Zhipeng Cao^1,2, Eduard Y. Chekmenev¹, and William A. Grissom^1,2
1Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States, ²Biomedical Engineering, Vanderbilt University, Nashville, TN, United States

The theory of Bloch-Seigert based RF frequency encoding is presented with a proposed pulse sequence and subsequent signal expression. They are validated by Bloch simulation and a non-linear image reconstruction as an initial demonstration.

Keigo Kawaji¹, Mehmet Akçakaya¹, Sébastien Roujol¹, Warren J Manning^1,2, and Reza Nezafat^1
1Medicine, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts, United States, ²Radiology, Beth Israel Deaconess Medical Center and Harvard Medical School, MA, United States

Golden angle interleaving of radial k-space lines can be incorporated into 3D coronary MRI, so that a large acquisition window can be acquired for retrospective reconstructions of any window subset with minimal cardiac motion. To enable such reconstructions with minimal artifacts, uniform distribution of the radial lines becomes desirable, but this coverage is dependent on parameters such as the rotation angle and acquisition window size. We present an approach to determine the optimal rotation angle so that the radial spokes are evenly distributed for any subset of the acquisition window in a 3D radial stack-of-stars coronary MRI sequence.

Koray Ertan^1,2 and Ergin Atalar^1,2
1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey, ²National Magnetic Resonance Research Center (UMRAM), Ankara, Turkey

Effect of nonlinear gradients on the RF excitation profile is formulated by using independent k-space variables. During pulse design, target excitation profiles can be calculated in order to eliminate the effect of B1 inhomogeneity. Use of independent k-space variable for nonlinear gradients leads to design technique for optimally combining linear gradients and nonlinear gradients. Optimizations have been performed in order to achieve target excitation profile. An example case demonstrates that simultaneous use of linear gradients and nonlinear gradients can achieve more accurate excitation profile in comparison with the only linear or only nonlinear gradients.

Akash P. Kansagra¹ and Christopher P. Hess^1
1Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States

MR fingerprinting is based on the application of a pseudorandom pulse sequence to a tissue of interest and comparing the resulting signal to a large library of simulated signals corresponding to a variety of T1, T2, and resonance offsets. Computing this library is computationally expensive. Here, we outline a simultaneous simulation-matching algorithm that allows partial construction of the signal library. Using this method, we demonstrate increases in execution speed nearing two orders of magnitude. The algorithm described is a simple and straightforward means to accelerate the post-processing portion of the MR fingerprinting method.

Amaresha Shridhar Konar¹, Smera Lingesh¹, Ramesh Babu¹, and Sairam Geethanath^1
1Medical Imaging Research Center, Dayananda Sagar Institutions, Bangalore, karnataka, India

The current work provides a framework for undersampling patterns for the combination of compressed sensing and parallel imaging. The masks considered for this work are called “Rangoli”. These masks have pseudo-random distribution of points on Cartesian grid and are smoothly connected. In this work, a variable density spiral was plotted on rangoli and was tweaked by mapping spiral coordinates on to the nearest rangoli coordinates balancing consistency and smoothness error factors to obtain the final trajectory. The images were then evaluated based on the Point spread function (PSF) where PSF was undersampled at different acceleration factors and quantified by peak-signal-to-noise-ratio.

Henry Szu-Meng Chen^1,2, Angshul Majundar³, and Piotr Kozlowski^4
1UBC MRI Research Centre, Vancouver, British Columbia, Canada, ²Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada,³Indraprastha Institute of Informatino Technology, New Delha, DL, India, ⁴Radiology, University of British Columbia, Vancouver, BC, Canada

Myelin water imaging using CPMG sequence is slow, which is especially a concern in in vivo cases. Compressed sensing 3D CPMG experiments were performed on ex vivo cervical spinal cord sample using a variable density sampling scheme in the two phase encode directions. It was found that while the T2 weighted echo images were accurately reconstructed at 3x and 2x acceleration, the 3x accelerated myelin water fraction maps showed higher reading than the fully sampled reference dataset. The 2x accelerated scan did not have this issue.

Zhaoying Han¹, Eric Peterson¹, Sjoerd B. Vos^1,2, Rafael O'Halloran¹, Samantha Holdsworth¹, Eric Aboussouan¹, Nancy Fischbein¹, and Roland Bammer^1
1Radiology, Stanford University, Stanford, California, United States, ²Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands

Slice accelerated (multiband) DTI shows great potential for reducing scan times of lengthy acquisitions. Multiband imaging can provide significant time saving in long scans, such as high angular diffusion resolution imaging (HARDI) acquisition. DTI-based tractography seeded in the corpus callosum on the mid-sagittal slice were performed, showing comparable results for accelerated HARDI. This study shows that slice accelerations of 2-3 can be used to reduce the otherwise long scan time of HARDI to clinically feasible times, without losing image quality.

Fang Liu¹, Richard Kijowski², and Wally Block^1,3
1Department of Medical Physics, University of Wisconsin-Madison, Madison, Wisconsin, United States, ²Department of Radiology, University of Wisconsin-Madison, Wisconsin, United States, ³Department of Biomedical Engineering, University of Wisconsin-Madison, madison, Wisconsin, United States

Digital simulation dramatically facilitates the understanding and development of new MR imaging methods. In this work, we optimized and improved our proposed fast 3D parallel numerical MR simulation package 'MRiLab' for incorporating many new challenging simulation features including multiple RF transmitting and receiving, motion simulation, etc. The spin model is also extended to include multiple exchanging spin compartments to create more realistic MR signal for certain types of tissues with, for instance, magnetization transfer (MT) modulation. This abstract was aimed to demonstrate the feasibility of using MRiLab for studying multiple types of MR experiments that remain challenging for numerical simulation.

Katsusuke Kyotani¹, Yoshiko Ueno², Satoru Takahashi², Yu Ueda¹, Tomoyuki Okuaki³, Nobukazu Aoyama¹, Kazuhiro Kitajima², Hideaki Kawamitsu¹, and Kazuro Sugimura^2
1Division of Radiology, Kobe University Hospital, Kobe, Hyogo, Japan, ²Department of Radiology, Kobe University Graduate School of Medicine, Kobe, Hyogo, Japan, ³Philips healthcare Asia pacific, Minato-ku, Tokyo, Japan

Our aim was to evaluate the usefulness of tissue-specific VRFA 3D-TSE optimized with measured T1 and T2 values of prostate gland comparing to standard 3D-TSE for T2W imaging. Contrast ratio between PZ and PZ cancer of the tissue-specific VRFA 3D T2W-TSE was significantly higher than that of standard 3D T2W-TSE (P < 0.05). Contrast ratio between PZ and TZ of the tissue-specific VRFA 3D T2W-TSE was significantly higher than that of standard 3D T2W-TSE (P < 0.05). Spatial resolution ratio of tissue-specific VRFA 3D-TSE was improved from standard 3D-TSE (P < 0.05)

Jing Li¹, Shuhui Cai¹, Congbo Cai², and Zhong Chen^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²Department of Communication Engineering, Xiamen University, Xiamen, Fujian, China

Immunity to field perturbation, spatial resolution and signal-to-noise ratio (SNR) are three significant factors needed to concern in MRI. Recently, a spatiotemporally encoded (SPEN) imaging approach based on linear frequency-swept excitation has been proposed to overcome the artifacts induced by various field perturbations. To demonstrate the superiority of SPEN imaging approach in reduced field of view (rFOV) imaging, a flexible rFOV imaging approach is proposed in this paper. Experiments on phantom and in vivo rat brain show that the flexible rFOV scheme can produce images with higher SNR and spatial resolution. Meanwhile, the regions of interest can be flexibly imaged.

Neville Gai¹ and John Butman^1
1Radiology & Imaging Sciences, NIH, Bethesda, Maryland, United States

3D FLAIR imaging of the entire brain using an extended modulated refocusing pulse train is now possible within a clinically reasonable time. However, a longer repetition time (with dead time) is still used to allow for adequate signal recovery. We describe a technique which uses variable repetition and inversion time to reduce total scan time. The scan time is reduced considerably while maintaining near equal SNR and CNR when compared with the constant repetition time sequence.

Peter Mark Smith¹, Benjamin Freed², B. D. Allen¹, Bruce S Spottiswoode³, M. Carr¹, Karissa Fortney-Campione¹, Marie-Pierre Jolly⁴, Christoph Guetter⁴, James C Carr¹, and Jeremy D Collins^1
1Radiology, Northwestern University, Chicago, IL, United States, ²Cardiology, Northwestern University, Chicago, IL, United States, ³Cardiovascular R & D, Siemens Healthcare, Chicago, IL, United States, ⁴Imaging and Computer Vision, Corporate Technology, Siemens Corporation, Princeton, NJ, United States

Myocardial strain may be useful to evaluate early alterations in myocardial structure-function. Deformation field analysis now enables myocardial strain calculation from balanced steady state free precession (bSSFP) cine images. However, the influence of field strength and temporal resolution on strain parameters is unknown. We compared strains from bSSFP cine images at 1.5 T and 3T across a range of temporal resolutions (12.5 to 39.2 msec). There was no significant influence of field strength or temporal resolution on RV longitudinal or LV radial, circumferential, or longitudinal global strain. Our results suggest that myocardial strain can be calculated from routine bSSFP sequences.

Zhe Wang^1,2, A. Nasiraei Moghaddam^2,3, Yunpeng Zou⁴, Subashini Srinivasan^1,2, J.Paul Finn^2,5, and Daniel B. Ennis^1,2
1Department of Bioengineering, University of California, Los Angeles, CA, United States, ²Department of Radiological Science, University of California, Los Angeles, CA, United States, ³Department of Biomedical Engineering, Amirkabir University of Technology, Tehran, Iran, ⁴4. Sino-Dutch Biomedical and Information Engineering School, Northeastern University, Shenyang, Liaoning, China, ⁵Biomedical Physics Interdepartmental Program, University of California, Los Angeles, CA, United States

Complementary radial tagging (CRT) generates radial tags with enhanced contrast throughout the entire cardiac cycle. CRT is potentially useful for measuring left ventricular (LV) contraction and torsion especially during late systole and early to late diastole. The previous CRT method was validated to work in >92% of clinic cases, but required a table position shift and the CRT pattern was non-ideal for some patients. Our objective was to develop a new CRT method that incorporates continuous RF phase adjustments to improve that CRT pattern. The new method generates an excellent CRT pattern without the need for a table shift.

Aiming Lu¹, Mitsue Miyazaki¹, Cheng Ouyang¹, and Xiangzhi Zhou^1
1Toshiba Medical Research Institute USA, Vernon Hills, IL, United States

Phase information is often neglected in MR imaging except a few applications such as flow quantification and susceptibility-weighted imaging (SWI).SWI usually acquires gradient echo images with long TEs to create susceptibility weighed magnitude images and has found applications in mainly neuroimaging. More recently, phase contrasts have been reported in several ultra-short TE (UTE) imaging applications. It is demonstrated here that phase images obtained with ultra-short TEs provide useful contrasts in knee and lung imaging. A new approach for visualizing both the magnitude and phase information simultaneously without information loss is then presented to facilitate the assessment of tissue characteristics.

Masanori Ozaki¹, Ken Arai¹, Kunihiro Miyoshi¹, Pauline Wong Worters², Suchandrima Banerjee², Arnaud Guidon², Hitoshi Ikeda¹, and Hiroyuki Kabasawa^3
1MR Engineering, GE Healthcare Japan, Hino-shi, Tokyo, Japan, ²ASL, GE Healthcare, WI, United States, ³Global Applications and Workflow, GE Healthcare, Hino-shi, Tokyo, Japan

Current MRE method offers techniques for diagnosis for large organs such as liver. Recent papers report that MRE for kidney could be useful for assessment of renal fibrosis. However, in the renal MRE, high spatial resolution is needed because pathologic evaluation is valid only on the renal cortex which is thin. This study proposes high spatial resolution MRE using Single Shot � Spin Echo –Echo Planar Imaging with reduced FOV for kidney.

Novena A Rangwala¹, Gopal Varma¹, Olivier Girard², Guillaume Duhamel², David B Hackney¹, and David C Alsop^1
1Department of Radiology, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts, United States, ²CRMBM UMR 7339, CNRS / Aix-Marseille Université, Marseille, France

The inhomogeneous magnetization transfer (IHMT) approach generates images both sensitive and specific to myelinated tissues, especially white matter (WM). We investigated the specificity of IHMT to several tissues and organs not containing myelin. IHMT images were acquired in cervical and thoracic spine and in the abdomen; and IHMT ratios (IHMTR) were compared with MT ratios (MTR) in these tissues. Results showed MTR of 6-27% in different tissues, with MTR ~ 23% in WM. In contrast, IHMTR demonstrated specificity to myelin-containing tissues with 5% in WM, < 2% in gray matter and virtually no IHMTR in other investigated tissues. IHMT may be a reliable marker for myelin in the investigation of WM disorders.

Olivier M. Girard¹, Valentin Prevost¹, Gopal Varma², David C. Alsop², and Guillaume Duhamel^1
1CRMBM UMR 7339, CNRS / Aix-Marseille Université, Marseille, France, ²Radiology Department, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States

Inhomogeneous magnetization transfer (ihMT) imaging has been proposed as a new method for imaging myelinated tissues of the central nervous system. Previous studies have been performed at 1.5T and 3T using different MR systems, and high similarities were observed. In this work the ihMT contrast is quantified at both field strengths using the same acquisition protocol and volunteers to further address the potential field dependency of the ihMT effect. IhMT datasets were very consistent and show no field dependency. The ihMT technique is very robust and experiments may be performed at various field strengths with similar contrast properties.

Olivier M. Girard¹, Valentin Prevost¹, Gopal Varma², David C. Alsop², and Guillaume Duhamel^1
1CRMBM UMR 7339, CNRS / Aix-Marseille Université, Marseille, France, ²Radiology Department, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States

Inhomogeneous magnetization transfer (ihMT) imaging has been proposed as a new method for imaging myelinated tissues of the central nervous system. Previous studies have shown promising results on both humans and small animals, at various magnetic field strengths. Various sequence parameters were used and a better understanding of the contrast mechanisms is necessary to further optimize the ihMT effect. Here we characterize the ihMT contrast with respect to the main sequence parameters of a pulsed ihMT preparation module at 1.5T. Very useful information can be derived from presented results in order to optimize ihMT signal and contrast.

Michael Helle¹, Nicole Schadewaldt¹, Marloes Frantzen-Steneker², Heinrich Schulz¹, Christian Stehning¹, Uulke van der Heide², and Steffen Renisch^1
1Philips Research Laboratories, Hamburg, Germany, ²Department of Radiation Oncology, The Netherlands Cancer Institute, Amsterdam, Netherlands

Radiation therapy planning can benefit from the superior display of soft tissue contrasts and the delineation of tumor and critical organs in magnetic resonance imaging. In this study, an advanced approach on the basis of Dixon MR techniques with subsequent reconstruction workflow is introduced for soft tissue and bone classification in order to generate MR-derived simulated computed tomography (CT) maps. For the first time, this method is evaluated in patients with corresponding true CT data and radiation therapy plans.

Yongquan Ye¹ and E. Mark Haacke^1
1Radiology, Wayne State University, Detroit, Michigan, United States

MR angiography for retrobulbar ocular vessels has been limited due to the fact that the vessels are buried in the excessive fatty tissues around the optical nerves, which has short T1 and long T2 to show high signals in most MR images, and thus the vessel-tissue contrast are low. We proposed using a newly developed technique, namely Binomial Off-resonant Rectangular (BORR) pulse, to completely suppress the fat signal. Compared to other MRA techniques, the BORR method yielded the best angiography contrast, and has the potential to be used on very high fields due to the fast scanning speed and low SAR level.

Thomas Benkert¹, Martin Blaimer¹, Peter M. Jakob², and Felix A. Breuer^1
1Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Bavaria, Germany, ²Experimental Physics 5, University of Würzburg, Würzburg, Bavaria, Germany

Inversion recovery (IR) techniques allow the suppression of one particular species (e.g. CSF fluid or fat). However, these techniques have their limitation. A single inversion time (TI) has to be chosen prospectively, resulting in only one contrast per IR experiment. Here, we demonstrate that employing a radial trajectory in combination with a conventional radial reconstruction allows the reconstruction of several images with different tissue types cancelled out of one singleshot dataset. Thus, no fixed TI is necessary and the desired contrasts can be chosen retrospectively, what makes this technique a promising candidate for clinical practice.

Eva Alonso Ortiz¹ and Gilbert Bruce Pike^2
1Department of Physics, Medical Physics Unit, McGill University, Montreal, QC, Canada, ²Department of Radiology and Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada

Multi-component T2 relaxation time measurement in the central nervous system shows a component that originates from water trapped in the lipid bilayers of myelin. This myelin-water component is of significant interest as it provides a myelin specific MRI signal of value in assessing myelin changes in cerebral white matter in vivo. In this educational e-poster the various acquisition and analysis strategies proposed to date for myelin water imaging are reviewed and research conducted into their validity and clinical applicability is presented. Comparisons between the imaging methods are made with a discussion regarding potential difficulties and model limitations.

Sandra M. Meyers¹, Shannon H. Kolind², and Alex L. MacKay^1,3
1Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, ²Medicine, University of British Columbia, Vancouver, BC, Canada,³Radiology, University of British Columbia, Vancouver, BC, Canada

Myelin water imaging based on multi-component T2 relaxation is valuable for investigating neurological diseases such as multiple sclerosis. Myelin water fraction correlates strongly with histological staining for myelin, but is also affected by changes in total water content; therefore, simultaneous measurement of the two could be a valuable tool for studying brain disease processes such as edema and demyelination. We present a T2 relaxation based technique in which myelin water fraction and water content are measured simultaneously at 3T. Water content showed excellent consistency and accuracy (mean error 1.8%) in phantom validations and in vivo values were consistent with literature.

Yanqin Lin¹, Liandi Zhang¹, Zhiliang Wei¹, Liangjie Lin¹, Shuhui Cai¹, and Zhong Chen^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China

The inherent heterogeneity of the samples or living organisms can lead to magnetic field fluctuations and losses of local homogeneity. Intermolecular multiple quantum coherences (iMQCs) can be used to obtain high resolution magnetic resonance spectroscopy (MRS) under field inhomogeneity. Since iMQC signals are phase-modulated, absolute value spectra are usually used for high resolution projection, which introduces unfavorable line broadening. Here, a pulse sequence is designed to simultaneously acquire intermolecular zero quantum coherence (iZQC) and intermolecular double quantum coherence (iDQC) signals. Double absorption lineshape is achieved through a combination of iZQC and iDQC signal, thus doubling the spectral resolution.

Haifeng Wang¹, Leo Tam¹, Emre Kopanoglu¹, Dana Peters¹, R. Todd Constable¹, and Gigi Galiana^1
1Department of Diagnostic Radiology, Yale University, New Haven, CT, United States

Recently, a developed approach to faster imaging has been spatial encoding with nonlinear magnetic fields, such as, PatLoc, O-Space, Null Space, 4D-RIO, etc. But, these efforts have focused on gradient echo imaging. Otherwise, Turbo Spin Echo (TSE) sequences provide much faster scan time than standard spin echo sequences. In this abstract, we demonstrate a hybrid of TSE and O-Space, developing an O-Space TSE sequence which is much faster than single echo O-Space imaging. Various techniques are applied to overcome the problems of artifacts and ambiguous T2 weighting. Simulations and experiment illustrate that the proposed method can inherit the advantages of both TSE and O-Space.

Peter Speier¹ and Michael S. Hansen^2
1Siemens AG Healthcare Sector, Erlangen, Germany, ²National Institutes of Health - NHLBI, MD, United States

Values for the Golden Angle depend on the symmetry of the problem. For radial imaging with symmetric echoes a value of 111.25° is typically used, but a value of 68.75°, corresponding to the adjacent angle, could be used as well. For GRE sequences with radial trajectories instead of Cartesian trajectories, the efficiency of in-plane gradient spoilers is reduced due to the non-constant spoiling direction; spoiling efficiency decreases with increasing projection advance angle. We conducted phantom measurements to demonstrate that artifacts from incomplete spoiling can be reduced by selecting the smaller Golden Angle value for the advance angle.

Zhiqiang Li¹, Dinghui Wang¹, Michael Schär^1,2, Nicholas R Zwart¹, and James G Pipe^1
1Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ, United States, ²Philips Healthcare, Cleveland, OH, United States

T1-weighted Cartesian SE imaging is sensitive to motion and flow artifacts. Several alternatives have been proposed including T1FLAIR PROPELLER and SE PROPELLER but each may have its limitations. Fat signal in T1-weighted brain imaging may be helpful or undesirable depending on the applications. But water/fat separation with current SE sequence results in long scan time. In this work we propose a SE technique employing a spiral readout with multiple TE shifts. The preliminary results demonstrate its insensitivity to flow artifacts, the capability for simultaneous water/fat separation, and the flexibility in scan time reduction.

Martin Blaimer¹, Daniel Neumann¹, Michael Völker¹, Peter M. Jakob¹, and Felix A. Breuer^1
1Research Center Magnetic-Resonance-Bavaria (MRB), Würzburg, Bavaria, Germany

Radial turbo spin-echo sequences (RARE, TSE, FSE) allow for an efficient acquisition of multi-contrast images. By applying the k-space weighted image contrast (KWIC) filter, images with the clinically important proton-density (PD) and T2 contrast are obtained from a single radial TSE acquisition. The purpose of this work is to extend this method towards an efficient suppression of the cerebrospinal fluid (CSF). This is accomplished by Fourier-analysis of the multi-contrast image series. In that way, CSF suppressed images with T2 contrast can be obtained in addition to the standard PD and T2 weighted images from a single radial TSE measurement.

Hans Weber¹, Jakob Assländer¹, Sebastian Littin¹, Jürgen Hennig¹, and Maxim Zaitsev^1
1Department of Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany

The ExLoc technique allows for excitation and geometrically matched local encoding of curved slices based on set of nonlinear magnetic encoding fields. For robust sequences such as gradient echo, the variation in image resolution resulting from the field nonlinearity can be compensated by the local FOV technique. In this study we explore the applicability of the local FOV technique to ExLoc echo planar imaging. ExLoc echo planar imaging is in particular promising for curved slice functional imaging. However, the underlying continuous trajectory is highly sensitive to B0 inhomogeneities.

Saikat Sengupta^1,2, David Smith^1,2, and E. Brian Welch^1,2
1Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, United States, ²Vanderbilt University Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States

Continuous moving bed MRI (COMBI) is a high throughput imaging technique for rapid whole body examination. In this abstract, golden angle increment (111.25o azimuthal step) based radial COMBI is introduced and compared to previously reported linear angle COMBI. The benefit of Golden angle sampling over a linear sampling scheme in retrospective profile binning for arbitrary slice thickness reconstruction is highlighted. We reconstruct gradient echo images acquired with GA sampling with 8, 10 (undersampled) and 13.4 mm (fully sampled) slice thickness and compare image quality with fully sampled, 13.4 mm linear angle radial reconstruction.

Melisa Okanovic^1,2, Martin Blaimer², Felix Breuer², and Peter Michael Jakob^2,3
1Comprehensive Heart Failure Center, University Hospital of Wuerzburg, Wuerzburg, Bavaria, Germany, ²Magnetic Resonance Bavaria (MRB), Wuerzburg, Bavaria, Germany, ³Department of Experimental Physics 5, University of Wuerzburg, Wuerzburg, Bavaria, Germany

In this work a helical bSSFP-sequence with a golden angle acquisition for abdominal imaging is presented. The technique allows continuously shifting the image plane along the slice axis. By this a continuous 3D volume is covered. Using sliding-window reconstruction slices of arbitrary positions can be obtained. The in-vivo applications on abdominal imaging during free breathing are demonstrated.

Ali Caglar Özen¹, Ute Ludwig¹, and Michael Bock^1
1Radiology - Medical Physics, University Medical Center Freiburg, Freiburg, Germany

Single Point Imaging (SPI) is a pure phase encoding sequence and it requires long scan times and it has high level gradients which are active during RF excitation. In this work, a novel k-space ordering scheme is proposed which results in less heating of gradient coils. Additionally, the sequence is modified such as the gradient levels are limited during RF excitation which provides a smoother excitation profile while preserving the low acoustic noise levels. The idea of limiting gradient amplitudes during RF excitation can be applied to other projection reconstruction sequences.

Rolf F Schulte¹ and Ralph Noeske^2
1GE Global Research, Munich, Germany, ²ASL Europe, GE Healthcare, Berlin, Germany

Peripheral nerve stimulation (PNS) is often a more limiting factor on modern MRI scanners than maximum gradient strength and slewrate. Typically, the slewrate is derated globally to adhere to PNS limitations. In this work, the PNS limitation is already included in the gradient waveform design in form of a time varying slewrate, hence shortening the overall gradient duration. The concept is demonstrated on exemplary spiral and EPI gradient waveforms.

Haifeng Wang¹, Leo Tam¹, R. Todd Constable¹, and Gigi Galiana^1
1Department of Diagnostic Radiology, Yale University, New Haven, CT, United States

Spatial encoding with nonlinear magnetic fields has been studied recently, such as, O-Space imaging, Null Space imaging, PatLoc, 4D-RIO, etc. Previous nonlinear gradient encoding works have explored single shot trajectories in nonlinear magnetic fields, where the nonlinear gradients are just aiding the linear encoding. In this abstract, a scheme of fast rotary nonlinear spatial encoding, named as FRONSAC imaging, is proposed to improve highly under-sampled data obtained with linear gradients by adding nonlinear gradients with low amplitudes and fast oscillations. The linear gradients are applied on the three standard linear encoding fields and the fast rotary gradients are applied on two second-order encoding fields. Images are reconstructed by Kaczmarz algorithm. The simulation results illustrate the proposed scheme can greatly aid many highly under-sampled linear trajectories, and improve image quality in accelerated data acquisitions.

Cheng Guan Koay^1
1Department of Medical Physics, University of Wisconsin-Madison, Madison, WI, United States

Uniformity in sampling measurements is a well known guiding principle in acquisition design for 3D radial MRI. Here, we argue that further improvement in image quality can be achieved by taking antipodal symmetry (or Hermitian symmetry) into account in the sampling design process. Herein, we developed a technique capable of generating very large, uniformly distributed points on the unit sphere that are endowed with antipodal symmetry. We observed a reduction of percent relative errors with respect to the ground truth, which was derived from a realistic numerical phantom, of about 10% as compared to the commonly used scheme.

Evan Levine^1,2, Manojkumar Saranathan², and Brian Hargreaves^2
1Electrical Engineering, Stanford University, Palo Alto, CA, United States, ²Radiology, Stanford University, Palo Alto, CA, United States

Many 3D dynamic MRI sequences use stochastic k-t sampling trajectories with elliptical-centrically sampled temporal phases. Poisson-disc k-space sampling is commonly used to balance sampling incoherence for compressed sensing (CS) and uniformity for g-factor in parallel imaging. Other temporally-uniform stochastic sampling patterns are used in approaches such as DISCO and allow view-sharing, where temporal phases are complementary and are used to reconstruct composite images. To allow both view-sharing and CS reconstruction of dynamic phases, we propose a method for generating sampling patterns, each with complementary Poisson-disc sample distributions.

Larry Hernandez¹, Pablo Irarrazaval², Kevin M Johnson¹, and Walter F Block^1,3
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ²Ingenieria Electrica, Pontificia Universidad Catolica de Chile, Santiago, Chile, ³Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States

Out and back 3D radial trajectories have successfully produced high resolution clinical imaging of the breast and knee when tight TR constraints are present, such as in bSSFP and/or fat/water separation methods where small echo spacing is crucial. We investigate replacing the bipolar gradients used in out and back trajectories with a 3D, single petal, rosette-like trajectory to cover k-space more efficiently while maintaining a tight TR constraint. Termed iCones due to the similarity of the trajectory to an ice cream cone, the number of excitations necessary to cover an equivalent k-space volume can be reduced by up to 1/3.

Tokunori Kimura¹, Mitsukazu Kamata¹, and Takashi Shigeta^1
1MRI Systems Development Department, Toshiba Medical Systems, Otawara, Tochigi, Japan

We proposed and assessed a new partial Fourier technique named RealAFI where the phase polarity can be preserved for inversion-prepared and Spin-Echo based imaging. 1D simulation and 3T double IR (DIR) proton density black-blood FSE (DIR-BB-FSE) with TI of shorter than nulled blood Mz were assesed here. POCS method for RealAFI was used modifying phase estimation; i.e., the back-ground phase was calculated from circular low-pass filtered data in k-space region that is smaller than the actually sampled. RealAFI provided almost comprable results as those from fully sampled data. RealAFI is practically useful technique for inversion-prepared spin-echo based imaging.

Luning Wang¹, Wei Tang², Zipeng Zhen², Hongming Chen², Jin Xie², and Qun Zhao^1
1Department of Physics and Astronomy, University of Georgia, Athens, Georgia, United States, ²Department of Chemistry, University of Georgia, Athens, Georgia, United States

Iron oxide nanoparticles (IONPs) have been widely used as a T2/T2* contrast agent in MRI. To avoid signal loss caused by IONPs, pulse sequences, such as Ultrashort TE (UTE) and Sweep Imaging with Fourier Transform (SWIFT), can be implemented to acquire MR signals of IONPs, since the echo times of these sequences are in a few microseconds. But in the UTE and SWIFT images, long T2 tissues and fat may also appear bright, posing a challenge to discriminate IONPs from surrounding tissues. In this work, IONPs were used to target tumor cells grafted in mice. In order to improve the detection specificity of the IONPs delivered to tumors, we hypothesized to embed saturation pulses into the SWIFT sequence to suppress long T2 tissues and fat.

Michael Tesch¹, Steen Moeller¹, Ryan Chamberlain¹, and Michael Garwood^1
1CMRR, University of Minnesota, Minneapolis, MN, United States

 

Martin Ott¹, Felix Breuer¹, David Grodzki², Martin Blaimer¹, Simon Triphan^1,3, and Peter Jakob^1,4
1MRB Forschungszentrum für Magnet-Resonanz-Bayern e.V., Würzburg, Bavaria, Germany, ²Siemens AG, Erlangen, Bavaria, Germany, ³Diagnostic and Interventional Radiology, University Hospital Heidelberg, Heidelberg, Germany, ⁴Department of Experimental Physics 5, University of Wuerzburg, Würzburg, Germany

In this work, the acoustic-noise of gradient echo (GRE) sequences is addressed. Thin slice thicknesses or short echo times usually require fast switching gradients which lead to high acoustic-noise levels. The focus is on acoustic-noise reduction of the data acquisition for images with these boundary conditions. Therefore, a prototype sequence is implemented which makes use of half pulse excitations as well as a novel k-space sampling scheme. First in-vivo results are presented as well as a significant reduction of acoustic-noise.

Florian Wiesinger¹, Laura Sacolick², Sandeep Kaushik³, Sangtae Ahn⁴, Gaspar Delso⁵, and Dattesh Shanbhag^3
1GE Global Research, Munich, Germany, ²GE Healthcare, Munich, Germany, ³GE Global Research, Bangalore, India, ⁴GE Global Research, Niskayuna, NY, United States, ⁵GE Healthcare, Zurich, Switzerland

In this abstract we investigate the feasibility of zero TE imaging for depiction of cortical bone structures in the head and pelvis. Proton density contrast in combination with logarithmic inverse scaling is used to highlight bone structures and differentiate them from soft tissues and background air. Different from prior art, no long T2 suppression methods (like echo subtraction or saturation prepulses) are required, rendering this method fast, robust and effective. In-vivo volunteer experiments indicate excellent 3D cortical bone depiction as required for instance for PET/MR attenuation correction and MR-based radiation therapy planning.

Markus Weiger¹, David Otto Brunner¹, Michael Wyss¹, Benjamin Emanuel Dietrich¹, Bertram Jakob Wilm¹, and Klaas Paul Pruessmann^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

Zero echo time (ZTE) imaging enables efficient direct MRI of tissues with very rapid transverse relaxation. In addition, this intrinsically three-dimensional technique is also very fast, has high robustness against eddy current and off-resonance effects, and the acoustic noise of the sequence is close to ambient level. All these beneficial properties make ZTE imaging also attractive beyond short-T2 applications. However, due to zero TE and usually small flip angles, intrinsic contrast mechanisms are somewhat limited. Therefore, in the present work, magnetisation preparation is used to create T1 contrast in ZTE images.

Jiang Du¹, Shihong Li¹, Eric Chang¹, Christine B Chung¹, and Graeme Bydder^1
1Radiology, University of California, San Diego, San Diego, CA, United States

Bone water occurs at various locations and in different binding states. In normal bone the majority of bone water is loosely bound to the organic matrix. There is also a significant amount of free water residing in the pores of bone which is responsible for nutrient diffusion and contributes to the viscoelastic properties of cortical bone. Separation of bound water from free water is of critical importance since the two make different contributions to the mechanical properties of bone. In this study we aimed to investigate the effects of inversion time (TI) on bone free water signal using an adiabatic inversion recovery prepared ultrashort echo time (IR-UTE) sequence on a clinical whole body 3T scanner.

Ute Ariane Ludwig¹, Jan-Bernd Hövener^1,2, Ali Özen¹, Lena Öhrström³, Andreas Bitzer⁴, Markus Walther⁴, Dominik Elverfeldt¹, Frank Rühli³, and Michael Bock^1
1Radiology - Medical Physics, University Medical Center Freiburg, Freiburg, Germany, ²German Consortium for Cancer Research (DKTK), German Cancer Research Center (DKFZ), Heidelberg, Germany, ³Centre for Evolutionary Medicine, Institute of Anatomy, University of Zürich, Zürich, Switzerland, ⁴Molecular and Optical Physics, University of Freiburg, Freiburg, Germany

Imaging of dehydrated objects is usually performed with CT due to the low water content. With the development of MR sequences with short or even zero echo times, MR is also able to detect tissues with fast decaying signal like bones or teeth. In this work the potential of the ZTE technique has been evaluated on an ancient Egyptian mummified fish and compared to other imaging modalities like CT and Terahertz imaging.

Anna-Katinka Bracher¹, Erich Hell², Johannes Ulrici², and Volker Rasche^1
1Internal Medicine II, University Hospital of Ulm, Ulm, BW, Germany, ²Sirona Dental Systems, Bensheim, HE, Germany

Ultra short echo time (UTE) imaging has proven its potential for radiation free detection of caries lesions and also for classification of lesion progression using ultra short T2* mapping. The UTE imaging sequence is very sensitive to gradient delay related k-space center shifts and off-resonance effects. C APRI allows automatic correction of gradient delays, reduces off-resonance and T2 induced blurring artefacts and provides T2* mapping for lesion classification.

Karl-Heinz Herrmann¹, Martin Krämer¹, Martin Stenzel², Hans-Joachim Mentzel², and Jürgen R Reichenbach^1
1Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany, ²Pediatric Radiology, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany

High resolution (0.55m isotropic) 3D radial center-out ultra-short echo time (UTE) MRI with interspersed fat saturation pulses is acquired for a case of a skull bone tumor and compared to CT images. In the presented case the tumor has not yet infiltrated into the soft tissue and the thin separating layer of bone can clearly be identified on both modalities. UTE images additionally show the integrity of the meninges, which are a short T2 species not visible at regular echo times or in CT. This suggests that UTE MRI might be a valuable additional modality in bone tumor diagnosis.

Karl-Heinz Herrmann¹, Martin Krämer¹, and Jürgen R. Reichenbach^1
1Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany

For spoiled gradient echo 3D radial center out UTE sequences standard, predictable RF and gradient spoiling can only be achieved by constant gradient moments every TR, which requires center-out-in trajectories to remove the direction dependency. The center-in trajectory provides acquisition of a second echo, but for many application the second echo is not required. Naively using center-out trajectories combined with RF spoiling and constant spoiler gradients in fixed direction can lead to stripe artifacts. By extending the duration of the rotated readouts by a variable, direction dependent the spoiling is optimized while the TR is simultaneously minimized for fast radial acquisition.

Joshua D. Trzasko¹, Daniel C. Mellema¹, Armando Manduca¹, Scott A. Kruse¹, Adil E. Bharucha¹, Kiaran P. McGee¹, and Philip A. Araoz^1
1Mayo Clinic, Rochester, MN, United States

In steady-state magnetic resonance elastography (MRE), quantitative tissue stiffness maps are constructed using the first temporal harmonic of a phase contrast image series. The accuracy of a stiffness map intrinsically depends on the estimated harmonic from which it is derived. Recently, a robust statistical framework for harmonic estimation was proposed, but the associated optimization strategy for performing the estimation was limited to phase-wrap free data. In this work, we describe a novel complex graph cut optimization strategy for harmonic estimation that can operate effectively in the presence of wrapping, and discuss several unique computational challenges associated with this problem.

Ruth J. Okamoto¹, Curtis L. Johnson², Matthew D. McGarry³, Andrew A. Badachhape⁴, Bradley P. Sutton², John G. Georgiadis², and Philip V. Bayly^1
1Mechanical Engineering and Materials Science, Washington University, Saint Louis, Missouri, United States, ²Beckman Institute for Advanced Science and Technology, University of Illinois, Urbana-Champaign, Illinois, United States, ³Dartmouth College, Hanover, New Hampshire, United States,⁴Biomedical Engineering, Washington University, Saint Louis, Missouri, United States

Three approaches, local frequency estimation (LFE), local direct inversion (LDI), and nonlinear inversion (NLI), were used to estimate viscoelastic parameters of the human brain from high-resolution magnetic resonance elastography (MRE) measurements. While each inversion method has been independently validated using gel phantoms, the complicated structure of the human head (comprising skull, meninges, CSF, and brain) and the complex material behavior of the brain (anisotropic, heterogeneous, and poroelastic) challenge the underlying assumptions of each method. Differences were found between average shear modulus estimates obtained by the three methods (NLI>LFE>LDI), however the relative differences between individuals were consistent among all methods.

Cemre Ariyurek^1,2, Yusuf Ziya Ider¹, Necip Gurler¹, Safa Ozdemir¹, Alp Emek¹, Arif Sanli Ergun³, and Ergin Atalar^1,2
1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey, ²National Magnetic Resonance Research Center (UMRAM), Ankara, Turkey, ³Department of Electrical and Electronics Engineering, TOBB-University of Economics and Technology, Ankara, Turkey

In this study, modes of shear waves that may form during MR elastography of brain are investigated. Results of eigenfrequency simulations indicate that there are eigenmodes that may be excited by actuators at correct frequencies. In the results of frequency domain simulations and human experiments peak displacement(s) are observed at certain frequencies, having similar shear wave patterns to eigenmodes. A shift in the frequency of the peak displacement is observed in simulations when stiffness value of brain is altered. Thus, this method can be used for diagnosing diseases altering stiffness by detecting a frequency shift in peak displacement.

Allen Q. Ye¹, Temel K. Yasar², Altaf Khan², Ziying Yin¹, Dieter Klatt¹, Thomas J. Royston¹, and Richard L. Magin^1
1Department of Bioengineering, University of Illinois at Chicago, Chicago, IL, United States, ²Department of Mechanical Engineering, University of Illinois at Chicago, Chicago, IL, United States

In this study we have utilized SampLing Interval Modulation (SLIM)-MRE on an in-vivo murine experiment. SLIM-MRE enables acquisition of displacement information in multiple directions without increasing total acquisition time. This enables us to acquire MRE images in one-third of the time compared to conventional MRE, which reduces the health problems that might arise due to prolonged anesthesia on mice and also reduces the image misregistration artifacts that might originate from motion of the internal organs such as the bladder and intestines throughout time.

Houchun Harry Hu¹, Tai-Wei Wu², Larry Yin³, Mimi S. Kim³, Jonathan M. Chia⁴, Thomas G. Perkins⁴, and Vicente Gilsanz^1
1Radiology, Children's Hospital Los Angeles, Los Angeles, California, United States, ²Neonatology, Children's Hospital Los Angeles, Los Angeles, California, United States, ³Pediatrics, Children's Hospital Los Angeles, Los Angeles, California, United States, ⁴Philips Healthcare, Cleveland, Ohio, United States

We report the observation of brown adipose tissue with low fat content in newborns with hypoxic-ischemic encephalopathy (HIE) after hypothermia therapy treatment in the neonatal intensive care unit. Ten HIE newborns were studied at 3T using a six-echo chemical-shift-encoded water-fat sequence. Fat-signal fraction (FF) measurements of the supraclavicular brown adipose tissue depot were compared to values from five non-HIE newborns. The average FF range was 10.3-29.9% and 23.7-42.2% for the HIE and non-HIE group, respectively (p<0.01). We speculate that hypothermia therapy in HIE neonates promotes BAT-mediated non-shivering thermogenesis, which subsequently leads to a depletion of the tissue’s intracellular fat stores.

Evan Levine^1,2, Brady Quist^1,2, Bruce Daniel², Brian Hargreaves², and Manojkumar Saranathan^2
1Electrical Engineering, Stanford University, Palo Alto, CA, United States, ²Radiology, Stanford University, Palo Alto, CA, United States

High spatiotemporal resolution multiphasic DCE-MRI entails tradeoffs of spatial and temporal resolution. To address the tradeoff, view-sharing schemes have been proposed, some of which use stochastic k-space trajectories. Compressed sensing offers a way to reduce view-sharing and temporal footprint. We propose a novel method of integrating view-sharing, compressed sensing, parallel imaging, and two-point Dixon based fat water separation that allows new tradeoffs in spatial and temporal resolution. A combination of CS, parallel imaging, and partial view-sharing is a suitable alternative to conventional view-sharing, especially when restricted to the water image using a two-point Dixon model.

Noam Ben-Eliezer¹, Li Feng¹, Kai Tobias Block¹, Daniel K. Sodickson¹, and Ricardo Otazo^1
1Department of Radiology, New York University School of Medicine, Bernard and Irene Schwartz Center for Biomedical Imaging, New York, NY, United States

Accurate in vivo estimation of T₂ relaxation values with high spatial resolution is very challenging in clinical settings. This work, presents the fusion of a recently-developed T₂ mapping technique – the echo-modulation curve (EMC) algorithm – with compressed sensing and model-based reconstruction using undersampled radial trajectories. The EMC approach avoids the common challenges associated with stimulated echoes in multi spin-echo protocols, while compressed sensing removes undersampling-related aliasing artifacts. The synergistic combination of the scanner-invariant EMC algorithm with a folding-free radial sampling scheme offers accelerated quantification of functional (T₂) and morphological (proton-density) information at sub-millimeter spatial resolutions and with reduced sensitivity to motion.

Dongwook Lee¹, Eung Yeop Kim², Huisu Yoon¹, Sunghong Park¹, and Jong Chul Ye^1
1Korea Advanced Institute of Science and Technology, Daejeon, Korea, ²Gachon University Gil Medical Center, Incheon, Korea

A compressed sensing algorithm is applied to T2 prime mapping which is useful for diagnosis of cerebrovascular disorders such as occlusive carotid disease and ischemic stroke. The quantitative measure of parameter maps with high resolution requires the acquisition of multiple images. That is usually associated with long acquisition times. To reduce the acquisition time of MR data, we use a compressed sensing algorithm using patch based low rank penalty for the reconstruction of T2 and T2-star weighted images to obtain the T2 prime map.

Dimitrios C Karampinos¹, Stefan Ruschke¹, Holger Eggers², Marcus Settles¹, Hendrik Kooijman³, Peter Börnert², Ernst J Rummeny¹, and Thomas Baum^1
1Department of Diagnostic and Interventional Radiology, Technische Universität München, Munich, Germany, ²Philips Research Laboratory, Hamburg, Germany, ³Philips Healthcare, Hamburg, Germany

It is known that T₂* decay effects can induce significant bias in the estimation of the fat fraction using a chemical shift based water-fat separation methods, especially in the presence of short T₂*. However, the need for T₂* correction has not been previously characterized in skeletal muscle fat quantification, where T₂* decay is moderate but susceptibility-induced fat resonance shift effects are also present. The present study shows the existence of a synergistic effect between T₂* decay and susceptibility shift on the estimation of the fat fraction. Single T₂* correction reduces fat fraction bias due to the synergistic effect of T₂* decay and susceptibility shift, independent of the choice of TEs.

Daniel C Mellema¹, Joshua D Trzasko¹, and Armando Manduca^1
1Mayo Clinic, Rochester, Minnesota, United States

Quantitative measurements of T1 values have been used to monitor pathology. Recently a maximum likelihood estimator was created to obtain the best estimate of T1 values in the presence of noise, with no a priori assumption about image structure, for a variable flip angle spoiled gradient-recalled echo sequence. While this improved estimations, individual voxel estimates exhibited high variance. It is hypothesized that adding a spatial prior promoting piecewise smoothness will improve estimations. Since standard continuous optimization techniques are inefficient and unstable for this generalized nonlinear least squares problem, an iterative graph cut strategy was developed for the regularized T1-estimation.

Nathalie Just¹, Olivier Reynaud², and Rolf Gruetter^1,3
1CIBM-AIT, EPFL, Lausanne, Switzerland, ²Laboratory for functional and metabolic Imaging, EPFL, Lausanne, Switzerland, ³Departments of Radiology, University of Lausanne and University of Geneva, Lausanne and Geneva, Switzerland

With the growing number of transgenic animal models of diseases, it is of paramount importance to develop high spatial resolution T1-weighted imaging techniques for structural imaging at ultra high fields. In the present work, the MP2RAGE technique was implemented at 14.1T and used to acquire T1-weighted images of a MnCl2-doped phantom and knockout mice. T1 mapping was also performed. The high translational value of MP2RAGE sequences was demonstrated.

Eva Alonso Ortiz¹ and Gilbert Bruce Pike^2
1Department of Physics, Medical Physics Unit, McGill University, Montreal, QC, Canada, ²Department of Radiology and Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada

Se-Hong Oh^1,2, Sung Suk Oh², Joon Yul Choi², Jang-Yeon Park³, and Jongho Lee^2
1Imaging Institute, Cleveland Clinic, Cleveland, Ohio, United States, ²Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, United States, ³School of Biomedical Engineering, College of Biomedical and Health Science, Konkuk University, Chungju, Korea

In this study, a new 3D myelin water imaging sequence that covers the entire brain (FOV = 240 ¡¿ 240 ¡¿ 128 mm^3; resolution = 1.5 ¡¿ 1.5 ¡¿ 4 mm^3) in less than 8 minutes was developed. This method is based on a recently proposed novel approach to MWI, called Direct Visualization of Short Transverse Relaxation Time Component (ViSTa). It provides a high quality myelin water fraction map.

Anna Combes¹, Tobias C Wood¹, Gareth J Barker¹, and Steven C R Williams^1
1Neuroimaging, King's College London, Institute of Psychiatry, London, England, United Kingdom

We present the results of a reproducibility study of quantitative T1 and T2 mapping using DESPOT at 3.0T.

Ville Renvall^1,2, Thomas Witzel^1,2, Lawrence L. Wald^1,2, and Jonathan R. Polimeni^1,2
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, ²Department of Radiology, Harvard Medical School, Boston, MA, United States

Fast T₁ mapping method using permuted slice acquisition order and TI skipping on an IR EPI scan was developed. The resulting T₁ and T₂*/ spin density images were used to synthesize inversion recovery EPI data with an arbitrary, selectable TI. Automatic brain segmentation and cortical surface reconstruction of the synthesized data using FreeSurfer were successful and could serve as basis for distortion matched anatomical reference for functional MRI.

Jinseong Jang¹, Tae-Joon Eo¹, Narae Choi¹, Minoh Kim¹, Dongyeob Han¹, Dong-Hyun Kim¹, and Dosik Hwang^1
1School of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea

Magnetic resonance fingerprinting is a method that can quantitatively estimate MR parameters such as T1, T2 of specific tissues, by matching pattern of signal evolution obtained from the scanner with the pattern of signal evolution that is generated from MR forward modelling. the well-accepted Cartesian trajectory needs to be considered for robust implementation of MRF for fast processing In this study, efficient iterative compressed sensing (CS) reconstruction method is proposed to highly accelerate the Cartesian-trajectory based acquisition for MRF, leading to the reduction factor up to 16.

Ross A Little¹, Geoff J M Parker¹, and Chris Rose^1
1Centre for Imaging Sciences, University of Manchester, Manchester, United Kingdom

We present a novel method for directly measuring morphological parameters from highly undersampled data without need for image reconstruction. We introduce the theory underpinning the approach and demonstrate it using synthetic and physical phantoms. The physical phantom measurements showed excellent agreement with ground truth (R²>0.999). The mean measurement bias was less than one voxel. We conclude that SPI is a method with the potential to allow quantitative measurements of interest to be made from highly undersampled k-space data without need for explicit image reconstruction.

Ouri Cohen¹, Brandon D. Armstrong^1,2, Mathieu Sarracanie^1,2, Christian T. Farrar¹, Jerome L. Ackerman¹, and Matthew S. Rosen^1,2
1Department of Radiology, MGH/Athinoula A. Martinos Center for Biomedical Imaging, Massachussets General Hospital, Charlestown, MA, United States,²Department of Physics, Harvard University, Cambridge, MA, United States

We demonstrate an application of Magnetic Resonance fingerprinting at 15T optimized to utilize only 10 measurements to obtain T1,T2,PD,B0 and B1 maps in an in vivo healthy mouse brain.

Tae-joon Eo¹, Jinseong Jang¹, Minoh Kim¹, Narae Choi¹, Dongyeob Han¹, Dong-hyun Kim¹, and Dosik Hwang^1
1School of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea

This study demonstrates that the accurate parameter maps can be obtained from extremely undersampled Cartesian trajectory data in MRF by applying the proposed data sharing method.

Debra F. McGivney¹, Dan Ma², Haris Saybasili³, Yun Jiang², and Mark A. Griswold^1,2
1Radiology, Case Western Reserve University, Cleveland, Ohio, United States, ²Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States, ³Siemens Healthcare, Chicago, Illinois, United States

Magnetic resonance fingerprinting is a technique that can provide quantitative maps of tissue parameters (T1, T2, and off-resonance) through matching observed signals to a precomputed dictionary of modeled signal evolutions. To retrieve the parameters, the inner product between the signal and each dictionary entry is computed to find the entry corresponding to the maximum. We propose to compress the size of the dictionary and observed signals in the time domain by applying the singular value decomposition (SVD), thereby reducing the number of computations required while retaining the most relevant information from the dictionary.

Sarah Thiesson¹ and Richard B Thompson^1
1Biomedical Engineering, University of Alberta, Edmonton, Alberta, Canada

A new approach for simultaneous quantification of water T1 and T2, fat fraction (FF) and off-resonance frequency is proposed using a saturation-recovery prepared phase sweep SSFP method. The signal intensity profile in each pixel, as a function of RF phase increment (conventional SSPF has a constant 180° increment), is fit as a weighted sum of simulated fat and water basis sets (Bloch equations) to derive T1, T2, FF and off-resonance. This method was validated with Monte Carlo simulations (to characterize accuracy and precession), T1/T2 and fat phantoms, and in-vivo with characterization of tibialis, gastroc and soleus skeletal muscle groups

Mathieu Sarracanie^1,2, Brandon D. Armstrong^1,2, and Matthew S. Rosen^1,2
1Department of Radiology, MGH/Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ²Department of Physics, Harvard University, Cambridge, MA, United States

We have demonstrated MR Fingerprinting at low magnetic field, which results in simultaneous measurement of 4 quantitative parameters, and thus provides 4 different image contrasts in a single acquisition (proton density, T₁, T₂ and off-resonance) in less than 15 minutes. This technique is of particular relevance at low magnetic field where SNR and contrast are tied to long acquisition times. The combination of MRF with low field MRI scanners has great potential to revolutionize future transportable MRI systems.

Yun Jiang¹, Dan Ma¹, Nicole Seiberlich¹, Vikas Gulani², and Mark A. Griswold^1,2
1Department of Biomedical Enginneering, Case Western Reserve University, Cleveland, OH, United States, ²Department of Radiology, Case Western Reserve University, Cleveland, OH, United States

Synopsis: MR Fingerprinting (MRF) has been shown to be extremely efficient in generating multiple parametric maps. Here we explore MRF using a FISP-based sequence with a spiral trajectory to generate T₁, T₂, and M₀ maps simultaneously. The result shows that high quality quantitative relaxation parameters can be estimated using the proposed method. With the introduction of the additional unbalanced gradients, the sequence is insensitive to inhomogeneities in the main magnetic field. This method has its potential to expand the areas of application of MRF, and extend the quantification to other MR parameters.

Eric Z.C Wu¹, Maxwell L Wong², and Eric C Wong^3
1University of Southern California, Los Angeles, California, United States, ²UCSD, San Diego, California, United States, ³Department of Radiology and Psychiatry, UCSD, San Diego, California, United States

In this study, we simulate the effects of motion in MR Fingerprinting. We established that motion related artifacts are sensitive to the time at which motion occurs in the scan, and that T1 maps are more sensitive to the position of the subject near the beginning of the scan, likely due to T1 sensitivity from the initial inversion pulse.

Alex Menys¹, Valentin Hamy¹, Caroline Hoad², Jesica Makanyanga¹, Freddy Odille³, Penny Gowland², Stuart A Taylor¹, and David Atkinson^1
1UCL, London, United Kingdom, ²Sir Peter Mansfield MRI Centre, University of Nottingham, Nottingham, United Kingdom, ³INSERM, Nancy, France

At present, registration-based quantification of bowel motility from dynamic MR is limited to breath-hold studies. Here we validate a dual-registration technique robust to respiratory motion for the assessment of small bowel and colonic motility.

Dattesh D Shanbhag¹, Venkata V Chebrolu¹, Sandeep N Gupta², Patrice Hervo³, and Rakesh Mullick^4
1Medical Image Analysis Laboratory, GE Global Research, Bangalore, Karnataka, India, ²Clinical Systems and Signal Processing, GE Global Research, Niskayuna, NY, United States, ³GE Healthcare, Buc, France, ⁴Diagnostics and Biomedical Technologies, GE Global Research, Bangalore, Karnataka, India

Local correlation and dispersion based measures have been introduced to reliably reflect the improvement in dynamic 4D MRI data tissue alignment post motion correction. These measures can be used as part of motion correction workflow and provide means for quantifying the efficacy of motion correction schemes in dynamic data across different anatomies and clinical sites.

Mauro Zucchelli¹, Eleftherios Garyfallidis², Michael Paquette², Sylvain Merlet³, Gloria Menegaz¹, and Maxime Descoteaux^2
1Department of Computer Science, University of Verona, Verona, Italy, ²Sherbrooke Connectivity Imaging Lab (SCIL), Université de Sherbrooke, Sherbrooke, Quebec, Canada, ³INRIA Sophia Antipolis-Méditerranée, Sophia Antipolis, France

DSI was one of the first techniques used to infer complex fiber configurations. However, DSI discrete propagator representation suffers from a limited frequency band, which makes infinite integration impossible. This abstract aims at comparing DSI propagator derived indices with those obtained on a continuous analytical model like SHORE. The models were tested both on simulated and human data on a Cartesian grid sampling scheme. The indices used for the comparison were the orientation distribution function, the return to the origin probability and the mean square displacement. Results indicate that the 3D-SHORE provides an improved estimation of the signal properties.

Di Xu¹, Haris Saybasili², Aravindan Kolandaivelu³, Henry Halperin^1,3, Menekhem M. Zviman³, Mark A. Griswold^4,5, Nicole Seiberlich⁵, and Daniel A. Herzka^1
1Department of Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, MD, United States, ²Siemens Healthcare USA, Inc., Chicago, IL, United States, ³Department of Medicine, Division of Cardiology, Johns Hopkins School of Medicine, Baltimore, MD, United States, ⁴Department of Radiology, Case Western Reserve University, Cleveland, OH, United States, ⁵Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States

T₂-weighted (T₂W) images have widespread application in MRI and are particularly valued in interventional MRI. However, most existing T₂W imaging techniques are not applicable to interventional imaging and are much slower than other competing imaging modalities used in interventional guidance. We present T₂-weighted radial interrupted balanced SSFP (T₂W-riSSFP), a technique allows high speed real-time imaging as well as edema visualization with significant T₂ contrast. This technique was evaluated on phantoms, normal human subjects, and animal model with acute myocardial infarction and edema. T₂W-riSSFP can be applied to real-time interventional guidance where heavily T₂-weighted images are needed.

Keigo Kawaji¹, Sebastien Roujol¹, Warren J Manning^1,2, and Reza Nezafat^1
1Medicine, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States, ²Radiology, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States

A novel coronary imaging method that allows cardiac motion assessment within the acquisition window of the k-space segmented volume has recently been proposed using 3D stack-of-stars (SOS) with golden-angle-derived interleaving for encoding temporal information within different segments of the acquisition window. However, non-Cartesian radial reconstructions are typically unsuitable for immediate visualization due to the computationally expensive gridding operation. This step may bottleneck the visualization and disturb the workflow for interactive clinical assessment. We have developed an accelerated reconstruction workflow that uses a graphic processing unit (GPU) for rapid gridding of the acquired radial data, and examine its implementation and performance.

Florian Knoll¹, Andreas Schwarzl², Clemens Diwoky², and Daniel K Sodickson^1
1Bernard & Irene Schwartz Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, New York, United States,²Institute of Medical Engineering, Graz University of Technology, Graz, Styria, Austria

Image reconstruction for 3d non-Cartesian trajectories is still challenging due to expensive computation times. Parallel implementations using GPUs have evolved as a computationally effective way to tackle this problem. Current software packages are often restricted 2d imaging and are usually integrated in larger reconstruction frameworks. This makes re-use challenging especially when considering that the large majority of code for image reconstruction is written in Matlab. This work introduces gpuNUFFT, a new open-source 3d regridding GPU implementation with a built in Matlab interface that is straightforward to include in all implementations of iterative image reconstruction.

Souheil J Inati¹, Peter Kellman², Joseph Naegele¹, Sonia Nielles-Vallespin², Vinai Roopchansingh¹, Thomas S Sorensen³, Kaveh Vahedipour⁴, Hui Xue², Nicholas R Zwart⁵, and Michael S Hansen^2
1National Institute of Mental Health, National Institutes of Health, Bethesda, MD, United States, ²National Heart Lung and Blood Institute, National Institutes of Health, Bethesda, MD, United States, ³Department of Computer Science, Aarhus University, Aarhus, Denmark, ⁴Institute of Neurosciences and Medicine, Forschungszentrum Juelich, Juelich, Germany, ⁵Barrow Neurological Institute, Phoenix, AZ, United States

A common raw data format is a prerequisite for sharing MR image reconstruction algorithms and code, and is a necessary component of reproducible research. Ideally, this common format would be vendor neutral, and would capture the data fields needed to describe the details of the MR experiment so as to permit image reconstruction from the raw data. We propose the ISMRM Raw Data (ISMRMRD) format, which is described by an XML schema and several C-style structs. We demonstrate an implementation using HDF5 files for storage, along with C++, Python, MATLAB, and JAVA libraries for reading and writing ISMRMRD files.

Peter Adany¹, Phil Lee², Douglas R. Denney³, Sharon G. Lynch⁴, and In-Young Choi^4
1Hoglund Brain Imaging Center, University of Kansas Medical Center, Kansas City, KS, United States, ²Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, KS, United States, ³Department of Clinical Psychology, University of Kansas, Lawrence, KS, United States, ⁴Department of Neurology, University of Kansas Medical Center, Kansas City, KS, United States

Processing of region of interest (ROI) shapes for medical image analysis is often encumbered by the poor results of intensity interpolation when resampling axially from low to high resolution, e.g. when editing ROIs using original images with few slices prior to resampling to high resolution ROI shape images. A straightforward shape interpolation algorithm is proposed based on the Radon transform and filtered back projection. We investigate the capability of this algorithm to preserve fine image details and to transversally merge information between the adjacent slices. Results show interpolating characteristics greatly superior to intensity interpolation.

Yasheng Chen¹, Meher Juttukonda¹, Yi Su², Tammie Benzinger², Brian Rubin³, Yueh Z Lee¹, Felipe Espinoza⁴, Weili Lin¹, Dinggang Shen¹, David S Lalush⁵, and Hongyu An^1
1Radiology and BRIC, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States, ²Radiology, Washington University in St. Louis, St. Louis, MO, United States, ³Surgery and Radiology, Washington University in St. Louis, St. Louis, MO, United States, ⁴Radiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States, ⁵Biomedical Engineering, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States

An atlas based method has been developed to derive pseudo continuous CT images using MR T1w images for PET attenuation correction. Population probabilistic map for air space segmentation and a sparse regression to enhance the local structure similarity were employed to achieve an improved AC accuracy.

Jeiran Jahani¹, Glyn Johnson¹, and Kamiar Rahnama Rad^2
1Department of Radiology, Bernard and Irene Schwartz Center for Biomedical Imaging, New York University School of Medicine, New York City, New York, United States, ²Department of Statistics, Zicklin School of Business, City University of New York, New York City, New York, United States

denoise

Kenichi Ueno¹ and Kang Cheng^1
1RIKEN Brain Science Institute, Wako-shi, Saitama, Japan

MRI images can contain intensity inhomogeneity in space because of various reasons. Recent progress in MRI technologies such as multi reception/transmission of RF and ultra-high-field MRI systems has made this problem even more serious. Although many types of algorithms have been proposed and evaluated to deal with the problem, so far there has been no perfect solution. We developed a robust algorithm to correct intensity inhomogeneity of MRI signals without relying on additional scans. It was demonstrated that our newly proposed algorithm is very robust and corrects MRI images with inhomogeneous intensity satisfactorily.

Rezwan Ghassemi¹, Robert Brown¹, Brenda Banwell^2,3, Sridar Narayanan¹, Kunio Nakamura¹, and Douglas Arnold^1
1Montreal Neurological Institute, McGill University, Montreal, QC, Canada, ²The Children's Hospital of Philadelphia, Philadelphia, PA, United States,³Neurology, The Hospital for Sick Children, Toronto, ON, Canada

 

Dustin K Ragan¹ and Jose A Pineda^1
1Department of Pediatrics, Washington University School of Medicine, St. Louis, MO, United States

We developed a novel segmentation routine for determining the feeding vessels of the brain from gated phase-contrast MRI data. This has potential to simplify MR measurements of vascular flow and ICP.

Sébastien Tourbier^1,2, Xavier Bresson^1,2, Patric Hagmann², Maud Cagneaux³, Marie Schaer⁴, Laurent Guibaud³, Jean-Philippe Thiran^2,5, Reto Meuli², and Meritxell Bach Cuadra^1,2
1Centre d'Imagerie BioMédicale (CIBM), Lausanne, Vaud, Switzerland, ²Department of Radiology, University Hospital Center (CHUV) and University of Lausanne, Lausanne, Vaud, Switzerland, ³Department of Radiology, Hôpital Femme-Mère-Enfant (HFME), Lyon, Rhône, France, ⁴Department of Psychiatry, School of Medicine, University of Geneva, Geneva, Switzerland, ⁵Signal Processing Laboratory (LTS5), Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Vaud, Switzerland

In fetal brain MRI, most of the high-resolution reconstruction algorithms rely on brain segmentation as a preprocessing step. Manual brain segmentation is highly time-consuming and therefore not a realistic solution for large-scale studies. Only few works have addressed this automatic extraction problem. In this study, we assess the validity of Multiple Atlas Fusion (MAF) strategies to automatically segment the fetal brain in MR imaging. We show that MAF performance is increased in healthy brain by increasing the number of atlases. Secondly, we also show that MAF can be applied to pathological brains even when large anatomical differences are present.

Jason Su^1,2, Thomas Tourdias^2,3, Manojkumar Saranathan², and Brian Rutt^2
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States, ³Neuroradiology, Bordeaux University Hospital, Bordeaux, Drôme, France

Automatic segmentation of thalamic nuclei using a high resolution study-specific template and label fusion is surveyed in a pilot study of multiple sclerosis at 7T. The template is created from 17 subjects (6 controls, 11 patients) via ANTS and the white matter nulled MPRAGE contrast shown recently by Tourdias et al. The STEPS algorithm by Cardoso et al. is employed to segment a new patient using manual delineations of 6 controls. Performance in this single case is comparable to other published methods that require DTI. The Dice coefficients for whole thalamus (0.87), pulvinar (0.77), and mediodorsal nucleus (0.71) were notable.

Jiehua Li¹, Jason M Zara¹, Rao P Gullapalli², and Jiachen Zhuo^2
1Electrical and Computer Engineering, George Washington University, Washington, District of Columbia, United States, ²Diagnostic Radiology and Nuclear Medicine, Universit of Maryland School of Medicine, Baltimore, Maryland, United States

There has been growing number of MRI studies using guinea pigs as animal models, however there has been no automatic brain extraction algorithm available for guinea pig brain extraction. In this study, we describe a guinea pig brain deformable model method (GBD) based on our previous rat brain extraction method (RBD). GBD is a template-based approach with adjusted model parameters compared to RBD. We validated GBD on 20 guinea pig T2-weighted images and achieved an extraction accuracy of above 90%.

Zhengyi Yang¹, Jurgen Fripp², Craig Engstrom¹, Shekhar Chandra², Ying Xia², Anthony Paproki², Mark Strudwick¹, Ales Neubert², and Stuart Crozier^1
1University of Queensland, Brisbane, Queensland, Australia, ²CSIRO, Brisbane, Queensland, Australia

In conditions such as shoulder osteoarthritis and impingement syndrome, it is important to quantify the subtle changes in the morphology of the glenohumeral cartilages, which can be measured from image segmentation. However, automatic cartilage segmentation from MR images is challenging. As a critical step stone, we present a fully automatic shoulder bone segmentation pipeline using statistical shape models. The mean volume overlap calculated as Dice Similarity Coefficient between automatic and manual segmentation is 0.94 and 0.76 for humerus and scapula, respectively. These promising results imply a high likelihood of the proposed pipeline being integrated into a fully automatic solution to shoulder cartilage segmentation and quantitative analysis on cartilage morphometry.

Ali Gholipour¹, Onur Afacan¹, Iman Aganj², and Simon K Warfield^1
1Department of Radiology, Boston Children's Hospital and Harvard Medical School, Boston, Massachusetts, United States, ²Athinoula Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, Massachusetts, United States

This educational electronic poster provides an in-depth review of super-resolution techniques in MRI. This review discusses techniques based on image, frequency, and wavelet domains and considers the effect of slice profile and point spread function estimations on the performance of reconstruction techniques.

Caroline L. Hoad¹, Kathryn Murray¹, Jill Garratt², Jan Smith², David J. Humes², Susan T. Francis¹, Luca Marciani², Robin C. Spiller², and Penny A. Gowland^1
1Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, Nottinghamshire, United Kingdom, ²Nottingham Digestive Diseases Centre, NIHR Biomedical Research Unit in GI and Liver Diseases, University Hospitals NHS Trust and the University of Nottingham, Nottingham, Nottinghamshire, United Kingdom

This study describes a semi-automatic segmentation algorithm to separate subcutaneous, visceral and total adipose fat from mDIXON MRI data. Data from 10 subjects with a wide range of BMIs were used to validate the algorithm. The algorithm used standard image processing techniques and did not use any training data. Excellent agreement between the algorithm and manual segmentation of the same data was found. Bland-Altman analysis found a small bias in the subcutaneous adipose tissue between manual and semi-automatic methods. Excellent agreement was also found between the results of 2 different observers.

Vincent Ugarte¹, Vadim Malis¹, Usha Sinha¹, Robert Csapo², and Shantanu Sinha^2
1Physics, SDSU, San Diego, CA, United States, ²Radiology, UCSD, San Diego, CA, United States

The study of the correlation between muscle loss, increase of intramuscular adipose (IMAT) and connective (IMCT) tissues with age is important topic for understanding how human physiology changes with age. A Dual echo Ultrashort TE (UTE) sequences is used to map the very short T2 connective tissue. A 3D spatial fuzzy c-means segmentation approach is used with intensity and TE maps and associated structure maps evaluated from the Hessian tensor. The spatial term was able to classify tissues in the presence of fairly severe shading artifact present in the input TE volumes.

Rene C. W. Mandl¹, Martijn P. van den Heuvel¹, Rachel M. Brouwer¹, and Hilleke E. Hulshoff Pol^1
1Psychiatry, Brain Center Rudolf Magnus, UMC Utrecht, Utrecht, Utrecht, Netherlands

Numerous cortical thickness studies using MRI operating at conventional field strengths (e.g. 3T) showed cortical reductions for various psychiatric diseases. But the question of which of the cortical layers were implicated could not be answered. Here we propose an automatic method that can extract additional cortical information from MRI scans acquired at conventional field strength. Initial experiments indicate that the method yields robust results and that it can be applied in large cohort studies to extract more detailed cortical information.

Christopher Leslie Adamson¹, Richard Leslie Beare¹, Mark Walterfang², and Marc Seal^1
1Developmental Imaging, Murdoch Childrens Research Institute, Parkville, VIC, Australia, ²Royal Melbourne Hospital, Parkville, VIC, Australia

To present a fully automated software pipeline for thickness profile based morphological analysis of the midsagittal section of the corpus callosum (CC) using 3D structural T1-weighted images. This pipeline contains a novel CC segmentation algorithm that is demonstrated to be efficient and highly accurate. The pipeline contains midsagittal slice extraction, CC segmentation, quality control tools, thickness profile generation, and group-wise statistical analysis routines with display scripts. The pipeline is implemented in MATLAB and is computationally efficient, completing thickness profile generation in 10 seconds per subject on average.

Kunio Nakamura¹, Nicolas Guizard¹, Vladimir S. Fonov¹, Sridar Narayanan¹, D. Louis Collins¹, and Douglas L. Arnold^1
1McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada

We developed an MRI simulation dataset with accurately known amount of brain volume and atrophy. We used tissue probability maps from 20 subjects and simulated central brain atrophy (enlarging lateral ventricles and contraction of surrounding brain tissue) by modifying the probability maps. The MRIs were simulated by mrisim/MINC (BrainWeb). Using this simulation dataset, we evaluated common brain atrophy measurement methods (SPM/SIENAX/SIENA/longitudinal FreeSurfer/Jacobian integration method) against the gold standard atrophy rates, which were measured from probability maps. SIENA, Jacobian, and FreeSurfer performed well in terms of correlation and/or accuracy. The dataset will be publicly available for validation of future methods.

Hamied A Haroon^1,2, Ross Little^1,2, Kola Babalola^1,2, Chris Miller^1,2, Neal Sherratt^1,2, Barry Whitnall^1,2, Tim Cootes^1,2, Chris Taylor^1,2, Geoff J Parker^1,2, and Chris Rose^1,2
1Centre for Imaging Sciences, The University of Manchester, Manchester, England, United Kingdom, ²Biomedical Imaging Institute, The University of Manchester, Manchester, England, United Kingdom

We present a novel method for measuring left-ventricular end-diastolic volume (EDV) from highly under-sampled k-space, without need for explicit image reconstruction. Using retrospectively under-sampled k-space data (8%), from 31 healthy volunteers, we show that the method can accurately (r=0.91,p < 0.001) estimate EDV with a mean bias of just 11 ml. The ability to parameterize features in the way we describe allows for much faster, tailored quantitative imaging.

Ali Caglar Özen¹, Ute Ludwig¹, Lena Maria Öhrström², Frank Jakobus Rühli², and Michael Bock^1
1Radiology - Medical Physics, University Medical Center Freiburg, Freiburg, Germany, ²Centre for Evolutionary Medicine, Institute of Anatomy, University of Zürich, Zürich, Switzerland

In this work, MRI of the left hand of an embalmed Egyptian mummy is implemented with UTE, PETRA and SPI sequences using a home-made Tx/Rx solenoid coil. Sequences are compared in their performance of SNR and extraction of anatomical details. Although PETRA has higher SNR, UTE reveal anatomical details better. SPI image is sharper yet it requires long scan times. Performances of the sequences are evaluated and UTE is found more suitable to MRI of ancient remains.

Dingxin Wang^1,2, Peter Kollasch¹, Xiufeng Li², An Vu², Edward Auerbach², Steen Moeller², Essa Yacoub², Kamil Ugurbil², and Vibhas Deshpande^3
1Siemens Medical Solutions USA, Inc., Minneapolis, MN, United States, ²Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, United States, ³Siemens Medical Solutions USA, Inc., Austin, TX, United States

Our study demonstrates the application of multiband slice accelerated TSE for T2-weighted brain imaging at 3T. Multiband slice acceleration improves the acquisition efficiency of TSE. Our results suggest the utility of multiband slice acceleration as a complementary and/or alternative solution to traditional 1D parallel imaging for time reduction.

Martin Ott¹, Martin Blaimer¹, Felix Breuer¹, David Grodzki², and Peter Jakob^1,3
1MRB Forschungszentrum für Magnet-Resonanz-Bayern e.V., Würzburg, Bavaria, Germany, ²Siemens AG, Erlangen, Bavaria, Germany, ³Department of Experimental Physics 5, University of Wuerzburg, Würzburg, Bavaria, Germany

Acoustic noise reduction is an upcoming field in MRI. We target on the challenging short echo time regime in TSE-imaging to obtain images with PD- or T1-weighting. We present present a functionally approach for significantly reducing acoustic noise without the need to modify hardware.

Pavel Falkovskiy^1,2, Daniel Brenner³, Thorsten Feiweier⁴, Stephan Kannengiesser⁴, Bénédicte Maréchal^1,2, Tobias Kober^1,2, Alexis Roche^1,5, Kaely Thostenson⁶, Denise Reyes⁶, Matthias Seeger⁷, Tony Stoecker³, Matt Bernstein⁶, and Gunnar Krueger^1,2
1Advanced Clinical Imaging Technology, Siemens Healthcare IM BM PI, Lausanne, Switzerland, ²CIBM - AIT, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, ³German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, ⁴Healthcare Sector, Siemens AG, Erlangen, Germany,⁵Department of Radiology, University Hospital (CHUV), Lausanne, Switzerland, ⁶Department of Radiology, Mayo Clinic, Rochester, Minnesota, United States, ⁷Laboratory for Probabilistic Machine Learning, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

The MPRAGE pulse sequence is often used for structural brain imaging because of its good GM/WM contrast. Recently, several strategies beyond conventional parallel imaging have been proposed that can be applied to further shorten the several-minute-long 3D whole brain acquisition time, e.g. 2D-GRAPPA, CAIPIRINHA, and segmented MPRAGE. This work qualitatively and quantitatively compares the performance of four 3-minute MPRAGE variants of the frequently employed 5-minute ADNI-2 3D brain imaging protocol. All images obtained with a 3-minute protocol are rated as being of clinically useful image quality, though the fast protocols are found to introduce systematic deviations in automated volumetric measurements.

James A. Rioux¹, Ives R. Levesque^1,2, Manojkumar Saranathan¹, and Brian K. Rutt^1
1Radiology, Stanford University, Stanford, CA, United States, ²Medical Physics, Oncology, and RI-MUHC, McGill University, Montreal, QC, Canada

Three methods for whole-brain high-resolution T₁ mapping are evaluated at 7T: 3-TI MPRAGE, MP2RAGE and MP3RAGE. Each acquires a number of T₁-weighted MPRAGE images, allowing lookup of T₁ values instead of a least-squares fit, and is B₁-insensitive to varying degrees. All acquisitions were matched at 10min total scan time and 1mm isotropic resolution. MP2RAGE demonstrated the highest T₁-to-Noise-Ratio, while 3-TI MPRAGE had the lowest difference from a reference map but has reduced T₁NR due to the acceleration required to achieve the desired scan time. MP3RAGE is a good compromise method, with.higher T₁NR than 3-TI MPRAGE, but higher accuracy than MP2RAGE.

Andreas Berg^1,2 and Karin Wiesauer^3
1Center for Medical Physics and Biomedcial engineering, Medical University of Viennna, Vienna, Austria, ²MR Centre of Excellence, Medical University of Vienna, Vienna, Austria, ³RECENDT Research Center for Non-destructive Testing GmbH, Linz, Austria

The MR-visualization of the very rigid composite materials for dental care appears to be impossible by MR methodology based on spin echo detection due to the very short T2* of the polymer and ceramic compounds. However we demonstrate here the possibility to visualize the solidifying process during illumination at high spatial resolution in these very rigid Biocompatible polymer-ceramic composites using UTE T2* micro-mapping on a 7T Ultra-High-Field human scanner. Quality control analysis allows for the detection of possible sources of artifacts in UTE-imaging. Inhomogeneities in the composite as possible source for mechanical failure are detected.

Takakiyo Nomura¹, Tetsu Niwa², Takashi Okazaki², Takuya Hara², Tatsuya Sekiguchi³, Shuhei Shibukawa², Hiroaki Nishio², Noriharu Yanagimachi², and Yutaka Imai^2
1Radiology, Tokai University of School of Medicine, Isehara, Kanagawa, Japan, ²Tokai University of School of Medicine, Kanagawa, Japan, ³Tokai University of School of Medicine, Tokai University of School of Medicine, Japan

Fast single-shot three-dimensional k-space acquisition (free factor) with balanced turbo field-echo (bTFE) of the thoracic duct was performed. We compared the visualization of the thoracicdact among three sequences (free factor with bTFE, 3D turbo spin-echo sequence, and conventional bTFE). We divided the thoracic duct into three segments (upper, middle, and lower). Each segment was assessed using a five-point scale. We found good visualization of the thoracic duct using 3D free factor with bTFE.

Ziyue Wu¹, Yoon-chul Kim², Michael C.K. Khoo¹, and Krishna S. Nayak^2
1Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, ²Electrical Engineering, University of Southern California, Los Angeles, CA, United States

Obstructive sleep apnea (OSA) is a disease characterized by repetitive upper airway (UA) narrowing or collapse during sleep. Upper airway compliance, computed as the change in UA cross-sectional area per unit pressure, is a measure of airway collapsibility. Invasive fiberoptic endoscopy has been previously used to measure the UA area and compliance while recently MRI has been used as an alternative noninvasive approach. Current 2D cartesian MRI methods do not provide the spatio-temporal resolution needed to fully resolve UA collapsing dynamics during inspiratory load. We demonstrate the use of dynamic golden-angle radial FLASH imaging in combination with parallel imaging and compressed sensing reconstruction to achieve 1mm spatial and 90ms temporal resolution. From the results, UA dynamic changes can be better visualized and compliance measurement is more accurate with finer temporal and spatial resolution.

Sébastien Bär¹, Matthias Weigel^1,2, Jürgen Hennig¹, Dominik Von Elverfeldt¹, and Jochen Leupold^1
1Department of Radiology, Medical Physics, University Hospital Freiburg, Freiburg, Germany, ²Radiological Physics, University of Basel Hospital, Basel, Switzerland

Sub-millimeter image resolution, necessary in fields like molecular imaging, requires high gradient amplitudes of several hundred mT/m, mostly available at ultra-high field small animal scanner. The associated high b-values induce strong diffusion effects. Diffusion effects specific to the phase encoding gradients leading to modulation of the steady state signal were observed for a bSSFP sequence. In this work, we could observe an important loss of SNR in the images by comparing the theoretical expectation with the measurements. Simulations confirmed that this loss of SNR is associated to the diffusion effects of the PE gradients.

Mayur Narsude^1,2, Wietske van der Zwaag¹, Daniel Gallichan¹, Rolf Gruetter¹, and Jose Marques^2
1CIBM, EPFL, Lausanne, Vaud, Switzerland, ²University of Lausanne, Lausanne, Vaud, Switzerland

In this study we evaluate the importance of increased temporal resolution when analysing resting state networks. Furthermore, we evaluate the performance of two alternative approaches to obtain 3-fold acceleration in terms of temporal resolution in respect to standard multi-slice 2D-EPI: 3D-EPI-CAIPI and SMS-EPI. Both accelerated sequences offered improved statistical power in the RSN detection. Similar RSNs were detected with 3D-EPI-CAIPI and SMS-EPI and both could detect additional networks compared to standard 2D-EPI.

Borjan Gagoski¹, Cornelius Eichner², Mukund Balasubramanian³, Himanshu Bhat⁴, Kawin Setsompop², and Patricia Ellen Grant^1
1Fetal-Neonatal Neuroimaging & Developmental Science Center, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States, ²A.A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA, United States, ³Department of Radiology, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States, ⁴Siemens Healthcare, Charlestown, MA, United States

This work shows preliminary results demonstrating the feasibility and potential for reducing the readout length of the HASTE acquisition for 3T fetal imaging while keeping the original voxel sizes, by applying outer volume skewed saturation pulses oriented perpendicular to the PE direction. Further optimization of the application of spatial saturation pulses in the pregnant abdomen is yet to be done. We believe that these methods are promising way to reduce the HASTE readout length and thus will make the HASTE acquisitions less susceptible to fetal motion in the future.

Shuhei Shibukawa^1,2, Toshiaki Miyati², Hiroaki Nishio¹, Tomoya Nakamura¹, Yutaka Imai³, Testuo Ogino⁴, and Isao Muro^1
1Department of Radiology, Tokai university hospital, Isehara, Kanagawa, Japan, ²Division of Health Science,Graduate School of Medical Sciences, Kanazawa university, Kanazawa, Ishikawa, Japan, ³Radiology, Tokai university hospital, Isehara, Kanagawa, Japan, ⁴Healthcare department, Philips electronics Japan, minato-ku, Tokyo, Japan

In MRI, Yamada et al. reported a method which used Time-SLIP other than observation of CSF flow dynamics using phase contrast technique. we used a pencil beam pulse as the selective pulse to improve the selectivity of the target region in Time-SLIP. We report the validation results for CSF flow dynamics observations in comparison with Time-SLIP with pencil beam pulse (PB Time-SLIP) and that with slab pulse (SP Time-SLIP). In the flow phantom, it was found that pencil beam pulse was affected by the irradiation time prolonged at high flow rate. In human study, no significant difference was found in observation of CSF flow dynamics between SP Time-SLIP and PB Time-SLIP. In the case of slow flow rates such as CSF, PB Time-SLIP enables more selective visualization than SP Time-SLIP.

Albert Jang^1,2, Naoharu Kobayashi¹, Jianyi Zhang^1,3, and Michael Garwood^1
1Center for Magnetic Resonance Research and Department of Radiology, University of Minnesota, Minneapolis, MN, United States, ²Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN, United States, ³Cardiovascular Division, Department of Medicine, University of Minnesota, Minneapolis, MN, United States

Using the low-tip angle approximation based k-space formulation, a two-dimensional frequency modulated (2D FM) pulse has been designed for spatiotemporal MRI. Utilizing a spiral k-trajectory and radial symmetry, the 2D FM pulse selectively excites a cylindrical-shaped volume by sweeping the resonance region along a spiral trajectory during the pulse for spatiotemporal encoding. Furthermore, simulation and experimental results show that these pulses are able to maintain spatial selectivity even in the presence of frequency offsets that arise from chemical shifts and/or B0 inhomogeneity.

Emre Kopanoglu¹, Leo K. Tam¹, Haifeng Wang¹, and Robert Todd Constable^1
1Dept. Diagnostic Radiology, Yale University School of Medicine, New Haven, Connecticut, United States

Spatial encoding functions (SEFs) generated using nonlinear gradient fields (NLGFs) have curved spatial patterns. In Cartesian coordinates, such SEFs are non-orthogonal, which may lead to undesired flip-angle variations in excitation profiles. Although such effects can be suppressed using coordinate transformations, such a transformation may introduce loss-of-detail and introduce undesired constraints on RF pulse design. In this study, an iterative method that utilizes Matching-Pursuit and Conjugate-Gradient algorithms to design RF pulses in Cartesian coordinates is proposed. The method is compared to MP and CG algorithms, as well as non-iterative pulse design using nonlinear coordinate transformations, using simulations.

Florent Eggenschwiler¹, Kieran R. O'Brien^2,3, Bénédicte Maréchal^3,4, Rolf Gruetter^1,2, and José P. Marques^5
1Laboratory for Functional and Metabolic Imaging, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Vaud, Switzerland, ²Department of Radiology, University of Geneva, Geneva, Geneva, Switzerland, ³CIBM - AIT, École Polytechnique Fédérale de Lausanne, Lausanne, Vaud, Switzerland, ⁴Advanced Clinical Imaging Technology, Siemens H IM BM PI, Lausanne, Vaud, Switzerland, ⁵Department of Radiology, University of Lausanne, Lausanne, Vaud, Switzerland

At high field, the B₁⁺ inhomogeneity causes undesirable signal and contrast variations across the brain. It has been shown that those artifacts can be corrected by designing a single k_T-point pulse in the STA regime to replace all the hard pulses of a TSE sequence. In this work, to further improve T₂-weighted imaging, a specific k_T-point pulse is designed for each sequence pulse (dynamic k_T-point design). K_T-points were thus included to the SR-EPG formalism in order to optimize the magnetization state throughout the TSE sequence such that the signal across the brain matches the targeted one for each sequence echo.

Shaihan J Malik¹ and Joseph V Hajnal^1,2
1Division of Imaging Sciences and Biomedical Engineering, Kings College London, London, London, United Kingdom, ²Centre for the Developing Brain, Kings College London, London, United Kingdom

In fast spin echo (FSE) sequences pathways that meet the CPMG condition (90° phase between excited magnetization and refocusing pulses) are sustained, whereas non-CPMG pathways die away. We explore the suppression of non-CPMG pathways as an additional design flexibility for localized excitation pulses for inner-volume 3D-FSE imaging. The RF design problem is reformulated by splitting real and imaginary components and then a weighted optimization is used to unevenly distribute error to the non-CPMG channel. The approach is tested by designing multi-dimensional localized excitation pulses for a 3D-FSE imaging sequence and is shown to reduce excitation error and preserve background suppression.

Adrian Martin¹, Bastien Guerin², Yigitcan Eryaman^2,3, Joaquin L Herraiz³, Borjan Gagoski⁴, Elfar Adalsteinsson^5,6, Lawrence L. Wald^2,6, and Emanuele Schiavi^1
1Applied Mathematics, Universidad Rey Juan Carlos, Mostoles, Madrid, Spain, ²Radiology, Martinos Center, Massachusetts General Hospital, Charlestown, Massachusetts, United States, ³Madrid-MIT M+Vision Consortium in RLE, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States, ⁴Fetal Neonatal Neuroimaging & Developmental Science Center, Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts, United States, ⁵Electrical Engineering and Computer Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States, ⁶Health Sciencies and Technology, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States

Despite of the recent advances in parallel transmission RF pulse design for local and global SAR management, to the best of our knowledge errors in the transmit chain have never been taken into account. We present a method to design pTx pulses that satisfy the SAR limits using the knowledge about the transmit chain fidelity. This method has been validated for spokes pulses using simulated data for a 3Tbody coil with 8 Tx channels and a 7T head array with 8 channels.

Xiaoping Wu¹, Sebastian Schmitter¹, Kamil Ugurbil¹, and Pierre-Francois Van de Moortele^1
1CMRR, Radiology, University of Minnesota, Minneapolis, Minnesota, United States

Simultaneous MultiSlice (SMS) MR imaging using MultiBand (MB) RF pulses is playing an increasingly important role in neuroimaging. Recently, we have proposed a slice-wise parallel transmit (pTx) MB pulse design that can be used to improve transmit B1 homogeneity and/or reduce RF power consumption relative to a single channel Circular Polarized mode application. However, directly utilizing this slice-wise approach and designing pTx MB pulses in a slice-by-slice fashion for all necessary imaging slices for covering a large volume of interest, such as the whole brain, is impractical within clinical time constraints with current instrumentation, especially when using high through-plane spatial resolutions. In the present study, we propose a novel, effective and practical slab-wise pTx MB pulse design targeting volumetric coverage in SMS/MB imaging and describe some of its important properties by designing RF pulses based on electromagnetic modeling of a head array at 7T.

Alessandro Sbrizzi¹, Shaihan J Malik², Cornelis A van den Berg³, Peter R Luijten³, and Hans Hoogduin^3
1UMC Utrecht, Utrecht, NL, Netherlands, ²King’s College London, London, United Kingdom, ³UMC Utrecht, Utrecht, Netherlands

During the last decade, several numerical methods for RF pulse design have been developed. Recently, the magnitude least squares (MLS) and the multi-shift conjugate gradient least squares (msCGLS) algorithms have been proposed to relax the phase constraint in the resulting magnetization and to efficiently design RF pulses with minimum power. In this study, we present a multi-shift MLS (msMLS) algorithm which presents both advantages of the MLS and the multi-shift approach

Yi-Cheng Hsu¹, Ying-Hua Chu¹, Riccardo Lattanzi², Daniel K Sodickson², and Fa-Hsuan Lin^1,3
1Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan, ²The Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, New York University School of Medicine, NY, United States, ³Aalto University, Finland

To mitigate B1+ inhomogeneity in high field MRI, we proposed to combine RF shimming and B1+ remapping using SpAtially selective RF excitation with Generalized Spatial encoding magnetic fields (SAGS) method. Using both numerical simulations and experimental data, we demonstrate that such a combined strategy can provide the most homogeneous transverse magnetization distribution without the complexity of a full parallel RF transmission system. Compared to using RF shimming alone, our method can improve B1+ inhomogeneity by approximately 50% in simulations and 30% in experimental data.

Desmond H. Y. Tse¹, Michael S. Poole¹, Arthur W. Magill¹, Jörg Felder¹, Daniel Brenner^1,2, and N. Jon Shah^1,3
1INM - 4, Research Centre Jülich GmbH, Jülich, Germany, ²German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, ³Department of Neurology, RWTH Aachen University, Aachen, Germany

Four different encoding methods for B₁⁺ mapping in parallel transmission system, namely 1-channel-on, all-channels-on-except-1, all-channels-on-1-inverted and RF phase encoding (PE), were evaluated using dual refocusing acquisition mode (DREAM) at 9.4 T. RF PE, which encodes each transmit channel by phase rotation, was the least susceptible to artefacts caused by destructive RF interference. It showed negligible dependency on the initial RF phase setting and provides a flexible way to increase the number of measurements to increase SNR and reduce artefacts by weighted decoding. These advantages of RF PE make it a good choice for B₁⁺ mapping at ultra high field.

Alois Martin Sprinkart^1,2, Georg Schmitz², Frank Träber¹, Wolfgang Block¹, Jürgen Gieseke³, Hans Schild¹, Peter Börnert^4,5, and Kay Nehrke^4
1Dept. of Radiology, University of Bonn, Bonn, Germany, ²Institute of Medical Engineering, Ruhr-University Bochum, Bochum, Germany, ³Philips Healthcare, Hamburg, Germany, ⁴Philips Research Laboratory, Hamburg, Germany, ⁵Dept. of Radiology, LUMC, Leiden, Netherlands

To test if the recently proposed ultra-fast B₁⁺ mapping approach DREAM is applicable in clinical routine for patient-adaptive radiofrequency(RF)-shimming in body MRI at 3T, the performance of DREAM, double-angle (DA), and actual-flip-angle (AFI) B₁⁺ mapping methods was compared in 10 volunteers. Therefore, RF-shimsets obtained by DA-, AFI-, and DREAM- based B₁⁺ calibration data were analyzed by AFI- based B₁⁺ mapping. Results suggest that DREAM-based RF-shimming yields similar results as DA and AFI with respect to both, flip angle accuracy and B₁⁺ homogeneity, while the total acquisition time of the calibration sequence is shortened from one or two long breath-holds to about 1 second when DREAM is applied.

Peter Börnert^1,2, Kay Nehrke¹, and Jinnan Wang^3
1Philips Research, Hamburg, Germany, ²Radiology, LUMC, Leiden, Netherlands, ³Philips Research North America, Briarcliff Manor, New York, United States

Fast and robust in vivo B1+ mapping is an essential prerequisite for multi-element transmit applications like RF-shimming and transmit SENSE. DREAM, a recently introduced B1+ mapping approach, is very fast and promising, allowing single shot B1+ mapping within a fraction of a second. However, DREAM might be sensitive to flow which potentially degrades the B1+ maps for the blood pool signal in large vessels and in the heart. In the present work this has been solved successfully by adding an efficient and robust magnetization preparation approach which is studied in detail here for high field applications.

Kay Nehrke¹ and Peter Börnert^1
1Philips Research, Hamburg, Germany

In the present work, an efficient RF spoiling scheme is theoretically deduced for the DREAM B1+ mapping sequence. It allows a frequent re-acquisition of data from the same location without accuracy degradation due to the formation of a steady state. The performance of the approach is investigated in simulations and demonstrated in phantom experiments.

Julian Maclaren¹, Melvyn B. Ooi¹, Murat Aksoy¹, Jakob Ehrl¹, and Roland Bammer^1
1Dept. of Radiology, Stanford University, Stanford, CA, United States

In this work, we present a fast means to calibrate an optical tracking system used for prospective motion correction. A calibration tool comprising wireless active markers and an optical marker is used to simultaneously measure motion in both coordinate systems. A ‘hand-eye calibration’, often used in robotics, is then applied to compute the required coordinate transformation. Other quality assurance information, such as optical system latency, accuracy and precision can be derived from the same data.

Nicolas Adrien Pannetier^1,2, Theano Stravinos², Peter Ng², Michael Herbst³, Maxim Zaitsev³, Gerald Matson^1,2, and Norbert Schuff^1,2
1Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States, ²Centre for Imaging of Neurodegenerative Diseases, VA Medical Center, San Francisco, CA, United States, ³Department of Radiology, University Medical Center Freiburg, Freiburg, Germany

Motion artifacts in MRI images can be reduced with prospective motion correction system. However, results are sensitive to the accuracy of the motion tracking system and finding a sensitive metric for quantifying the effectiveness of the correction has been elusive. We tested two different marker positions; either mounted on the nose bridge or mounted on a mouth guard, and proposed a quantitative approach to compare effectiveness of motion correction. We found that the mouth guard outperforms the nose bridge fixation for accurately tracking the head motion and produce motion-free images.

Alexander Aranovitch^1,2, Maximilian Haeberlin¹, Axel Haase², and Klaas Paul Pruessmann^1
1Institute for Biomedical Engineering, University and ETH, Zurich, Zurich, Switzerland, ²Zentralinstitut für Medizintechnik, Technische Universität München, Munich, Bavaria, Germany

Very fast ehad motion cannot be neglected in long readouts such as single-shot EPI. The resulting k-space trajectory distortions cause image artefacts such as blurring, ghosting and geometric distortions if left untreated. Concurrent field monitoring in the head frame of reference is proposed to correct for fast motion during an EPI readout. Potential applications include tremor patients and other non-cooperative patients.

Steven Kecskemeti¹ and Andrew L Alexander^2
1Waisman Center, University of Wisconsin, Madison, WI, United States, ²Medical Physics, University of Wisconsin, WI, United States

Conventional MPRAGE (Magnetization-Prepared Rapid Gradient Echo) T1w acquisitions are highly sensitive to head motion and physiologic motions from flow and eye movements. The PROPELLER technique2 original introduced for T2 weighted imaging and later modified for T1 FLAIR imaging3 corrects well for in-plane motion, but not for through plane motion. The aim of this work is to develop a technique for T1w imaging that is (1) less sensitive to motion (2) can detect motion in 3D and finally (3) can retrospectively correct for occasional 3D motions from twitching, swallowing, coughing, or adjusting for comfort.

Shayan Guhaniyogi¹, Hing-Chiu Chang¹, Mei-Lan Chu¹, Allen W. Song¹, and Nan-Kuei Chen^1
1Brain Imaging and Analysis Center, Duke University, Durham, North Carolina, United States

Fast Spin-Echo (FSE) images are routinely acquired clinically due to their contrast range and fast acquisition. In the presence of rigid-body patient motion, however, the excitations in FSE correspond to different patient positions, resulting in motion-corrupted images when using standard reconstruction. Here we present a reconstruction method which can inherently estimate the level of rigid-body motion in FSE data, and produce images without blurring and artifacts. We demonstrate that the method provides better image quality than conventional reconstruction in the presence of rigid-body motion, and is useful for clinical FSE investigations requiring high quality images free of motion corruption.

Daniel Gallichan¹, José P Marques², and Rolf Gruetter^1,3
1CIBM-AIT, EPFL, Lausanne, VD, Switzerland, ²Dept. of Radiology, University of Lausanne, VD, Switzerland, ³Depts. of Radiology, Universities of Lausanne and Geneva, VD/GE, Switzerland

We recently demonstrated that the natural sparsity of fat images of the human head can be used to achieve extremely high acceleration factors. Here we demonstrate that this also allows conventional GRAPPA reconstructions at 64x acceleration, and use this for a motion navigator in a high-resolution MP-RAGE structural scan.

Martina F. Callaghan¹, Siawoosh Mohammadi¹, and Nikolaus Weiskopf^1
1Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, London, United Kingdom

Quantitative MRI (qMRI) with high resolution and whole brain coverage is sensitive to motion artefacts because of typically long acquisition times. We used a multi-parameter mapping (MPM) protocol to acquire quantitative maps of longitudinal relaxation rate (R1), magnetisation transfer (MT) and effective transverse relaxation rate (R2*). These quantitative maps were combined in a linear relaxometry model to reduce motion artefacts by exploiting the inconsistent manifestation of the artefacts across the maps.

Benjamin Zahneisen¹, Brian Keating¹, Aditya Singh¹, and Thomas Ernst^1
1University of Hawaii, Honolulu, HI, United States

Single-shot echo-planar imaging (EPI) is commonly assumed to be immune to subject motion. However, head movements can reach velocities in the 100mm/s or 100°/s range [1], resulting in translations and rotations of several mm or ° over the course of a typical readout train (tens of ms). Therefore, a simulation was performed to determine the effect of such fast movements on EPI scan quality.

Mari Elyse Boesen^1,2, Jerome Yerly^2,3, Robert Marc Lebel^2,4, and Richard Frayne^2,5
1Biomedical Engineering, University of Calgary, Calgary, AB, Canada, ²Seaman Family MR Research Centre, Calgary, AB, Canada, ³CardioVascular MR Research Center, Centre d'Imagerie BioMedicale, Lausanne, Switzerland, ⁴Applied Sciences Laboratory, GE Healthcare, AB, Canada, ⁵Radiology & Clinical Neuroscience, Hotchkiss Brain Institute, University of Calgary, AB, Canada

Brain pulsation with the cardiac cycle is a limiting factor in artifact-free high-resolution brain imaging. A non-uniformly sampled k-space with oversampled low frequency data reduced the effect of pulsatile motion. Further artifact reduction was achieved by retrospectively gating this non-uniformly sampled FSE data (cineFSE) and removing the components that demonstrated significant temporal variation.

Erik Beall¹ and Mark Lowe^1
1Imaging Institute, Cleveland Clinic, Cleveland, OH, United States

Head motion artifact is a major unsolved problem for fMRI and fcMRI. The primary reason why current methods do not solve motion is the limitation of volume-synchronized or -interpolated motion. Real motion happens at the slice level, and this has been shown to seriously limit accuracy of volumetric measures. Improved correction requires knowledge of slice-level motion, and in a major advance for the field, we present an algorithm that obtains this. Using a motion-injection sequence to acquire realistic motion-corrupted data in cadavers, we show the algorithm obtains the known motion parameters with accuracy on the level of external trackers.

Ali Gholipour¹, Onur Afacan¹, Benoit Scherrer¹, Burak Erem¹, and Simon K Warfield^1
1Department of Radiology, Boston Children's Hospital and Harvard Medical School, Boston, Massachusetts, United States

An image based motion correction and reconstruction approach has been developed for diffusion weighted imaging of subjects that move continuously and irregularly during scans. Subject motion poses random sampling in image space and q-space. We purposely perform oversampling to achieve sufficient coverage of the image space and q-space under motion conditions.

Shayan Guhaniyogi¹, Mei-Lan Chu¹, Hing-Chiu Chang¹, Allen W. Song¹, and Nan-Kuei Chen^1
1Brain Imaging and Analysis Center, Duke University, Durham, NC, United States

Multishot diffusion tensor EPI offers several advantages over singleshot acquisitions, including improved spatial resolution and reduced geometric distortions. However, it also has increased sensitivity to patient motion. While phase errors and pixel misregistrations among shots are commonly addressed in correction schemes, the altered diffusion-encoding of each shot due to motion is often neglected. We therefore present a new reconstruction technique to correct all three motion-induced errors in multishot diffusion EPI. The technique is shown to improve both image quality and tensor calculations, and is expected to be valuable for clinical and neuroscience applications requiring accurate high resolution diffusion tensor information.

Yuji Iwadate¹, Anja C.S. Brau², and Hiroyuki Kabasawa^1
1Global MR Applications and Workflow, GE Healthcare Japan, Hino, Tokyo, Japan, ²Global MR Applications and Workflow, GE Healthcare, Munich, DE, Germany

The self-navigator gating technique with the k-space center (DC) signal has less navigator signal deterioration than the conventional pencil-beam navigator gating when high imaging flip angles are used. We combined DC signals from the superior slices to improve respiratory motion detection in DC-based self-navigator sequence. The proposed method was tested with a 2D SPGR sequence, and it averaged DC signal fluctuations caused by spin saturation differences and cardiac motion, which resulted in improved motion detection and correction.

James H. Holmes¹, Diego Hernando², Yuji Iwadate³, Ann Shimakawa⁴, Gavin Hamilton⁵, Utaroh Motosugi², and Scott B Reeder^2,6
1Global MR Applications and Workflow, GE Healthcare, Madison, WI, United States, ²Radiology, Univeristy of Wisconsin-Madison, Madison, WI, United States,³Global MR Applications and Workflow, GE Healthcare, Hino, Tokyo, Japan, ⁴Global MR Applications and Workflow, GE Healthcare, Menlo Park, CA, United States, ⁵Radiology, University of California San Diego, San Diego, CA, United States, ⁶Medical Physics, University of Wisconsin-Madison, Madison, WI, United States

Fat and R2* quantification methods, in combination with free-breathing navigator-based methods provide a significant advantage for accommodating patients by removing the need for breath-holding. A trade-off exists between high signal for navigator motion detection and high steady-state signal for imaging without introducing T1 bias. This work evaluates the measured fat-fraction and R2* as a function of navigator flip angle. Results suggest the navigator excitation flip angle should be 10 degrees or lower to minimize fat-fraction bias at the location of the navigator excitation and enable fat-fraction and R2* measurements in good agreement with stationary phantoms.

Shujing Cao¹, Feng Huang², Bida Zhang³, and Rui Li^1
1Center for Biomedical Imaging Research, School of Medicine, Tsinghua University, Beijing, Beijing, China, ²Philips Healthcare, Gainesville, FL, United States,³Philips Research China, Beijing, China

An effective retrospective motion compensation method using iterative k-space convolution and selective coil combination was proposed for free breathing T2 weighted liver imaging without acquisition interruption. By iteratively implementing the GRAPPA like k-space convolution with gradually improved convolution kernel, widespread respiration motion induced errors can be sufficiently dispersed and cancelled out. Convergence of the designed image index was adopted to adaptively terminate the iteration. Meanwhile data from coil elements with low sensitivity to motion was selectively kept during iteration to preserve the signal noise ratio (SNR). In-vivo studies demonstrated its robustness in artifact reduction.

Maryam Seif¹, Laila Yasmin Mani², Huanxiang Lu³, Mauricio Reyes³, Bruno Vogt², Chris Boesch¹, and Peter Vermathen^1
1Depts Clinical Research and Radiology, University of Bern, Bern, Bern, Switzerland, ²Dept. Nephrology, Hypertension and Clinical Pharmacology, Inselspital Bern, Bern, Bern, Switzerland, ³Institute for Surgical Technology and Biomechanics, University of Bern, Bern, Bern, Switzerland

DTI-scans are performed employing respiratory-triggering in native kidneys to reduce severe physiological motion artifacts caused by respiration. Previous diffusion measurements in transplanted kidneys have been performed with and without controlling for respiratory motion. The aim of this study was to investigate if performing co-registration on non-triggered DTI may allow omitting respiratory triggering for transplanted kidneys. The clear improvement due to co-registration and the small difference between triggered and non-triggered images suggest that patients with renal allografts can be measured without respiratory triggering, but employing co-registration to improve the stability.

Wei-Ching Lo¹, Jesse I. Hamilton¹, Kestutis J. Barkauskas¹, Katherine L. Wright¹, Yong Chen², Mark A. Griswold^1,2, Nicole Seiberlich¹, and Vikas Gulani^1,2
1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, ²Radiology, University Hospitals Case Medical Center\Case Western Reserve University, Cleveland, OH, United States

3D abdominal MR imaging at high spatial resolution requires long breathholds to avoid motion artifacts. This study presents the combination of a 3D golden angle radial trajectory and non-Cartesian parallel imaging for high-resolution abdominal imaging. This trajectory provides a ¡§self-navigation¡¨ signal to detect the inevitable transition from breathhold to free-breathing, and the golden angle radial trajectory yields nearly uniform angular undersampling regardless of breathhold duration. This method can automatically and retrospectively provide clinically useful images even in the presence of motion while preserving high resolution for 3D abdominal imaging.

Yang Yang¹, Craig H Meyer^1,2, Christopher M Kramer^2,3, and Michael Salerno^2,3
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, ²Radiology, University of Virginia, Charlottesville, VA, United States,³Medcine, University of Virginia, Charlottesville, VA, United States

Compressed sensing (CS) has shown promise for highly accelerated first-pass perfusion imaging. However CS techniques suffer from temporal blurring in the presence of respiratory motion which is inevitable in clinical breath-held perfusion scans. Spiral pulse sequences have multiple advantages for myocardial perfusion imaging and their relatively incoherent aliasing pattern is advantageous for CS reconstruction. We have developed a parallel imaging and CS reconstruction technique for spiral first-pass perfusion imaging which incorporates a registration operator directly into the reconstruction model and thus outputs a series of motion-free perfusion images. We demonstrate high image quality and robustness to respiratory motion.

Nadia Paschke¹, Olaf Dössel¹, Tobias Schaeffter², Claudia Prieto², and Christoph Kolbitsch^2
1Institute of Biomedical Engineering, Karlsruhe Institute of Technology, Karlsruhe, Baden-Württemberg, Germany, ²Division of Imaging Sciences and Biomedical Engineering, King’s College London, London, United Kingdom

Navigator-gated acquisitions are in general used in 3D whole-heart MRI to reduce respiratory motion artefacts with the final scan time depending strongly on the breathing pattern of the subject. Motion correction approaches have been proposed to avoid these subject dependencies and ensure fast 3D high-resolution scans due to increased scan efficiency. Here two different respiratory motion correction techniques are compared in 4 volunteers and results show that incorporating motion information directly into an iterative reconstruction leads to the best image quality with a similar depiction of the coronary arteries and a scan time reduction of 43% compared to respiratory gating.

Muhammad Usman¹, Christian Baumgartner¹, Andrew King¹, David Atkinson², Tobias Schaeffter¹, and Claudia Prieto^1,3
1Division of Imaging Sciences and Biomedical Engineering, King's College London, London, Greater London, United Kingdom, ²Department of Image Computing, University College London, London, United Kingdom, ³Escuela de Ingenieria, Pontificia Universidad Catolica de Chile, Santiago, Chile

Multislice 2D CINE MRI is a common approach for assessing cardiac function and anatomy. This approach requires multiple breath-holds and usually suffers from slice-misalignments due to motion between the acquisitions. To overcome these problems, free-breathing respiratory-gated 3D CINE MRI have been proposed. However, these methods require long acquisition times and suffer from poorer contrast between blood and myocardium than 2D acquisitions1. In this work, we propose to combine the contrast of 2D acquisitions with the coverage of the 3D scans by generating a “simulated 3D” scan. This is achieved by addressing the issue of misalignment, which can occur between different slices in a multi-slice acquisition, using a self-gated Simultaneous Groupwise Manifold Alignment (SGA) technique. Prospective golden radial cardiac MR acquisitions, performed in 3 volunteers, demonstrate the feasibility of proposed framework to achieve gated “simulated 3D” CINE from free-breathing multi-slice 2D acquisitions.

Li Feng¹, Daniel K Sodickson¹, and Ricardo Otazo^1
1Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, New York University, School of Medicine, New York, New York, United States

Self-navigation provides an alternative to ECG triggering/gating in cardiac MRI and has been used in radial imaging where inherent navigator signal is obtained from the centers of k-space. However, this signal, which is an average of image over entire FOV, includes unwanted signals outside the region of interest and reduces the capability of accurate motion detection, particularly for patients with arrhythmias or irregular breathing. This study proposes a novel approach to automatic detect both respiratory and cardiac signal for free-breathing cardiac imaging by restricting the radial spokes to include signal from only region of interest based on the Fourier-slice theorem.

Jan Paul¹, Evica Divkovic¹, Stefan Wundrak¹, Peter Bernhardt¹, Wolfgang Rottbauer¹, Heiko Neumann², and Volker Rasche^1
1Internal Medicine II, University Hospital of Ulm, Ulm, Germany, ²Institute of Neural Information Processing, University of Ulm, Ulm, Germany

Several methods for self-gating (SG) are described in literature. However, to our knowledge, no comprehensive comparison of these variants has been performed yet. We compare different respiratory self-gating methods from radial acquisition by means of increase in image sharpness relative to non-gated reconstructions. In addition, high temporal resolution k-t SPARSE SENSE reconstruction of the gated data is investigated, which reveals more details of the heart motion.

Muhammad Usman¹, David Atkinson², Tobias Schaeffter¹, and Claudia Prieto^1,3
1Division of Imaging Sciences and Biomedical Engineering, King's College London, London, Greater London, United Kingdom, ²Department of Image Computing, University College London, London, United Kingdom, ³Escuela de Ingenieria, Pontificia Universidad Católica de Chile, Santiago, Chile

Manifold learning approaches can be applied in MRI to extract meaningful dimensions (manifolds) from the high-dimensional set of images. In this work, we propose a novel manifold learning based framework for cardiac and respiratory self-gating cardiac CINE MRI. Results show that the proposed approach estimates accurate cardiac and respiratory gating signals from ECG-free free breathing data and use these to achieve high spatial and temporal quality in retrospectively reconstructed CINE images.

Hsin-Jung yang¹, Jianing Pang¹, Behzad Sharif¹, Avinash Kali¹, Xiaoming Bi², Ivan Cokic¹, Diabao Li¹, and Rohan Dharmakumar^1
1Cedars Sinai Medical Center, Los Angeles, California, United States, ²Siemens Medical Solutions, IL, United States

Myocardial BOLD CMR is an appealing alternative to contrast-enhanced methods for the functional assessment of coronary artery disease. While significant advances have been made in BOLD CMR, limitations still exist. An ideal BOLD CMR method would be able to overcome the limitations by permitting acquisition of 3D myocardial BOLD images as maps (T2 or T2*) within 5 minutes, without the need to suspend breathing. In this work, we developed a stack-of-stars k-space acquisition scheme, which permits 100% acquisition efficiency, to construct 3D T2 maps with full LV coverage within 5 minutes. We tested this approach in healthy human volunteers.

Yoon-Chul Kim¹ and Krishna S. Nayak^1
1Electrical Engineering, Univ. of Southern California, Los Angeles, CA, United States

We demonstrate a continuous 3D balanced SSFP cardiac cine imaging in humans at 3Tesla. The technique is based on a golden-ratio Cartesian radial sampling with isotropic spatial resolution. The SAR reduction is achieved with a variable flip angle scheduling in which the flip angle is linearly weighted based on the distance from the k-space origin within the 2D phase encoding plane. Results demonstrate adequate depiction of the blood and myocardium borders, which allows for quantification of the ventricular function.

Yinghua Zhu¹, Yi Guo¹, R. Marc Lebel², Meng Law³, and Krishna Nayak^1
1Electrical Engineering Department, University of Southern California, Los Angeles, CA, United States, ²GE Healthcare, Calgary, Alberta, Canada,³Radiology Department, University of Southern California, Los Angeles, CA, United States

Compressed sensing has shown great potential in accelerating dynamic contrast enhanced MRI. Conventional Poisson-disc (PD) and Cartesian golden ratio (GR) radial schemes on the ky-kz plane of the 3D k-space are inefficient due to computation time and sub-optimal sparsity, respectively. We propose a novel randomized GR (RGR) sampling that is fast in sampling pattern generation on an MRI scanner, and flexible in temporal resolution selection in the reconstruction. We show and compare the results from PD, GR and RGR in retrospective studies using clinical DCE data. The proposed method yields promising results for highly accelerated DCE-MRI.

Robert Grimm¹, Dominik Nickel², Jana Hutter¹, Christoph Forman¹, Berthold Kiefer², Joachim Hornegger¹, and Kai Tobias Block^3
1Pattern Recognition Lab, FAU Erlangen-Nuremberg, Erlangen, Germany, ²MR Application Development, Siemens AG, Healthcare Sector, Erlangen, Germany, ³Department of Radiology, NYU Langone Medical Center, New York City, New York, United States

Modern imaging techniques for Dynamic Contrast Enhanced MR Imaging (DCE-MRI) can achieve high temporal resolution by exploiting correlation between different time-points in the reconstruction. In order to reduce the vast number of images as well as the reconstruction time while keeping the relevant clinical information, we propose to use a variable temporal resolution for the reconstructed time-points. Specifically, this is implemented for Golden-Angle Radial Sparse Parallel DCE-MRI and illustrated for prostate and liver imaging. In these applications the time resolution is critical in the initial phase after contrast injection and less important for the pre-contrast and later phases.

Benjamin Paul Berman¹, Abhishek Pandey², Zhitao Li², Theodore Trouard^3,4, Isabel Oliva⁴, Diego R. Martin⁴, Maria Altbach⁴, and Ali Bilgin^2,3
1Applied Mathematics, University of Arizona, Tucson, Arizona, United States, ²Electrical and Computer Engineering, University of Arizona, Tucson, Arizona, United States, ³Biomedical Engineering, University of Arizona, Tucson, Arizona, United States, ⁴Medical Imaging, University of Arizona, Tucson, Arizona, United States

Spirometry measurements during expiration are standard for diagnosis of many lung diseases. However, these measurements contain no spatial information. Highly accelerated 4D MRI with compressed sensing can be used to image the lungs during forced respiratory maneuvers. Golden angle stack-of-stars view ordering is utilized to distribute the data into undersampled temporal bins. Total variation between neighboring frames, as well as reference frames acquired during the same scan, provide enough sparsity to reconstruct images with as few as 1 readout per slice per time, ie. 100x acceleration, ie. 46ms resolution.

Yongwan Lim¹, Yeji Han¹, and HyunWook Park^1
1Department of Electrical Engineering, KAIST, Daejeon, Korea

In this abstract, a free-breathing data acquisition method was proposed for retrospective reconstruction in the abdomen. Based on a variable density sampling, multiple signals of the same PE are equidistantly acquired during a respiratory cycle and the consecutive PE steps are randomly distributed. Thus, this scheme prevented the adjacent PE lines from being densely acquired at non-quiescent respiratory phases. The proposed method was validated on a numerical simulation and in-vivo experiments from five healthy volunteers. The results showed high image quality comparable to the conventional data acquisition methods, thereby achieving reduction of respiratory motion artifacts.

Huisu Yoon¹, Dongwook Lee¹, Seung Hong Choi², Sung-Hong Park¹, and Jong Chul Ye^1
1Dept. of Bio & Brain Engineering, KAIST, Dae-jeon, Korea, ²Radiology, Seoul National University, Seoul, Korea

One of important applications of dynamic contrast-enhanced (DCE) MRI is to observe brain tissue dynamics. In this work, we applied a dynamic compressed sensing (CS) with patch-based non-convex low-rank penalty1 for 3D DCE-MRI to exploit temporal redundancy of the dynamic images to improve the spatiotemporal resolution of the conventional GRAPPA2 reconstruction. Experiments show that the proposed method provided improved spatio-temporal resolution than GRAPPA reconstruction.

Rebecca Ramb¹, Elias Kellner¹, Iulius Dragonu¹, Frederik Testud¹, Irina Mader², Jürgen Hennig¹, Maxim Zaitsev¹, and Bernd Jung^1
1Dept. of Radiology, Medical Physics, University Medical Center, Freiburg, Baden-Württemberg, Germany, ²Dept. of Neuroradiology, University Medical Center, Freiburg Brain Imaging, Freiburg, Baden-Württemberg, Germany

Parallel imaging offers the potential to decrease blurring and susceptibility artifacts in ssEPI acquisitions by shortening echo train lengths, however, at the expense of SNR. We propose k-t-EPI: k-t-undersampled ssEPI acquisition with k-t-GRAPPA reconstruction, in order to overcome SNR limitations. The developed sequence with two acquisition strategies and reconstruction is presented and applied in first-pass cerebral perfusion measurements at reduction factor R=4 and high spatial resolution. The gained flexibility in increasing spatial and/or temporal resolution by higher achievable reduction factors requires further investigation, also in the context of other applications such as diffusion or fMRI.

Mark Chiew¹, Stephen M Smith¹, Thomas Blumensath², and Karla L Miller^1
1FMRIB Centre, University of Oxford, Oxford, Oxfordshire, United Kingdom, ²IVSR, University of Southampton, Southampton, Hampshire, United Kingdom

Recently, we introduced the k-t FASTER method for rank-constrained acceleration of FMRI data acquisition. The method used an iterative hard thresholding with matrix shrinkage algorithm (IHT+MS) for image reconstruction, and showed robust recovery of FMRI resting state networks at modest (4x) acceleration factors using only rank constraints. Here, we introduce an enhancement of the IHT+MS algorithm that includes multi-coil information to improve the data-consistency by increasing the effective number of measurements. The multi-coil k-t FASTER method is demonstrated in 8-fold under-sampled data, using joint rank- and coil-constraints to recover k-t FMRI data with high fidelity.

Eugene Yip¹, Jihyun Yun², Keith Wachowicz¹, Zsolt Gabos¹, Satyapal Rathee¹, and Gino Fallone^1,2
1Department of Oncology, University of Alberta, Edmonton, AB, Canada, ²Department of Physics, University of Alberta, Edmonton, AB, Canada

Compressed Sensing (CS) can be beneficial to real time MRI guided interventions by significantly improving imaging frame rates. Spatial-temporal (k-t) CS can increase the acceleration potential of conventional CS but requires significantly longer reconstruction times. We have devised a novel spatial-temporal CS imaging strategy – Prior Data Assisted Compressed Sensing (PDACS), that is capable of near real time reconstruction (0.3s), and improves the reconstruction accuracy of conventional 2D-CS by using pre-acquired data to support reconstruction. In this work, we demonstrated the effectiveness of the PDACS technique in a lung tumour tracking study of cancer patients undergoing free breathing.

Aurélien Julien Trotier¹, William Lefrançois¹, Emeline Julie Ribot¹, Eric Thiaudière¹, Jean-Michel Franconi¹, and Sylvain Miraux^1
1CNRS UMR5536, RMSB, Bordeaux, Aquitaine, France

3D time-resolved cartesian TOF MR Angiography is a powerful method to assess blood flow velocity. Nevertheless total acquisition time required can be prohibitive. The purpose of this work was to develop a 3D TOF MRA based on radial projection trajectories. A first golden angle is used to uniformly distributed projections along time and a second golden angle to obtain different trajectories between cine images. The efficiency and flexibility (retrospective filtering) of this method was performed on mice model at 7T and shown an acquisition time reduction of at least a factor 4 in comparison to cartesian sampling.

Anders Garpebring^1,2, Maarten J. Versluis^1,3, and Matthias J. P. van Osch^1,3
1CJ Gorter Center for high field MRI, Leiden University Medical Center, Leiden, Zuid-Holland, Netherlands, ²Radiation Sciences, Umeå University, Umeå, Sweden, ³Radiology, Leiden University Medical Center, Leiden, Zuid-Holland, Netherlands

High resolution time-of-flight angiograms are frequently degraded due to pulsation artifacts. By acquiring the data randomly and subdividing it based on the timing relative to the cardiac cycle data suitable for spatio-temporal compressed sensing can be obtained. Results based on data from a 7 T scanner show that this approach is successful in eliminating the pulsation artifacts and that it in addition enables visualization of vessel intensity changes and vessel motion.

Manoj Kumar Sarma¹, Stan Rapacchi¹, Peng Hu¹, Daniel B. Ennis¹, M. Albert Thomas¹, Percy Lee², Patrick Kupelian², and Ke Sheng^2
1Radiological Sciences, UCLA School of Medicine, Los Angeles, CA, United States, ²Radiation Oncology, UCLA School of Medicine, Los Angeles, CA, United States

Respiratory motion has posed significant challenges in lung cancer radiotherapy. For patients presented with lung cancer, dynamic 2D lung MRI is a safe and robust method to characterize internal organ motion. Since the MR speed depends on the number of data points sampled in a given time, under-sampling of the k-space is a practical approach to shorten imaging time. Recently, various compressed sensing techniques have been utilized to accelerate imaging acquisition. In the study, the combination of transform domain sparsity with rank deficiency is used to reconstruct spatial-temporal lung dynamic MRI data and its ability to track lung tumor motion is examined.

Julia V Velikina¹ and Alexey A Samsonov^2
1University of Wisconsin - Madison, Madison, Wisconsin, United States, ²University of Wisconsin - Madison, WI, United States

A novel method is proposed for image series reconstruction from incomplete data that improves performance of techniques relying on temporal basis representation for dimensionality reduction. The improvement is achieved by automated iterative partition of FOV into several clusters with similar temporal dynamics and choosing a different locally adapted basis for each cluster. The spatially adaptive reconstruction allows for better image representation and higher SNR.

Johannes F.M. Schmidt^*1, Claudio Santelli^*1,2, and Sebastian Kozerke^1,2
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland, ²Imaging Sciences and Biomedical Engineering, King's College London, London, United Kingdom

Combining parallel imaging and compressed sensing (CS) has shown improved reconstruction performance as compared to applying either of the two methods alone. Sampling patterns are mostly designed to fully sample the k-space center while randomly undersample higher phase encodes. Trajectories combining regular and random undersampling have been shown to improve reconstruction accuracy. In dynamic imaging, time-interleaved k-t sampling may be used to reduce aliasing in the spatial temporal Fourier domain. We propose a k-t sampling scheme combining time-interleaved regular and random undersampling. Using cardiac short-axis data, it is demonstrated that this approach improves image reconstruction relative to standard CS trajectories.

Ozan Sayin¹, Haris Saybasili², Mark Griswold^3,4, Nicole Seiberlich⁴, and Daniel A. Herzka^1
1Department of Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, MD, United States, ²Siemens Healthcare USA, Inc., Chicago, IL, United States, ³Department of Radiology, Case Western Reserve University, Cleveland, OH, United States, ⁴Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States

Undersampled non-Cartesian trajectories permit high acceleration factors for real-time acquisitions with parallel imaging. Thus, accurate and efficient calibration schemes for such methods are important. Recently, a well-established parallel imaging technique GRAPPA, originally proposed for Cartesian trajectories, has been successfully extended to non-Cartesian imaging. This was enabled via an improved calibration formalism that extends the calibration to the time dimension (through-time calibration), and can be implemented for rapid imaging. The current study aims reducing the number of calibration frames required, thereby speeding up the real-time reconstructions significantly. A new calibration method that includes compression of the GRAPPA weights is proposed.

Peng Lai¹, Shreyas S Vasanawala², and Anja CS Brau^3
1Global MR Applications & Workflow, GE Healthcare, Menlo Park, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States, ³Global MR Applications & Workflow, GE Healthcare, Munich, Germany

3D cine MRI with high spatiotemporal resolution remains challenging due to residual artifacts and temporal flickering. Reconstruction for k-t accelerated 3D cine with high-density coil poses high computation demand, which is difficult for online processing. This work developed a new variable-density random k-t sampling scheme that produces less-disturbing noise-like artifacts and can be further suppressed by denoising. Also, this work combines a few newly developed algorithms (e.g. coil compression and data decoupling calibration) to address the computation challenge. The proposed sampling scheme and reconstruction framework improves the robustness of 3D cine with 9x acceleration and enables rapid online processing.

Stefan Wundrak^1,2, Jan Paul¹, Johannes Ulrici², Erich Hell², and Volker Rasche^1
1Ulm University Hospital, Ulm, Germany, ²Sirona Dental Systems, Bensheim, Germany

In clinical practice breath-hold ECG-synchronized cine protocols remain the preferred method for the assessment of the ejection fraction of the ventricles. Recent research aims for real-time free-breathing CMR by combining fast imaging se-quences with parallel imaging and compressed sensing. However, recent work reports a 10% underestimation of the ejec-tion fraction by real-time CMR. We reproduce these results with golden-angle radial sparse SENSE and show that the underestimation might be explained by the strong temporal regularization that is needed to reconstruct the highly under-sampled CMR images.

Angshul Majumdar¹ and Rabab Ward^2
1Indraprastha Institute of Information Technology, New Delhi, Delhi, India, ²Electrical and Computer Engineering, University of British Columbia, Vancouver, BC, Canada

This paper proposes an improvement over the recently proposed Blind Compressive Sensing (BCS) framework for dynamic MRI reconstruction. Owing to temporal correlation among the frames, the Casorati matrix formed by stacking the dynamic MRI frames as columns is a rank deficient matrix. BCS fails to capture this rank deficiency. We incorporate a rank deficiency penalty into the BCS framework and thereby improve the reconstruction results.

Huisu Yoon¹ and Jong Chul Ye^1
1Dept. of Bio & Brain Engineering, KAIST, Dae-jeon, Korea

One of important applications of dynamic contrast-enhanced (DCE) MRI is to observe brain tissue dynamics. In this work, we applied a dynamic compressed sensing (CS) with patch-based non-convex low-rank penalty1 for 3D DCE-MRI to exploit temporal redundancy of the dynamic images to improve the spatiotemporal resolution of the conventional GRAPPA2 reconstruction. Experiments show that the proposed method provided improved spatio-temporal resolution than GRAPPA reconstruction.

Muhammad Usman¹, David Atkinson², Gerald Greil¹, Tobias Schaeffter¹, and Claudia Prieto^1,3
1Division of Imaging Sciences and Biomedical Engineering, King's College London, London, Greater London, United Kingdom, ²Department of Image Computing, University College London, London, United Kingdom, ³Escuela de Ingenieria, Pontificia Universidad Católica de Chile, Santiago, Chile

In this work, we propose a novel method titled ‘Motion Corrected Sparse SENSE (MC-SS) that can give motion corrected reconstruction from free breathing data acquired in scan duration of a single breath-hold. Prospective multi-slice free-breathing golden radial cardiac MR acquisitions of 2 minutes (10 sec per slice), performed in 5 volunteers and 2 patients, demonstrate the feasibility of MC-SS framework for highly accelerated multislice CINE.

Stefan Wundrak¹, Jan Paul¹, Johannes Ulrici², Erich Hell², Sebastian Kozerke³, and Volker Rasche^1
1Ulm University Hospital, Ulm, Germany, ²Sirona Dental Systems, Bensheim, Germany, ³Institute for Biomedical Engineering, University of Zurich and Swiss Federal Institute of Technology, Zurich, Switzerland

The application of compressed sensing presumes a sampling scheme with incoherent, noise-like undersampling artifacts. In practical cardiovascular MRI this objective is usually not fully met. In particular, the very inhomogeneous sensitivity profile of a 32-channel coil array leads to strong streak artifacts around the chest-wall. In this work, the k-t radial SPARSE SENSE reconstruction method is extended with a regularization matrix Λ which adapts the temporal regularization strength to the local magnitude of the aliasing artifacts. The proposed method reduces the residual aliasing artifacts with-out compromising the temporal fidelity of the cardiac region.

Samuel T Ting¹, Yu Ding¹, Hui Xue², Shivraman Giri³, Ning Jin³, Rizwan Ahmad¹, and Orlando P Simonetti^1
1The Ohio State University, Columbus, OH, United States, ²National Heart Lung and Blood Institute, Bethesda, MD, United States, ³Siemens Healthcare, Chicago, IL, United States

We combine the Variable density Incoherent Spatio-Temporal Acquisition (VISTA) sampling pattern with the Fast Iterative Shrinkage Thresholding Algorithm (FISTA) implementation of SPIRiT to achieve online real-time, free-breathing cardiac cine imaging at sub 30ms temporal resolution. We test our method in six healthy volunteers. Visual scoring compared to conventional segmented techniques show little degradation in artifacts, temporal fidelity, and image quality. EDV and ESV measurements from segmented and real-time data show mean differences of 2.20% and 3.83% respectively. Reconstruction times of less than one minute within the Gadgetron framework promise a clinically practical implementation of this technique.

Nikolaos Dikaios¹, Benjamin Tremoulheac¹, Alex Menys², Simon Arridge¹, and David Atkinson^2
1Dept. of Medical Physics and Bioengineering, University College London, London, Great London, United Kingdom, ²Centre of Medical Imaging, University College London, Great London, United Kingdom

Quantification of small bowel motility correlates with disorders such as Crohn’s disease. The motility metric can be derived from non rigid registration of dynamic MR images, and its accuracy depends on their spatial/temporal resolution. We propose a split Bregman algorithm to reconstruct alias free dynamic MR images from undersampled (k,t)-space data, improving either the temporal resolution or maintaining the same temporal resolution and improving the spatial resolution. The proposed algorithm uses shearlets as an optimal sparsifying transform, and assumes that the recovered image consists of a low-rank plus a sparse component.

Yoon-Chul Kim¹, Biswas Joshi², Shirleen Loloyan², Roberta Kato², Michael C.K. Khoo¹, Sally L.Davidson Ward², and Krishna S. Nayak^1
1Univ. of Southern California, Los Angeles, CA, United States, ²Children's Hospital Los Angeles, Los Angeles, CA, United States

Our group has recently developed a novel real-time 3D MRI of the upper airway imaging which has the potential to provide complete anatomical information of the pharyngeal airway with sub-second temporal resolution and <2 mm isotropic spatial resolution. We have applied this technique to overweight and obese adolescents who snore, and have been able to observe and classify the patterns of upper airway narrowing and obstruction during respiratory events such as central or obstructive sleep apneas during natural sleep.

Kang Wang¹, James H Holmes¹, Shaorong Chang², Philip J Beatty³, Ann Shimakawa⁴, Lloyd Estkowski⁴, Scott B Reeder^5,6, and Ersin Bayram^7
1Global MR Applications and Workflow, GE Healthcare, Madison, WI, United States, ²MR Engineering, GE Healthcare, Waukesha, WI, United States,³Physical Sciences, Sunnybrook Research Institute, Toronto, ON, Canada, ⁴Global MR Applications and Workflow, GE Healthcare, Menlo Park, CA, United States, ⁵Radiology, University of Wisconsin-Madison, WI, United States, ⁶Medical Physics, University of Wisconsin-Madison, WI, United States, ⁷Global MR Applications and Workflow, GE Healthcare, Houston, TX, United States

Direct Virtual Coil (DVC) has been previously demonstrated to be an efficient method to reduce computation time for data-driven parallel imaging. It has also been shown that DVC preserves phase information, and is compatible with phase-sensitive applications, such as 2-pt Dixon qualitative water-fat separation. In this work, we demonstrate the feasibility of DVC for quantitative fat fraction imaging.

Alessandro Sbrizzi¹, Bjorn Stemkens², Sjoerd Crijns², Cornelis A van den Berg², Jan J Lagendijk², Peter R Luijten², Rob Tijssen², and Anna Andreychenko^2
1UMC Utrecht, Utrecht, NL, Netherlands, ²UMC Utrecht, Utrecht, Netherlands

Multi-band based imaging techniques require phase cycling between the different slices when the sensitivity maps are not sufficiently distinct. The phase cycling has the effect of shifting the image by a fraction of the FOV. The phase difference is therefore fundamental for the successful application of this kind of techniques. In previous studies, the negative effect of signal leakage and cross talk between simultaneously excited slices has been shown. In this work, we show that inter-slice leakage can be partially caused by an imperfect phase cycling and we derive a mathematical model to quantify it. Based on the model, correction of imperfect phase cycling could be done.

J Gabriel Castrillon¹, Valentin Riedl^1,2, Martin Bührer³, and Christine Preibisch^1
1Neuroradiology, Klinikum Rechts der Isar, Technische Universität München, München, Bayern, Germany, ²TUM-Neuroimaging Center, Klinikum Rechts der Isar, Technische Universität München, München, Bayern, Germany, ³Gyrotools, Zurich, Switzerland

This work evaluates the results of r-fMRI analysis when different multiband factors are used in data acquisition. Four subjects were scanned on a Philips Ingenia 3 T using a 32 channel head coil. Multiple r-fMRI data were acquired for 7 min using four protocols with different multiband factors. Even though the temporal SNR decreases as could be expected from the heavily undersampled data combining a SENSE factor of 2 with multiband factors up to 4, the mean z-values and spatial representation of the four investigated networks were remarkably stable. Multiband EPI technique could become a new standard in (r-)fMRI.

Xinran Zhong^1,2, Jingyuan Lyu², Chuangjian Cai¹, Kui Ying³, and Leslie Ying^2
1Department of Biomedical Engineering, Tsinghua University, Beijing, Beijing, China, ²Department of Biomedical Engineering, University at Buffalo, Buffalo, NY, United States, ³Department of Engineering Physics, Tsinghua University, Beijing, Beijing, China

Recently a novel algorithm called nonlinear GRAPPA is proposed to improve the SNR with high acceleration factors. However, nonlinear GRAPPA usually requires more ACS lines, which limits the net acceleration factor that can be achieved. In this abstract, we present a method that combines the advantage of GRAPPA and NL-GRAPPA to achieve even higher net reduction factors than GRAPPA and NL-GRAPPA alone. Experimental results demonstrate that the proposed method is able to achieve a high reduction factor of 3.76 in 2D imaging without significant SNR loss/artifacts.

Enhao Gong¹ and John M Pauly^1
1Electrical Engineering, Stanford University, Stanford, CA, United States

The estimation of GRAPPA and SPIRiT auto-calibration kernel, which is usually formed as an inverse problem, is an essential step for Parallel Imaging (PI). Regularizations for the kernel coefficients have been discussed before to achieve more accurate kernel estimation. However, the weighting for each measurement in the inverse problem has not been fully investigated. In this work, we propose a novel scheme for auto-calibration PI, which consider both measurement and kernel coefficients to achieve an optimal solution under a statistical model. Experiments compared with previous proposed calibration methods demonstrated advantages in kernel value estimation and reconstruction accuracy.

Stephen F Cauley¹, Berkin Bilgic^1,2, Jonathan R Polimeni^1,2, Himanshu Bhat³, Lawrence L Wald^1,4, and Kawin Setsompop^1,2
1A.A. Martinos Center for Biomedical Imaging, Dept. of Radiology, MGH, Charlestown, MA, United States, ²Harvard Medical School, Boston, MA, United States, ³Siemens Medical Solutions Inc, Malvern, PA, United States, ⁴Harvard-MIT Division of Health Sciences and Technology, Cambridge, MA, United States

High channel-count array coils have enabled accurate parallel imaging (PI) reconstruction at very high acceleration factors. However, the computational scaling of many PI algorithms leads to long reconstruction times. Methods such as SVD are applicable to a wide range of k-space sampling patterns but produce poor image quality. Other improved methods such as Geometric-decomposition Coil Compression are tailored for Cartesian sampling. In this work, we introduce a Hierarchical Mapping Framework (HMF) for coil compression that improves upon previously proposed algorithms. The additional flexibility provided by HMF should enable accurate PI reconstruction for many acquisition types.

Dara Bahri¹, Martin Uecker¹, and Michael Lustig^1
1Electrical Engineering and Computer Sciences, University of California, Berkeley, California, United States

ESPIRiT is a subspace-based method that estimates coil sensitivity maps from calibration data. It produces maps where only the relative phase is known. Often, the phase of a single coil is used as reference, which can result in non-smooth phase. This is problematic for phase-sensitive applications and compressed sensing as it reduces the sparsity of the image. In this work we propose a technique for determining basis vectors for ESPIRiT maps which are smoothly aligned in space. In addition to producing smooth phase in the maps, it prevents mixing of signal components as might occur when imaging a tight FOV.

Zechen Zhou¹, Jinnan Wang^2,3, Niranjan Balu³, and Chun Yuan^1,3
1Center for Biomedical Imaging Research, Tsinghua University, Beijing, China, ²Philips Research North America, Seattle, Washington, United States,³Radiology, University of Washington, Seattle, Washington, United States

Recently, a parallel reconstruction technique SAKE has been developed using Singular Value Decomposition (SVD) to impose low rank property for calibrationless parallel reconstruction, which can improve the result of SPIRiT. We hypothesize that a learned dictionary rather than SVD method can better adapt to acquired data and improve the reconstruction result. In this study, we propose a new patch-based dictionary learning method to estimate the local signal features in k-space and demonstrate its improved performance in-vivo.

Steen Moeller¹, An T Vu¹, Edward Auerbach¹, Kamil Ugurbil¹, and Essa Yacoub^1
1Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, Minnesota, United States

The use of the SENSE/GRAPPA algorithm in combination with multiband imaging using shifted FOV is established as feasible. Comparison with the conventional slice-GRAPPA algorithm is considered and the effect on different datatypes is investigated, with some datatypes exhibiting poor performance using the slice-GRAPPA algortihm

Jonathan Rizzo Polimeni¹, Kawin Setsompop¹, and W. Scott Hoge^2
1Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Harvard Medical School, Massachusetts General Hospital, Charlestown, MA, United States, ²Department of Radiology, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, United States

It has been recently demonstrated that phase errors across segments of multi-shot EPI data used for GRAPPA calibration can have a dramatic impact on image quality, causing severe SNR loss and ghosting/aliasing artifacts. Here we characterize several proposed calibration methods in terms of artifact level and tSNR to determine to what extent calibration data should match image data. We find that the ideal calibration data must match the accelerated EPI data in terms of phase errors and geometric distortion to provide high quality GRAPPA reconstructions. Target audience: Clinicians/researchers using accelerated echo planar imaging, especially in high-field or high-resolution applications.

Andrew S. Nencka¹, Daniel L. Shefchik¹, Andrew M. Huettner¹, and Andrzej Jesmanowicz^1
1Department of Biophysics, Medical College of Wisconsin, Milwaukee, WI, United States

An auto-calibrated multiband acquisition technology is presented. Hadamard unaliasing is used in a time series acquisition to generate reference images for subsequent SENSE unaliasing of the same data. This abstract offers an empirical optimization of the number of repetitions to use for generating reference images. In an ideal case, the full time series should be used for reference image generation, but system instabilities lead to an optimal window width for a given set of acquisition parameters and imaging hardware.

Martin Blaimer¹, Marius Heim², Daniel Neumann¹, Peter M. Jakob^1,2, and Felix A. Breuer^1
1Research Center Magnetic-Resonance-Bavaria (MRB), Würzburg, Bavaria, Germany, ²Department of Experimental Physics 5, University of Würzburg, Würzburg, Bavaria, Germany

One major drawback of parallel MRI (pMRI) is the noise amplification due to the reconstruction process. The noise amplification originates from the formulation of pMRI as an inverse problem. Phase-constrained algorithms achieve a significant SNR improvement by constraining all elements of the solution to be real-valued. In the original formulation, the SENSE reconstruction problem is split into real- and imaginary parts and is referred to as phase-constrained SENSE. An alternative formulation utilizes conjugate k-space symmetry by generating additional virtual coils. The purpose of this work is to demonstrate by mathematical analysis and imaging experiments that both formulations are equivalent.

Matthew J. Muckley^1,2, Douglas C. Noll^1,2, and Jeffrey A. Fessler^1,2
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, ²Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States

Algorithms that minimize SENSE-type image reconstruction cost functions almost always compute the gradient of a data consistency term at each iteration of the algorithm. Incremental gradient methods approximate the full gradient of the data consistency term by computing the gradient using a subset of the data. Since these subset gradients require less computation time, using them as a proxy for the full gradient significantly accelerates convergence. The method is general enough to be applied to any MR image reconstruction problem involving multiple receive coils with a SENSE model. Four-fold acceleration is shown with a low rank plus sparse model.

Kie Tae Kwon¹, Bob S Hu², and Dwight G Nishimura^1
1Electrical Engineering, Stanford University, Stanford, California, United States, ²Palo Alto Medical Foundation, Palo Alto, California, United States

A sliding interleaved cylinder (SLINCY) acquisition employs a 3D concentric cylinders trajectory as the readout instead of a 3DFT sequence. Due to the helical sampling geometry of the trajectory and the sliding-nature of the acquisition scheme, a dedicated parallel imaging strategy is required for SLINCY to achieve a clinically feasible scan time. In this work, we developed a new parallel imaging strategy for SLINCY, which decomposes the 3D helical structure of SLINCY into a series of 2D Cartesian planes. We demonstrated that the proposed parallel imaging strategy is feasible for SLINCY, which reduced the scan time down to almost 50%.

Yuchou Chang¹, James G. Pipe¹, and Michael Schär^1,2
1Neuroimaging Research, Barrow Neurological Institute, Phoenix, Arizona, United States, ²Philips Healthcare, Cleveland, Ohio, United States

In Cartesian SENSE parallel imaging the g-factor measures the noise propagation due to ill-conditioning of the matrix inversion during the reconstruction. To the best of our knowledge, the g-factor of SENSE images acquired with the PROPELLER method has never been studied. In this abstract, we propose a method to directly calculate g-factor map for SENSE reconstruction with PROPELLER trajectories and compare it to a Monte-Carlo simulation. In comparison with Cartesian imaging, experimental results demonstrate that the proposed method provides a useful tool for identifying noise behavior in PROPELLER SENSE and is computationally feasible in practice.

Kangrong Zhu¹, Robert F. Dougherty², John M. Pauly¹, and Adam B. Kerr^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²CNI, Stanford University, Stanford, CA, United States

In conventional simultaneous multislice (SMS) acquisitions, such as CAIPIRINHA and blipped-CAIPI, the simultaneously excited slices alias coherently, because these SMS acquisition techniques repeatedly sample the DFT encoding frequencies in the frequency spectrum of the simultaneous slices. In this work, we describe a new SMS acquisition approach called 'Multislice acquisition with InCoherent Aliasing (MICA)', which uses irregular encoding in order to create incoherent aliasing of the simultaneously excited slices. We demonstrate the feasibility of using MICA for SMS imaging through simulation and in vivo experiments. Our preliminary results show that the performance of MICA is comparable to CAIPI.

Stephen F Cauley¹, Micheal Lustig², Berkin Bilgic^1,3, Himanshu Bhat⁴, Lawrence L Wald^1,5, and Kawin Setsompop^1,3
1A.A. Martinos Center for Biomedical Imaging, Dept. of Radiology, MGH, Charlestown, MA, United States, ²Electrical Engineering and Computer Sciences, UC Berkeley, CA, United States, ³Harvard Medical School, Boston, MA, United States, ⁴Siemens Medical Solutions Inc, Malvern, PA, United States, ⁵Division of Health Sciences and Technology, Harvard-MIT, Cambridge, MA, United States

Simultaneous MultiSlice EPI acquisition significantly increases the temporal efficiency for both diffusion-weighted imaging and functional MRI. With the Blipped-CAIPI modification and a large channel-count receive coil array, high resolution whole brain images are obtained in sub-second with little SNR penalty and artifact level. However, this breakthrough poses a challenge for the rapid reconstruction of these large datasets; a critical criteria for high patient throughput in clinical and research settings. We use Geometric-decomposition Coil Compression to ameliorate the computational challenges associated with SMS-EPI acquisitions at high MB factors. This enables real-time reconstruction of large datasets using vendor's provided computational hardware.

Grzegorz Tomasz Kowalik¹, Daniel Steven Knight^1,2, Jennifer Anne Steeden¹, Oliver Tann^1,3, Freddy Odille^4,5, David Atkinson⁶, Andrew Taylor^1,3, and Vivek Muthurangu^1,3
1UCL Institute of Cardiovascular Science, London, United Kingdom, ²Royal Free Campus, UCL Division of Medicine, London, United Kingdom,³Cardiorespiratory Unit, Great Ormond Street Hospital for Children, London, United Kingdom, ⁴IADI, INSERM, Nancy, France, ⁵Université de Lorraine, Nancy, France, ⁶Centre for Medical Imaging, UCL Division of Medicine, London, United Kingdom

The abstract presents the image reconstruction technique that combines parallel imaging (SENSE) and temporal encoding (UNFOLD) techniques for very high temporal acquisitions (~15 ms). The reconstruction was tested and used for the assessment of cardiac function.

Vladimír Jellúš¹ and Stephan A.R. Kannengiesser^1
1MR Applications Development, Siemens AG, Healthcare Sector, Erlangen, Germany

Combining images from multi-channel coils is challenging with respect to maximizing signal-to-noise ratio (SNR), providing an image phase for phase-sensitive techniques, and avoiding artifacts. This work combines advantages of two previously proposed groups of techniques: coil sensitivity estimations based on a separate calibration prescan including the body coil, and adaptive combination based on the image data themselves. The approach is to use the former for phase correction, prior to application of the latter. High quality, artifact free combined images with smooth phase can be achieved robustly and efficiently.

Souheil J Inati¹, Michael S Hansen², and Peter Kellman^2
1National Institute of Mental Health, National Institutes of Health, Bethesda, MD, United States, ²National Heart Lung and Blood Institute, National Institutes of Health, Bethesda, MD, United States

We propose a simple, SNR optimal method for coil sensitivity estimation and adaptive combination suitable for applications in which complex valued images are required. The proposed method enforces smoothness in both the magnitude and phase of the estimated coil sensitivities and overcomes the limitations inherent in previous methods. It has a very fast implementation and can be easily incorporated into a reconstruction pipeline. We demonstrate how this approach can mitigate the problems with phase. It is likely to be useful in applications such as susceptibility weighted imaging, complex interpolation, complex motion correction, and referenceless thermometry.

Yu Li^1
1Radiology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States

We developed a novel parallel imaging technique that performs image reconstruction in wavelet space. Since wavelet space is undersampled, aliasing exists in reconstructed images. However, if this aliasing is well controlled, aliasing-free images can be generated from inverse wavelet transform. Compared with conventional image- or k-space reconstruction, wavelet-space reconstruction provides better performance for aliasing suppression and noise control in imaging acceleration.

William Dominguez-Viqueira¹, Angus Z Lau², Albert P Chen³, and Charles H Cunningham^1,4
1Physical Sciences, Sunnybrook Research Institute, Toronto, Ontario, Canada, ²University of Oxford, United Kingdom, ³GE healthcare, Toronto, Ontario, Canada, ⁴Medical Biophysics, University of Toronto, Toronto, Ontario, Canada

Rapid multislice imaging of hyperpolarized 13C pyruvate was demonstrated by using a single-shot spiral pulse-sequence with up to 5-channel receiver-coils. SNR and coverage can be increased by using multi-channel reception as presented before; but final images are biased by the coil sensitivity when a simple sum-of-squares channel combination was used. In this work coil sensitivity coefficients were numerically calculated in Matlab and used for optimal linear combination reconstruction with in vivo measurements. SNR improvements of up to a 100 % in areas closer to the base of the heart were demonstrated by using the optimal reconstruction, as compared with sum-of-squares.

M. Dylan Tisdall^1,2
1A. A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, Massachusetts, United States, ²Radiology, Harvard Medical School, Boston, Massachusetts, United States

We show initial results with a novel reconstruction algorithm and small modification to previous undersampling strategies used in parallel imaging. Our novel algorithm constrains the coil sensitivities to be compact in the Fourier domain, and uses a small set of samples acquired with the body coil to further constrain the solution space. We demonstrate high-quality magnitude and phase images can be estimated from 3D FLASH with at least 7x reductions in scan time.

Benjamin Zahneisen¹, Rasim Boyacioglu², Thomas Ernst¹, and Benedikt A. Poser^3
1School of Medicine, University of Hawaii, Honolulu, HI, United States, ²Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, Netherlands, ³Maastricht University, Maastricht, Netherlands

Simultaneous multi-slice (SMS) imaging has recently gained in popularity, especially for 2D single-shot sequences like EPI. Established reconstruction techniques for SMS acquisitions are SENSE/GRAPPA or “slice-GRAPPA” when CAIPIRINHA is used. With a few exceptions, most SMS reconstructions require a two-step approach to first disentangle the aliased slices, and then perform in-plane parallel reconstruction, or vice versa. We propose a 2D-SENSE reconstruction as a general one-step approach to reconstruct SMS data with arbitrarily undersampled Cartesian k-space (CAIPI-like patterns) in phase and/or slice directions. In case of additional in-plane undersampling it provides an easy-to-implement, contrast independent, one-step reconstruction along both undersampled dimensions.