor imaging (DTI) data has not been evaluated by using the digital DTI phantoms. We extended the previous DTI phantom model to the partial volume description of diffusion measurements, and investigated the performance of segmentation using the conventional method on the DTI phantom with a partial volume effect (PVE) model. The values of overlap measure between segmentation and the ground truth in the data with a PVE are lower than those without a PVE. These results indicate that a method taking a PVE into account can improve the segmentation performance on DTI data.

                  2876.     MR ONCO-TREAT:  a New Tool for Volumetric and Functional Analysis of Hepatic Tumors Monitored with Multi-Modal MRI

Mehmet Akif Gulsun1, Clifford Raabe Weiss2,3, Ralph Strecker3, Glenn Meredith1, Ihab R. Kamel2

1Imaging and Visualization, Siemens Corporate Research, Princeton, NJ, USA; 2The Russell H. Morgan Department of Radiology and Radiologic Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA; 3MR Oncology, Siemens Healthcare, Erlangen, Germany

Multi-parametric MRI is becoming the gold standard for assessing hepatic tumor response to intra-arterial therapy. We present the MR-ONCO-TREAT software package for 3D evaluation of multi-parametric MR studies. The software includes: deformable image registration, interactive segmentation, and volumetric and functional analysis. In six patients with hepatocellular carcinoma, this software was able to register pre and post intra-arterial treatment MRI scans, segment tumor from surrounding liver, and provide volumetric and functional data analysis. On average, percent arterial enhancement decreased by 35.7, and the ADC increased by 9.5%. The MR-ONCO-TREAT software can automate the evaluation of hepatic tumors, possibly decreasing interpretive errors.

                  2877.     Comparison of Brain Segmentation Results Using Automated FSL-FAST with DTI Channel Inputs

Elizabeth Zakszewski1,2, Deok Han1, Jee Eun Lee1, Vikas Singh1, Andrew L. Alexander1,2

1Waisman Laboratory for Brain Imaging and Behavior, University of Wisconsin, Madison, WI, USA; 2Dept. of Medical Physics, University of Wisconsin, Madison, WI, USA

The performance of the FSL-FAST algorithm for segmenting brain tissue from DTI data is evaluated in comparison with hand-drawn white matter segmentations. Various ways of combining different DTI parameter maps with FSL's built-in option are compared and evaluated.

                  2878.     Optimizing N3 Parameters Leads to Better Segmentation Accuracy on 3T Scanners

Weili Zheng1, Michael WL Chee2, Vitali Zagorodnov1

1Computer Engineering, Nanyang Technological University, Singapore, Singapore; 2Cognitive Neuroscience Laboratory, Duke-NUS Graduate Medical School, Singapore, Singapore

A recent study by Boyes at al. have found that performance of non-parametric nonuniformity correction approach N3 on 3T scanners could be improved by reducing the value of the parameter that controls the smoothness of the estimated bias field. The present study not only confirms this finding but also demonstrates the benefits of using smaller smoothing distances (30-50mm compared with default 200mm) on the quality of white matter surface estimation and reliability of cortical thickness and subcortical structures volumes, which is promising to increase the feasibility of longitudinal neurological studies.

                  2879.     Use of Lesion Walker Classification in Patients with Multiple Sclerosis

David Scott Wack1,2, Michael G. Dwyer, Jackie Durfee, Sara Hussein, Deepa P. Ramasamy, Niels Bergsland, Robert Zivadinov1

1 Buffalo Neuroimaging Analysis Center, Jacobs Neurological Institute, Dept. of Neurology, State University of New York at Buffalo, Buffalo, NY, USA; 2Dept. of Nuclear Medicine, State University of New York at Buffalo, Buffalo, NY, USA

Software for the analysis and serial classification of multiple sclerosis lesions was developed and tested. The system is based on the definition of a lesion complex (LC) which maintains a one to one correspondence between scans. A grammatically consistent labeling system was used to classify LCs in a chronological progression. The software forces a quality control check to ensure that the operator has not missed ROIs prior to the classification stage. The analysis time is significantly reduced for an operator; while the result has a higher credence by virtue of the forced quality control and grammatical consistency of the labels.

                  2880.     Validation of Fat Volume Quantification with IDEAL MRI

Abdullah Alabousi1, Salam Al-Attar2, Tisha R. Joy1,3, Robert A. Hegele1,3, Charles A. McKenzie3,4

1Schulich School of Medicine and Dentistry, London, Ontario, Canada; 2University of Ottawa School of Medicine, Ottawa, Ontario; 3The Robarts Research Institute, London, Ontario; 4Department of Medical Biophysics, Schulich School of Medicine and Dentistry , University of Western Ontario

We will be validating IDEAL (Iterative Decomposition of water and fat with Echo Asymmetry and Least-squares estimation), an MRI technique that allows images to be produced only from fat containing tissues. Four normal controls were recruited to validate IDEAL for fat volume measurements. IDEAL correlates strongly with our gold standard (1.5T imaging) except for visceral abdomi-nal fat measurements. The lower value of visceral fat volume given by IDEAL may reflect an improved ability to differentiate be-tween true visceral fat and other tissues, however, this will require evaluation of a larger number of subjects with a range of visceral fat mass.

                  2881.     Three-Dimensional (3-D) Segmentation of Knee Cartilage from 3.0 T MR Images Using a Graph-Cuts Technique

Hackjoon Shim1,2, Samuel Chang3, Cheng Tao1, Jinhong Wang1, Kyongtae Ty Bae1

1Department of Radiology, University of Pittsburgh, Pittsburgh, PA, USA; 2School of Electrical Engineering, Seoul National University, Seoul, Korea; 3Department of Radiology, Washington University, St. Louise, MO, USA

In this study, a 3-D semi-automated segmentation method of knee cartilages from MR images was proposed. The segmentation process is based on a graph-cuts algorithm and was evaluated with experiments on MR images from osteoarthritis initiative (OAI) database. It was significantly more efficient and reproducible than the manual delineation method.

                  2882.     Deterministic Automatic Segmentation in MRI, CT and CTA: A Robust Method Based on Anatomical Structures Modelling and Local Recursive Intensity Analysis

Manfredo Atzori1,2, Gianluca Rambaldelli1, Paolo Brambilla3, Alessandra Bertoldo2

1Medicine and Public Health, Università di Verona, Verona, Italy; 2Information Engineering, Università di Padova, Padova, Italy; 3Pathology & Exp. Clin. Medicine, Università di Udine, Udine, Italy

At the state of the art some semiautomatic and automatic segmentation procedures have been described. Current methods don’t mind geometric and physical constitution of anatomical structures, not making a physically justified segmentation. This implies 8-9% volume errors, low methods concordance and low adaptability. We realized an automatic method for MRI, CT, CTA deterministic segmentation, based on mathematical modelling and analysis of anatomical structures physical and geometrical properties. It has been applied for several structures segmentation, as brain, grey-white matter, heart, kidneys, urinary tracts and bone. It is characterized by high precision (volume error 4.5%), non subjective parameters and high adaptability.

                  2883.     Towards Localization of Transcranial Magnetic Stimulation Inside the MRI Scanner

Aranee Techawiboonwong1, Alvaro Pascual-Leone2, Daeshik Kim1, Itamar Ronen1

1Boston University School of Medicine, Boston, MA, USA; 2Beth Israel Deaconess Medical Center, Boston, MA, USA

Transcranial magnetic stimulation (TMS) has been used as a non-invasive tool in various neuroscience applications. One of the greatest challenges TMS faces is an exact coil positioning to deliver the pulse to the intended area. In this work, we propose a new method of localizing the TMS stimulation by exploiting the fact that interference of TMS field inside the MR magnet lead to intensity variation of MR images. Our phantom studies showed that origins of the TMS fields could be localized and mapping of the TMS fields can potentially lead to quantification of neurostimulations that the TMS produce.

                  2884.     Changes in in Vivo Tibiofemoral Cartilage-To-Cartilage Contact Area Under Acute Loading; Comparison of Two Sequences (3D-SPGR Vs. T2-Weighted FSE)

Choongsoo S. Shin1, Richard B. Souza1, Xiaojuan Li1, Brad Wyman2, Thomas M. Link1, C Benjamin Ma3, Sharmila Majumdar1

1Department of Radiology and Biomedical Imaging, University of California at San Francisco, San Francisco, CA, USA; 2Pfizer Inc., New London, CT, USA; 3Department of Orthopaedic Surgery, University of California at San Francisco, San Francisco, CA, USA

The purpose of this study is 1) to compare human in vivo tibiofemoral cartilage-to-cartilage contact area using 3D-SPGR vs. T2-weighted FSE images; 2) to determine the influence of acute mechanical loading on tibiofemoral cartilage-to-cartilage contact area in the medial and lateral compartment. Cartilage contact area using both sequences showed strong positive correlation (Pearson Correlation coefficient =0.969). Acute loading significantly increases tibiofemoral cartilage contact area both at the medial and lateral compartments. Absolute changes as well as relative changes in cartilage contact under acute loading were more distinctive in the medial compartment.

                  2885.     Automated Segmentation of 3D Histological Mouse Brains Using a MRI-Based 3D Digital Atlas

Jessica Lebenberg1, Anne-Sophie Hérard1, Albertine Dubois2, Marc Dhenain1, Philippe Hantraye1, Vincent Frouin3, Thierry Delzescaux1

1MIRCen, CEA, Fontenay-aux-Roses, France; 2NeuroSpin, CEA, Saclay, France; 3SCSR, CEA, Evry, France

We propose a method to register MRI-based 3D digital atlas on histological and autoradiographic mouse brain images to automatically segment those post mortem data. To co-register the atlas with these data, we first deformed the MRI volume to match with 3D-reconstructed post mortem volumes. We then applied the deformation parameters to warp the MRI-based 3D digital atlas. The reliability of this method was evaluated by qualitatively and quantitatively comparing the atlas-based with manual segmentations, considered as gold standard. Our results show that this approach is promising to investigate post mortem datasets in the mouse, and make group or strain comparisons.

                  2886.     Semi-Automatic Segmentation of Mouse Embryo MRIs

Leila Baghdadi1, John G. Sled1,2, Jurgen E. Schneider3, R Mark Henkelman1,2, Jason P. Lerch1,2

1Mouse Imaging Center, Toronto, Ontario, Canada; 2University of Toronto, Toronto, Ontario, Canada; 3University of Oxford, Department of Cardiovascular Medicine, UK

A semi-automatic technique to segment individual embryos from the 32 stacked in the MRI dataset using a deformable model algorithm.

                  2887.     Automatic Identification and Segmentation of Infarct Lesions from Diffusion Weighted MR Images and ADC Maps

Suja Saraswathy1, Saurabh Sharma2, Dattesh Shanbhag1, Uday Patil1, Rakesh Mullick1

1Imaging Technologies Lab, GE Global Research, Bangalore, Karnataka, India; 2Dept of Computer Science, IIIT,Hyderabad, Hyderabad, A.P., India

Stroke is a major cause of death and disability in both the more developed and the less developed world. The core of the lesion that is dead shows up in dwi image as hyper intense region. Previous methods for infarct segmentation from DWI images had many false positives, as they did not use the complementary information from apparent diffusion coefficient (ADC) maps available with DWI images. In the present work, we have combined the contrast information provided by DWI and physiological information provided by the apparent diffusion coefficient (ADC) maps to segment the infarct core automatically.

                  2888.     A Multi-Strategy Method for MRI Segmentation

Miguel Martin-Landrove1,2, Gabriel Padilla3, Giovanni Figueroa3, Marco Paluszny3,4, Wuilian Torres3,5

1Center for Molecular and Medical Physics, Universidad Central de Venezuela, Caracas, Venezuela; 2Centro de Diagnóstico Docente, Las Mercedes, Caracas, Venezuela; 3Center of Geometry, Universidad Central de Venezuela, Caracas, Venezuela; 4Department of Mathematics, Universidad Nacional de Colombia, Medellín, Colombia; 5Fundación Instituto de Ingeniería, FII, Caracas, Venezuela

A method for T2-weighted MRI segmentation according to tissue transversal magnetization decay rates is presented. A log-convexity filter is applied first to regularize image intensity decay and control noise, followed by curve fitting with de Prony pseudo-interpolating and Montecarlo-Vandermonde robustness filters. Decay rate distributions and decay modes are obtained for tissue classification. Linear regression analysis is used for image segmentation assuming that pixel intensity decay is a linear superposition of decay modes. The proposed approach shows accuracy and computational speed compared to Inverse Laplace Transform or Vandermonde like methods. It can be extended to other images such as diffusion-weighted MRI.

                  2889.     Identification of Functional Subunits of the Human Cortex Using Resting State FMRI

Xilin Shen1, Fuyuze S. Tokoglu1, Jennifer K. Roth1, Xenophon Papademetris1, Robert Todd Constable1

1Diagnostic Radiology, Yale University, New Haven, CT, USA

Since the first lesion studies of Broca (1861) and Wernicke (1874) scientists

                  2890.     Improved Skull Stripping Using Graph Cuts

Suresh Anand Sadananthan1, Weili Zheng1, Michael WL Chee2, Vitali Zagorodnov1

1Computer Engineering, Nanyang Technological University, Singapore, Singapore; 2Cognitive Neuroscience Laboratory, Duke-NUS Graduate Medical School, Singapore, Singapore

Many recent skull stripping approaches rely on iterative surface deformation to fit the brain boundary and tend to leave residual dura, which may cause cortical thickness overestimation. The approach proposed here is based on intensity thresholding followed by removal of narrow connections using graph theoretic image segmentation to position cuts that serve to isolate and remove dura. Relative to the Hybrid Watershed Algorithm (HWA), the novel approach achieved at least 10-30% reduction in dura mater without having to trade off for increased brain tissue erosion. However, the main value of the new approach is significant reduction in cortical thickness overestimation when combined with HWA.

                  2891.     Automatic Extraction of 3D Cortical Profiles as the Basis for Anatomically-Based Cortical Parcellation

Marcel Weiss1, Gabriele Lohmann1, Gerik Scheuermann2, Robert Turner1

1Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Saxonia, Germany; 2Institute of Computer Science, University of Leipzig, Leipzig, Saxonia, Germany

We present a framework for the creation of three-dimensional cortical profiles based on cortical surface reconstruction.

                  2892.     The EMD-Test for Shape-Based Morphometry.

Nathan Faggian1,2, India Bohanna2, Leigh A. Johnston2,3, Nellie Georgiou-Karistianis4, Gary Francis Egan1,2

1Center for Neuroscience, Melbourne University, Melbourne, Victoria, Australia; 2Neuroimaging, Howard Florey Institute, Melbourne, Victoria, Australia; 3Department of Electrical and Electronic Engineering, Melbourne University, Melbourne, Victoria, Australia; 4Centre for Developmental Psychiatry, Monash University, Melbourne, Victoria, Australia

Shape analyses commonly assume normality and apply an appropriate statistical test correction. However, non-parametric methods such as the KS-test are more suitable when normality is not assured. Parametric methods are inaccurate when applied to non-normally distributed data and require correction. Our aim was to use non-parametric shape analysis methods to investigate shape differences between Huntingtons Disease (HD) patients and controls. Non-parametric shape differences were examined using a novel statistical method based on Earth Movers Distance (EMD).

                  2893.     Quantitative Voxel-Based Analysis of T1-Weighted MRI Signal Intensity

David Fenton Abbott1,2, Gaby S. Pell1,2, Heath Pardoe1,2, Graeme D. Jackson1,3

1Brain Research Institute, Florey Neuroscience Institutes (Austin), Melbourne, Victoria, Australia; 2Department of Medicine, The University of Melbourne, Melbourne, Victoria, Australia; 3Departments of Medicine and Radiology, The University of Melbourne, Melbourne, Victoria, Australia

Abnormalities in the brain generally manifest on MRI as changes in shape (morphometry) or changes in the nature of the tissue (signal intensity). Voxel Based Morphometry (VBM) is a well-known whole brain quantitative way of assessing morphometric changes. Here we demonstrate the effectiveness of a complementary method, Voxel-Based Iterative Sensitivity (VBIS) analysis, applied to directly assess signal intensity of T1-weighted MRI. In a group of patients with left hippocampal sclerosis, compared to healthy controls, VBIS-T1 was able to detect the major regions of tissue abnormality as well as more subtle pathology often not evident using conventional quantitative measures.

                  2894.     MR Brain 3D Contouring Using Atlas Matching and Snake Edge Detection Combined Approach

Robert Ambrosini1, Walter O'Dell1

1University of Rochester, Rochester, NY, USA

Brain extraction is a critical pre-processing step for enabling automated lesion detection algorithms to achieve a minimal false positive rate at a low computational cost for MR datasets. We have developed a method that combines atlas-based techniques with image edge detection capabilities provided by snakes for the purpose of performing brain extraction while preserving the MR appearance of lesions. We were able to achieve a Jaccard Index value of 0.87 and a Dice Coefficient of 0.93. These preliminary results indicate our approach’s potential value as a method for extracting the MR brain prior to employing a lesion detection algorithm.

                  2895.     A Geo Cut Algorithm for Brain MRI Segmentation

Jie Zhu1, Ashish Raj2, Ramin Zabih3

1Electrical Engineering, Cornell U, Ithaca, NY, USA; 2Radiology, Weill Medical College of Cornell University, New York, NY, USA; 3Computer Science, Cornell U, Ithaca, NY

Segmentation of MR brain images is an important step in many clinical applications, but is challenging due to the vast amount of fine structures. Conventional EMS algorithm fails in presence of noise, partial voluming or fine details and textures. Recent improvement were reported using graph cuts but standard inter-voxel edge weighting in graph cuts fails near small structures. We propose a graph-based combinatorial algorithm called geo-cuts, where image gradient magnitudes and gradient directions determine the weighting, to overcome this problem. Our results show that the overall percentage of correctly classified voxels is higher using the geo-cuts method except when noise is present. The Dice Similarity Measure indicates that both methods work well with geo-cuts generally outperforming standard graph cuts for white matter and CSF and vice versa for gray matter. Visually, geo-cuts gives better performance on both real and synthetic images with respect to fine structures of white matter and CSF.

                  2896.     Reliable Segmentation of Subcortical Gray Matter Structures Using Magnetization Transfer (MT) Maps

Nikolaus Weiskopf1, Bogdan Draganski1,2, Richard Frackowiak1, John Ashburner1, Gunther Helms3

1Wellcome Trust Centre for Neuroimaging, University College London, London, UK; 2Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany; 3MR-Research in Neurology and Psychiatry, Goettingen University, Goettingen, Germany

Segmentation and voxel-based morphometry (VBM) of subcortical structures is partly limited by the reduced contrast of these structures in T1w images that it is based on. We have developed a novel approach based on magnetization transfer (MT) maps instead of T1w images. MT maps are not affected by T1 shortening due to iron accumulation that degrades T1 contrast. We have demonstrated, in a group of 49 subjects, that MT-based segmentation improves the delineation of subcortical structures significantly. Thus, this method promises to be a powerful novel approach to MRI based morphometric descriptions of subcortical structures.

                  2897.     A Rat Brain Template of Class Distribution Maps Based on Diffeomorphic Image Registration

lei zheng1, Sarah Biedermann, Jürgen Hesser2, Alexander Sartorius, Wolfgang Weber-Fahr

1central institute of mental health, mannheim, mannheim, baden-württemberg, Germany; 2Medical Faculty of Mannheim, University of Heidelberg

Brain tissue segmentation is an important step to post-process MR images. From part of the data, the rat brain was extracted and segmented. The individual segmentation results were co-registered based on diffeomorphic image registration to generate a rat brain template of class distribution maps. With this template, later, the segmentation of rat brain can be achieved automatically.

Image Registration
Exhibit Hall 2-3                    Thursday 13:30-15:30

                  2898.     Assessing the Impact of Image Registration on Texture Analysis of MR Images

Henry Tran1,2, Roger Tam1,3, Anthony Traboulsee1,4, David Li1,3

1MS/MRI Research Group, Vancouver, BC, Canada; 2Faculty of Medicine, University of British Columbia (UBC), Vancouver, BC, Canada; 3Radiology, UBC; 4Neurology, UBC

The study looks at the impact of image registration on texture analysis of MR images. Texture analysis is a technique used to detect subtle changes in MR image to elucidate certain pathologies that are too subtle to detect visually. Using MR images of patients with definite MS, image resampling was performed using the linear, Blackman and B-spline interpolation methods. The polar Stockwell transform was then computed for the MS lesion regions of the original and resampled images. The study found that the differences in the texture analysis values after resampling was significant enough to affect the texture analysis results.

                  2899.     Integration of Model Information Into Model-Free Multivariate Analyses of Structural and Diffusion MRI Data

Stephen Smith1, Achim Gass2, Andreas U. Monsch3, Anil Rao4, Brandon Whitcher4, Paul M. Matthews4, Christian Beckmann1,5

1Oxford University FMRIB Centre, Oxford, Oxon, UK; 2Depts. Neurology and Neuroradiology, University Hospital Basel, Switzerland; 3Memory Clinic Basel, Switzerland; 4GlaxoSmithKline, Clinical Imaging Centre, London, UK; 5Clinical Neuroscience Department, Imperial College London, UK, UK

Model-based analysis methods are generally reliable in identifying expected responses, aiding interpretability of results. Conversely, model-free methods are able to find “surprising” effects in the data, or separate out structured confound processes from signals of interest. Little work has been carried out to combine the best aspects of the different approaches. We present initial results from inserting model information into a multivariate model-free approach and compare this approach to a GLM analysis as well as to PLS and CVA. We find that approaches containing both model-based and data-driven aspects are almost always more interpretable than rigidly enforcing model structure.

                  2900.     Establishing Probabilistic Chinese Human Brain Templates Using HAMMER Elastic Registration Algorithm

Zhentao Zuo1, Xing Wang2, Nan Chen2, Jing Luo1, Yan Zhuo1, Dinggang Shen3, Kuncheng Li2, Rong Xue1

1State Key Laboratory of Brain and Cognitive Science, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China; 2Department of Radiology, Xuan Wu Hospital, Capital Medical University, Beijing, China; 3Department of Radiology & BRIC, School of Medicine, University of North Carolina, Chapel Hill, NC, USA

We have successfully constructed Chinese human brain templates for male and female groups using the HAMMER automatic deformable registration algorithm. Sharp white matter and gray matter boundaries and brain ventricular edges of the templates were displayed compared with other normalization results. The average Chinese human brain templates had smaller brain volumes but wider brain shape than the MNI templates of the western society. In the future, we would further apply the HAMMER registration algorithm to a larger data set of Chinese people and generate the probabilistic Chinese brain templates, to promote the establishment of digitized standard brain (DSB) database for fMRI and disease-related neuroscience studies in China.

                  2901.     Fully Automated DTI Postprocessing Embedded in Scanner Control Software: Deghosting, Denoising, Smoothing, and Registration to a Standard Atlas

Kurt Hermann Bockhorst1, Jaivijay Ramu1, Venkata Kishore Mogatadakala1, Cheukkai K. Hui1, Ponnada A. Narayana1

1DII, University of Texas, Houston, TX, USA

Diffusion tensor imaging (DTI) produces large data sets and requires significant post-processing that includes minimization of image artifacts such as ghosting and geometric distortions. The post-processing generally requires significant user intervention that is tedious and could introduce bias and human errors. We implanted a fully automatic pipeline for processing the DTI data that is initiated from the scanner control program. The sole input by the user is the selection of the files to be processed. The automatic processing pipeline includes deghosting, extrameningial stripping, spatial normalization and calculation of DTI metrics. This pipeline allows processing multiple studies in a batch job.

                  2902.     Automated Registration of Sequential Breath-Hold DCE-MRI Images

Sivaramakrishnan Rajaraman1, Jeffery J. Rodriguez1, Christian Graff2, Maria I. Altbach2, Tomislav Dragovich3, Claude B. Sirlin4, Ronald L. Korn5, Natarajan Raghunand2

1Electrical & Computer Engineering, The University of Arizona, Tucson, AZ, USA; 2Radiology, The University of Arizona, Tucson, AZ, USA; 3Cancer Center, The University of Arizona, Tucson, AZ, USA; 4Radiology, The University of California, San Diego, CA, USA; 5Scottsdale Medical Imaging Ltd., Scottsdale, AZ, USA

We introduce a novel technique for registration of DCE-MRI images, the sequential elastic registration (SER) scheme. The performance of SER is compared with two recently published schemes for DCE-MRI registration, progressive principal components registration (PPCR) and pharmacokinetic model-driven registration (PMDR). When tested on a novel DCE-MRI software phantom, all 3 registration schemes were found to be comparably accurate for registering relatively large regions of interest (ROIs) such as muscle, liver, and spleen. However, SER outperformed the other two schemes in registering smaller but significant ROIs such as tumor rim. Quantitative results and performance statistics from clinical DCE-MRI data are presented.

                  2903.     Automatic Scan Plane Planning Method for Brain MRI

Xiaodong Tao1, Sandeep Narendra Gupta1

1Visualization and Computer Vision Lab, GE Global Research Center, Niskayuna, NY, USA

Consistent scan plane prescription for MRI is very important clinically for easy comparison of images across subjects and in serial evaluation of longitudinal exams. Typical workflow requires careful manual slice positioning and suffers from inter-operator variability. Existing automated methods use probabilistic atlases or segmentation of anatomic landmarks to align images. We present a novel scheme for registering brain images using an implicit symmetry axis determination. This algorithm is robust to asymmetry and large motion. We have incorporated this approach in a clinical MR system and demonstrated its usefulness in automatically obtaining consistent imaging planes across examinations irrespective of subject position.

                  2904.     Automatic Detection of Quiescent Cardiac Phases Using Navigator Echoes with Adjacent Complex Correlation Algorithm

Maggie M. Fung1, Vincent B. Ho2, Maureen N. Hood2, Yuval Zur3, Ehud J. Schmidt4

1Applied Science Lab, GE Healthcare, Waukesha, WI, USA; 2Radiology, Uniformed Services University, Bethesda, MD, USA; 3GE Healthcare, Tirat Carmel, Israel; 4Radiology, Brigham and Women’s Hospital, Boston, MA, USA

A novel automatic quiescent cardiac phase detection technique is presented. An adjacent complex correlation algorithm was used to analyse the cardiac navigator echoes acquired continously throughout the cardiac cycle. We have shown the feasibility of detecting the timing and duration of the quiescent phases using this technique. Respiratory motion can also be detected effectively. This technique could be applied to various cardiac pulse sequences for automatic and adaptive triggering, and therefore, increase the robustness of cardiac imaging.

                  2905.     Deformation Based Morphometry with Implicit Reference-Based Registration: Validation and Detection of Structural Changes in a Primate Model of Early-Life Stress

Xiujuan Geng1, Thomas J. Ross1, Svetlana Chefer1, Hong Gu1, Elliot Stein1, Yihong Yang1

1Neuroimaging Research Branch, National Institute on Drug Abuse, National Institutes of Health, Baltimore, MD, USA

A framework of deformation-based morphometry (DBM) with implicit reference-based group (IRG) registration is proposed. Comparisons between the proposed DBM method and the commonly used approach were made and results demonstrated more accurate volume change detection. Validation of cross group volumetric comparison using DBM was performed by simulating known volume change and deformation fields. The DBM technique was applied to a study of monkey brain morphological change due to early life stress, and compared to manual segmentation results. DBM provided consistent but more localized structural changes.

                  2906.     Assessment of SPM5’s Brain Registration Performance Using Landmark Points

Joao Miguel Pereira1, Guy B. Williams1, Li Xiong2,3, Julio Acosta-Cabronero1, George Pengas2, Peter J. Nestor2

1Wolfson Brain Imaging Centre, University of Cambridge, Cambridge, UK; 2Clinical Neurosciences, University of Cambridge, Cambridge, UK; 3Neurology Department, Wuhan University, Wuhan, China

The performance of registration algorithms is known to be one of the limiting factors of automated morphology methods. One such technique is voxel based morphometry (VBM), which is used to detect structural changes in diseased cohorts by contrasting them with healthy subjects – this is highly dependent on the performance of the registration step. The goal of this study was to quantify the misregistration of well characterised landmark points using standard SPM5 tools in both normal and diseased cohorts. Results showed that the standard unified segmentation and more complex methods such as DARTEL yield comparable results.

                  2907.     Retrospective Evaluation of Brain PET-MR Registration

Zuyao Y. Shan1, Sara J. Mateja2, Wilburn E. Reddick3, John O. Glass3, Barry L. Shulkin4

1Division of Translational Imaging Research, Department of Radiological Sciences  , St. Jude Children's Research Hospital , Memphis, TN, USA; 2Eckerd College, St. Petersburg, FL, USA; 3Division of Translational Imaging Research, Department of Radiological Sciences, St. Jude Children's Research Hospital, Memphis, TN, USA; 4Division of Nuclear Medicine, Department of Radiological Sciences, St. Jude Children's Research Hospital, Memphis, TN, USA

We retrospectively evaluated the accuracy of PET-MR atlas fusion of 5 well-accepted registration methods, using data acquired for purposes of clinic care. Both visual inspection and quantitative analysis showed that the affine registration based on ratio image uniformity is the most accurate for PET-MR atlas registration.

                  2908.     Scaling Methods for Registration of Pre- And Post-Androgen Ablation MR Prostate Images

Sophie Riches1, Mike Partridge2, Geoff S. Payne1, Vonnie A. Morgan1, Nandita M. deSouza1

1CRUK Clinical Magnetic Resonance Group, Institute of Cancer Research & Royal Marsden NHS Trust, Sutton, Surrey, UK; 2Department of Physics, Institute of Cancer Research & Royal Marsden NHS Trust, Sutton, Surrey, UK

Synposis: MR offers improved visualisation of the prostate gland over the standard CT imaging currently used for planning of radical radiotherapy for prostate cancer. Hormone therapy often prescribed prior to radiotherapy reduces the discrimination of tumor with functional MR techniques therefore necessitating registration of functional information acquired prior to androgen ablation to anatomical images acquired immediately prior to radiotherapy. Comparison of three scaling techniques to account for changes in prostate shape during hormone treatment found assuming bi-linear shrinkage of the prostate allowed pre-treatment tumor regions to be registered to post-treatment anatomical images.

                  2909.     Quantitative Comparison of Methods for Spatial Normalisation of CASL Perfusion MR Images

Stephen J. Wastling1, Owen O'Daly2, Fernando O. Zelaya2, Matthew Howard2, David C. Alsop3, Ruth L. O'Gorman2,4

1Medical Engineering and Physics, King's College Hospital, London, UK; 2Centre for Neuroimaging Sciences, Institute of Psychiatry, London, UK; 3Neuroradiology, Beth Israel Deaconess Medical Center, Boston, MA, USA; 4Neuroradiology, King's College Hospital, London, UK

This study quantitatively examined the accuracy and precision of various registration software methods, cost functions, and templates for the spatial normalisation of continuous ASL (CASL) perfusion MRI data. While the registration precision was broadly comparable across all methods, the registration accuracy varied considerably, although several methods yielded registrations accurate to 1 mm or less.

                  2910.     A MR Brain Template of Young Children

Zuyao Y. Shan1

1Division of Translational Imaging Research, Department of Radiological Sciences  , St. Jude Children's Research Hospital , Memphis, TN, USA

We developed a brain atlas of young children on the basis of MR images from 96 children aged from 6-8 yrs old and used it for spatial normalization of PET images. The spatial normalization results showed that a pediatric brain atlas is more appropriate for pediatric neuroimaging research than the adult brain atlas.

                  2911.     Image Registration of Whole-Body Mouse MRI

X Josette Chen1, Satheesh Maheswaran2, Daniel Rueckert2, R Mark Henkelman1

1Mouse Imaging Centre, Hospital for Sick Children, Toronto, Ontario, Canada; 2Department of Computing, Imperial College London, London, UK

We registered together 13 whole-body MR images of perfusion-fixed mice.

                  2912.     Diffeomorphic Normalization Via Constant Acceleration Field

Yung-Chin Hsu1, Ching-Han Hsu1, Wen-Yih Isaac Tseng2

1Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, Hsinchu, Taiwan; 2Center for Optoelectronic Biomedicine and Department of Medical Imaging, National Taiwan University College of Medicine and Hospital, Taipei, Taiwan

Image registration plays an important role in neuroimaging research. Under diffeomorphic framework, it is possible to solve the large-deformation problem usually encountered in image registration. Both Beg¡¦s LDDMM and Ashburner¡¦s constant velocity filed methods are diffeomorphic. We extended Ashburner¡¦s method by using a constant acceleration filed. The result shows that LDDMM has shortest path length and lowest energy. Comparing our method and the constant velocity field method, the prior has lower energy and the later has shorter path length. The result suggests that these three methods might have their own strength and weakness in accuracy and computation cost.

                  2913.     Registration Methodology of Histology with in Vivo 3T MRI for Human Prostate

Hyunjin Park1, Morand Piert1, Bing Ma1, Charles Meyer1

1University of Michigan, Ann Arbor, MI, USA

Registration of histology and in vivo MRI is an important problem as registered histology onto the in vivo MRI enables pixel by pixel comparison of in vivo MRI imaging features with ground truth of ex vivo histology. For human prostate, registration of histology and in vivo MRI is a difficult as the prostate undergoes a complex set of deformations. A methodology has been proposed to register various in vivo scans with histology for human prostate. Here we have refined and applied the methodology to in vivo 3T diffusion MRI for two cases.

Image Correction Schemes
Exhibit Hall 2-3                    Monday 14:00-16:00

                  2914.     3D Phase Unwrapping Based on Expansion in Chebyshev Polynomials

Jason Langley1,2, Qun Zhao1,2

1Physics and Astronomy, The University of Georgia, Athens, GA, USA; 2Bioimaging Research Center, The University of Georgia, Athens, GA, USA

A three dimensional phase unwrapping algorithm is presented in this

                  2915.     Effects of Distortion and Inhomogeneity Correction on Brain Morphometry with 3T MRI

Osamu Abe1, Hidemasa Takao1, Hiroyuki Kabasawa2, Hidenori Yamasue3, Masami Goto4, Wataru Gonoi1, Harushi Mori1, Akira Kunimatsu1, Yoshitaka Masutani1, Kiyoto Kasai3, Kuni Ohtomo1

1Department of Radiology, University of Tokyo, Tokyo, Japan; 2GE Yokogawa Medical Systems, Tokyo, Japan; 3Department of Neuropsychiatry, University of Tokyo, Tokyo, Japan; 4Radiological Technology, University of Tokyo Hospital, Tokyo, Japan

We investigate the effects of distortion and inhomogeneity correction on brain morphometry with 3T MRI. Significant group difference was observed in the brainstem, bilateral cerebral and cerebellar cortices between uncorrected and corrected images using SPM5 and FreeSurfer. Gradient-nonlinearity-induced spatial distortion and signal inhomogeneity correction allow accurate cross-platform or longitudinal comparisons of quantitative morphometry results.

                  2916.     New Thresholding Correction Methods on BOLD FMRI for Interpreting Brain Function

Han-Su Baek1, Gwang-Woo Jeong1,2, Tae-Hoon Kim1, Hyeong-Jung Kim1, Thirunavukkarasu Sundaram1, Gwang-Won Kim1, Tae-Jin Park3

1Department of Biomedical Engineering, Chonnam National University Medical School, Gwangju, Korea; 2Department of Radiology, Chonnam National University Medical School, Gwangju, Korea; 3Department of Psychology, Chonnam National University, Gwangju, Korea

In order to reduce the type I and II errors and obtain more accurate brain activation map, fMR images were statistically analyzed with uncorrected t-test, false discovery rate(FDR) and family-wise error rate(FWE) in the general linear model (GLM) through statistic parameter mapping software(SPM). This study compared and evaluated the power of thresholding correction methods with uncorrected t-test, FDR and FWE, and further discussed about new thresholding correction method.

                  2917.     Evaluation of Methods for Removal for Background Field Inhomogenities in Susceptibility Weighted Phase Imaging

Ferdinand Schweser1, Andreas Deistung1, Berengar Wendel Lehr1, Juergen R. Reichenbach1

1Medical Physics Group, IDIR, University Clinics Jena, Friedrich-Schiller-University, Jena, Thueringen, Germany

We quantitavely compare different methods for removing background field inhomogenities in susceptibility weighted phase imaging. We discovered that the commonly used homodyne-filter changes phase differences while other methods e.g. spherical mean value estimation keep phase differences and properly reduce background field inhomogenities.

                  2918.     Evaluating Reliability of Performance Metrics for Bias Field Correction in MR Brain Images

Vitali Zagorodnov1, Zin Yan Chua1, Weili Zheng1

1Computer Engineering, Nanyang Technological University, Singapore, Singapore

Performance evaluation of nonuniformity correction approaches is usually performed indirectly, on the basis of remaining tissue intensity variability. However, inconsistencies between indirect measures reported in recent studies suggest that indirect measure might be poor proxies for true correction performance. This was confirmed by reliability study reported here. Reliability can be improved by evaluating the same metrics over conservative brain tissue masks and by applying slight smoothing to suppress image noise. On real brain MR data, due to potential difficulty in obtaining a conservative GM mask, we recommend using coefficient of variation (CV) measured on smoothed conservative WM mask.

                  2919.     Distortion Correction for Echo Planar MR Imaging Using the Regularized Inverse Solution of the Point Spread Function (PSF) Map ©

Yu Cai1, Qingwei Liu2, Hongyu An1, Weili Lin1, Craig Hamilton3

1Radiology, University of North Carolina-Chapel Hill, Chapel Hill, NC, USA; 2BME, University of North Carolina-Chapel Hill, Chapel Hill, NC, USA; 3BME, Wake Forest University, Winston Salem, NC, USA

The PSF mapping technique for distortion correction of EPI images is improved by solving for the inverse solution of the PSF map using Tikhonov regularization. the Tikhonov regularization-based method is robust for the correction of distortion caused by magnetic susceptibility. In particular, in light of the continuing increase of the magnetic field strength in both clinical and research arenas, the proposed Tikhonov method should further improve the image quality of high field MRI.

Quantitative Image Analysis
Exhibit Hall 2-3                    Tuesday 13:30-15:30

                  2920.     Fast and Automatic Quantification of Cardiac Perfusion MRI

Andreas Weng1, Christian Oliver Ritter1, Joachim Lotz2, Meinrad Joachim Beer1, Dietbert Hahn1, Herbert Köstler1

1Institut für Röntgendiagnostik, Uniklinik Würzburg, Würzburg, Bayern, Germany; 2Institut für Radiologie, Medizinische Hochschule Hannover, Hannover, Niedersachsen, Germany

An automatic postprocessing workflow for cardiac perfusion MRI was evaluated with perfusion studies of 75 subjects. Motion correction and image segmentation was performed both completely automatically and manually. Additionally the interobserver variability was investigated by manually segmenting 10 datasets by two different observers.

                  2921.     Orientation Dependence of Grey/white Matter Contrast in Ultra High Fields

Andreas Schäfer1, Richard Bowtell2, Robert Turner1

1Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany; 2Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, UK

Recently, it has been shown that high resolution T2*-weighted images acquired at ultra high field show much more T2* heterogeneity in white matter than at lower field, which alters with the orientation of fibre bundles with respect to the static field. Phase images show high contrast between different compartments. Since phase images reflect susceptibility variations more strongly than magnitude data, we have used phase images at 7 Tesla to address the orientation dependence of GM/WM contrast. Furthermore we compare measurements with simulations based on applying a field calculation to the HUGO body model.

                  2922.     Systematic Comparison of Quantitative T1 Mapping Methods at 7 T High Field

Thomas Trantzschel1, Kai Zhong1, Oliver Speck1

1Biomedical Magnetic Resonance, Otto-von-Guericke-University, Magdeburg, Germany

B1-inhomogenity leads to problems in quantitative T1-mapping case at high field. TESSA as a new rapid method has been presented. This study systematically studied the efficiency and accuracy of TESSA and other established methods. While most methods need an additional B1-map to determine T1 correctly, TESSA is able to estimate T1 and B1 from one measurement. The SNR and accuracy of TESSA can be increased by averaging without sacrificing scan time compared to IR. In addition The T1 fitting algorithm for TESSA is much faster and easy to implement.

                  2923.     Complex-Valued Voxel Thresholding Increases Image Contrast in SWI

Daniel B. Rowe1, E. Mark Haacke2

1Department of Biophysics, Medical College of Wisconsin, Milwaukee, WI, USA; 2The MRI Institute for Biomedical Research, Detroit, MI

It is often desirable to threshold signal plus noise voxels from pure noise voxels. Generally thresholding utilizes only the magnitude of the images. More recently a method has used both the magnitude and phase. This work is an extension and uses the normality of the real and imaginary values with phase coupled means. A statistic is derived that is F-distributed in large samples. In small samples Monte-Carlo critical values can be used. We apply this method to SWI images and show increased image contrast while it is found to be more robust to phase variations from unwanted field inhomogeneity effects.

                  2924.     Connectivity in Controls and Epilepsy: Demonstration of a Metabolic Network by MRSI

Jullie W. Pan1,2, Ruben Kuzniecky3, Susan Spencer4, Dennis Spencer1, Hoby P. Hetherington1,5

1Neurosurgery, Yale University School of Medicine, New Haven, CT, USA; 2BME, Yale University School of Medicine, New Haven, CT, USA; 3Neurology, New York University School of Medicine, New York, NY, USA; 4Neurology, Yale University School of Medicine, New Haven, CT, USA; 5Radiology, Yale University School of Medicine, New Haven, CT, USA

While resting fMRI connectivity studies have linked the thalamus and hippocampus in humans, whether such connectivity maps are manifest metabolically is unknown. Our goal is to establish the presence of such a network through 1H spectroscopic imaging data. We use 1H MRSI data from a subcortical set of loci with a common factor analysis to assess for metabolic connectivity in controls and epilepsy patients. Not only is the subcortical network readily detected in resting connectivity studies, it is important to evaluate this system in epilepsy as it may be thought to participate in nearly all seizures and epilepsy types.

                  2925.     Cortical Surface Structure Analysis in Sharks Using Magnetic Resonance Imaging (MRI)

Kara E. Yopak1, Greg T. Balls1, Lawrence R. Frank1

1Radiology, UCSD - Center for Scientific Computation in Imaging, La Jolla, CA, USA

The shark cerebellum has morphologically diversified throughout vertebrate evolution. We have previously developed a qualitative visual grading index for cerebellar foliation (n=60). Here we provide a more robust method using MRI in conjunction with rigorous 3D shape analysis methods to quantify our previous scheme. Quantification was based upon the number of parameters necessary to accurately describe the cortical surface decomposed onto a spherical harmonic basis using the spherical harmonics decomposition (SHD). These methods, previously used to characterize global brain shape in humans, were preliminarily applied to a small number of representative shark species. More foliated shark brains required a more complex SHD to accurately fit its cortical surface.

                  2926.     Computer-Aided Detection (CAD) for Breast MRI at 3.0T: A Comparison Between Four Observers with Different Levels of Experience.

carla meeuwis1, S M. van de Ven2, G Stapper3, A M. Fernandez Gallardo3, M A. van den Bosch3, 3, W P. Mali3, W B. Veldhuis3

1Radiology, Rijnstate Hospital, Arnhem, Netherlands; 2Radiological Sciences Laboratory Stanford, Lucas MRS Imaging Center; 3Department of Radiology, University Medical Center Utrecht

MRI of the breast is a promising diagnostic modality with a high sensitivity. The main drawback is its relatively low specificity. Computer-aided detection systems have been developed that automate processing and kinetic analysis. This study was performed to compare the accuracy of 3.0T breast MRI interpretation, between manual- and fully automated-kinetic analysis in 71 lesions for four observers with different levels of experience in reading breast MRI. Our study shows that CAD system significantly improved the specificity of interpretation compared with manual analysis of enhancement kinetics. Additionally there was no significant difference in diagnostic accuracy using CAD between the observers.

                  2927.     Prediction of Malignant Breast Lesions from MRI Features: A Comparison of Artificial Neural Network and Logistic Regression Techniques

Ke Nie1, Christine E. McLaren2, Wen-Pin Chen2, Jeon-Hor Chen1,3, Orhan Nalcioglu1, Min-Ying Lydia Su1

1Tu & Yuen Center for Functional Onco-Imaging, University of California, Irvine, Irvine, CA, USA; 2Department of Epidemiology, University of California, Irvine, Irvine, CA, USA; 3Department of Radiology, China Medical University, Taichung, Taiwan

This study compared two different approaches using artificial neural network (ANN) and logistic regression analysis (LRA) for selecting diagnostic models to differentiate between malignant (N=43) and benign (N=28) lesions. For each case, 8 parameters were used to characterize morphology, 10 GCLM and 14 Laws features were used to characterize enhancement texture (homogeneity) within the lesion. The diagnostic performance of classifiers obtained by ANN or different LRA models was similar. While ANN is more robust and does not require a high level of operator judgment, LRA may be used complimentarily to understand the role of the selected predictors.

                  2928.     Comparison of Normalized DTI Analytical Methods: Detection Powers of Voxel-Based Analysis and Sub-Atlas Based Analysis

Koji Sakai1,2, Susumu Mori2, Kenichi Oishi2, Andreia Faria2

1Kyoto University, Kyoto, Japan; 2Johns Hopkins University, USA

The voxel-based group analysis (VBA) is one of the most effective examination methods of the entire white matter (WM) of brain. However, the VBA often suffers from low statistical power (high false discovery rate), which caused by embedded noise in voxels. We attempted to further extend the ABA to obtain statistically stronger detection power than the VBA. We propose a sub-atlas-based analysis (SBA), which uses 3D plane made from the fitting curves to the WM atlas. We compared detection power among the VBA and the SBA by using ICBM-152 normal brain artificially embedded abnormal values.

                  2929.     Selective Evaluation of Fast MRI Reconstruction Artifacts Using Case-PDM

Jun Miao1, David L. Wilson1,2

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA; 2Radiology, University Hospitals of Cleveland, Cleveland, OH, USA

The perceptual difference model (Case-PDM) is being used to quantify image quality of fast, parallel MR acquisitions and reconstruction algorithms as compared to slower, full k-space, high quality reference images. To date, most perceptual difference models average image quality over a wide range of image degradation effects. Here, we develop and evaluate a modification, which allows one to selectively quantify noise, blur, and aliasing artifacts. Results from over 120 images indicate that the PDM compares favorably to human evaluation.

                  2930.     Comparison of Methods for the Measurement of Cardiac Function in Rats

Johannes Riegler1,2, Anthony Niel Price1, Kenneth King Cheung1, Jon Orlando Cleary1, Mark Francies Lythgoe1

1Centre for Advanced Biomedical Imaging (CABI), Department of Medicine and Institute of Child Health, University College London (UCL), London, UK; 2Centre for Mathematics and Physics in the Life Sciences and Experimental Biology (CoMPLEX), UCL, London, UK

Cardiac MRI is used for the assessment of animal models of heart failure. Manual image processing is frequently used to determine cardiac function. Since this is time consuming two different clinical cardiac analysis tools were compared against manual segmentation to establish the intra-observer variability and preclinical applicability for rats pre- and post-myocardial infarction (MI). Semi-automated methods offered faster analysis time, and showed similar intra-observer variability and acceptable limits of agreement compared to a manual method for rodent CMR. However, a different bias was found for the estimation of the ejection fraction for pre- and post-MI data.

                  2931.     Calculation of the Magnetic Susceptibility from Susceptibility Weighted Phase Images

Andreas Deistung1, Berengar W. Lehr1, Ferdinand Schweser1, Jürgen R. Reichenbach1

1Medical Physics Group, Institute for Diagnostic and Interventional Radiology, University Clinics, Friedrich-Schiller-University, Jena, Germany

We present a modified method for quantifying the magnetic susceptibility for arbitrary shaped objects from high-resolution susceptibility weighted phase images. The inverse problem of going from the relative difference field (RDF) to the magnetic susceptibility was extended by a shutter and a possible offset of the RDF. The field inhomogeneity pattern seen in phase images was efficiently removed in the susceptibility maps.

                  2932.     Magnetic Resonance Elastography of Polymer Gel Dosimetry Phantoms

Yogesh Kannan Mariappan1, Silvio L. Vieira2, Julianna P. Fernandes2, Antonio Adilton Carneiro2, Mostafa Fatemi3, Richard L. Ehman1

1Department of Radiology, Mayo clinic, Rochester, MN, USA; 2Departamento de Fisica e Mathematica da FFCLRP, Universidade de Sao paulo, Ribeorao Preto, SP, Brazil; 3Department of Physiology and Biomedical Engineering, Mayo clinic, Rochester, MN, USA

Polymer gel dosimetry involves the estimation of radiation dose distribution based on radiation induced changes. Since these phantoms undergo polymerization of monomers, we hypothesized that it would affect the mechanical properties. We tested this hypothesis in a series of phantom experiments and investigated the stiffness of radiated phantoms as a function of absorbed radiation doses using Magnetic Resonance Elastography (MRE). The results show that the stiffness of a polymer gel dosimeter is directly correlated to the amount of the applied radiation dose and hence it would be feasible to estimate radiation dose distribution with MRE.

                  2933.     Model Regularization with Blind Deconvolution

Jacob U. Fluckiger1, Matthias C. Schabel1, Edward VR DiBella1

1University of Utah, Salt Lake City, UT, USA

We have developed an alternating minimization with model (AMM) algorithm for estimating the arterial input function directly from measured tissue activity curves. This method uses an analytic form for the AIF to regularize noise in measured DCE-MRI data. Simulations show the AMM method performs nearly as well as conventional deconvolution with a known AIF.