Pulse Sequences - Spectroscopy

Wednesday 3 June 2015

Fan Lam¹, Bryan Clifford¹, Chao Ma², Curtis L. Johnson², and Zhi-Pei Liang^1
1Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ²Beckman Institute, University of Illinois at Urbana-Champaign, Urbana, IL, United States

This paper presents a novel data acquisition and processing method for subspace-based MRSI to enable ultra-high resolution 3D 1H-MRSI of the brain. Results from both phantom and in vivo experiments will be shown to demonstrate the unprecedented capability of the proposed method.

Jian-Xiong Wang^1,2, Matthew E Merritt^1,2, A Dean Sherry^1,2, and Craig R Malloy^1,2
1Advanced Imaging Research Center, University of Texas Southwestern Medical Center, Dallas, Texas, United States, ²Department of Radiology, University of Texas Southwestern Medical Center, Dallas, Texas, United States

By applying 3D compressed sensing (CS) method to accelerate the acquisition time of chemical-shift imaging (CSI) which doesn’t possess the sparseness required by conventional 2D slice type CS. When the entire CSI data set is treated as one unity, CS can be satisfactorily applied to spectroscopic CSI to reduce acquisition time which is very desirable for possible multiple CSI acquisition during the lifetime of hyperpolarized agent for metabolic dynamic analysis such as Hyperpolarized 13C metabolism research.

Ruomin Hu¹, Simon Konstandin², and Lothar R. Schad^1
1Computer Assisted Clinical Medicine, Heidelberg University, Mannheim, Baden-Württemberg, Germany, ²MR-Imaging and Spectroscopy, University of Bremen, Bremen, Bremen, Germany

Sodium magnetic resonance imaging sets its focus of research primarily on the measurement of tissue sodium concentration with the attempt to distinguish between intra- and extracellular space. An alternative is represented by sodium imaging exhibiting relaxation-based contrast. The main goal of this work is to develop a sequence to explore the feasibility and properties of relaxation-based sodium imaging using phantoms as well as in vivo. Structural changes of the macromolecular environment such as the demyelination of axons and the degradation of cartilage might be highlighted using this contrast-generating technique.

Amir Seginer¹, Rita Schmidt¹, and Lucio Frydman^1
1Chemical Physics Department, Weizmann Institute of Science, Rehovot, Israel

We introduce a new method for fast spectral-imaging which avoids folding-in of spetra from outside of the spectral-BW, and which can be accelerated to a single-shot acquisition when specific spectral points are of intereset. The new method borrows a spectral encoding block from ultrafast 2D NMR spectroscopy (Frydman et al.), allowing a 1D-spatial / 1D-spectral “image” to be collected using an EPI-like acquisition with blips along the “spectral” dimension. Examples are given for breast imaging at 3T, separating tissue, fat, and silicone implant; and at 14T, separating five peaks spanning about 3kHz. Under work is an application to hyperpolarized dynamic imaging.

James Bankson¹, Christopher Walker¹, Wolfgang Stefan¹, David Fuentes², Matthew Merritt³, Yunyun Chen⁴, Craig Malloy³, Dean Sherry³, Stephen Lai⁴, and John Hazle^1
1Department of Imaging Physics, UT MD Anderson Cancer Center, Houston, TX, United States, ²UT MD Anderson Cancer Center, Department of Imaging Physics, Houston, TX, United States, ³Advanced Imaging Research Center, UT Southwestern Medical Center, Dallas, TX, United States, ⁴Department of Head & Neck Surgery, UT MD Anderson Cancer Center, Houston, TX, United States

Hyperpolarized (HP) MRI has enabled observation of biochemical processes in vivo with unprecedented resolution and specificity. These measurements are challenging due to the transient and non-renewable nature of HP signal enhancement, and the need to distribute dynamic spatial observations over a sufficient span of time to allow substrate interaction with biological targets. We have developed a kinetic model to describe the evolution of HP [1-¹³C]-pyruvate in vivo, and integrated this model into a constrained reconstruction algorithm exploits prior information from ¹H MRI and allows estimation of under-sampled dynamic HP imaging data as continuous functions of position and time.

Victor D. Schepkin¹, Boris M. Odintsov², Ilya Litvak¹, Peter L. Gor'kov¹, William W. Brey¹, Andreas Neubauer³, and Thomas F. Budinger^4
1NHMFL/FSU, Tallahassee, Florida, United States, ²UIUC, Illinois, United States, ³Heidelberg University, Germany, ⁴LBNL/UCB, California, United States

A novel way to detect bound potassium and sodium in vivo was explored based on triple quantum signals without filtration. The method is based on time proportional phase increments, allowing the detection of bound ions as a separate peak on triple frequency relative to single quantum MR signals. Bound ions signals are acquired in the same way and efficiently for a wide range of binding strengths. This is advantageous compared to typical TQF methods, especially during diseases or interventions. A comparison of TQTPPI signals in rat head revealed almost two times more effective binding of potassium relative to sodium.

B1 Imaging

Wednesday 3 June 2015

Yu Ding¹ and Jinghua Wang^2
1Dorothy M. Davis Heart and Lung Research Institute, The Ohio State University, Columbus, Ohio, United States, ²Center for Cognitive and Behavioral Brain Imaging, The Ohio State University, Columbus, Ohio, United States

Imaging of large objects at high field generally introduces large signal inhomogeneity associated with wavelength effect, object configuration and object electromagnetic properties. Estimating transmit field B1+ maps is prerequisite for various techniques (such as RF shimming, parallel excitation and inhomogeneity correction) to mitigate signal inhomogeneity, quantitative MR imaging and electrical property mapping. The status quo gold standard utilizes pixel-wise division which could be problematic when the image SNR is low. Here we explored the feasibility of estimating B1+ maps accurately, rapidly and robustly in k-space domain using a convolution kernel.

Alexis Amadon¹, Franck Mauconduit², Alexandre Vignaud¹, and Nicolas Boulant^1
1I2BM / NeuroSpin / UNIRS, CEA, Gif-sur-Yvette, France, France, ²Siemens Healthcare, Saint-Denis, France, France

The XFL sequence is a very fast 2D multi-slice B1-mapping sequence relying on the measurement of the magnetization Flip Angle (FA) of a selective preparation saturation pulse immediately preceding a centric-ordered FLASH readout. Like most 2D B1-mapping sequences, imperfect slice profiles introduce biases in the B1 measurement. Here we correct for this bias by taking into account the preparation & imaging slice profiles as well as the B0 offset map.

Parnian Zarghamravanbakhsh¹, John M Pauly¹, and Greig Scott^1
1Electrical Engineering, Stanford University, Stanford, CA, United States

Prior knowledge of magnetic field distribution is necessary for RF shimming and calibration.In practice, sample loading change coil current ,also field pattern tends to vary with relative location of coil to sample. In this study, we assess a method to localize transmit coil in imaging space by placing fiducial markers on known location of coils. Then the optimal rotation and translation matrix is calculated for mapping the target regions from simulation to imaging space. knowing coil location B1 estimation of region of interest can be computed by FDTD based simulation. Knowledge of field pattern estimation along coil current sensor can be used in auto-calibration PTx systems.

Oliver Kraff¹, Andrea Lazik^1,2, Daniel Brenner³, Desmond H.Y. Tse^4,5, Qi Duan⁶, Soeren Johst¹, Harald H. Quick^1,7, and Mark E. Ladd^1,8
1Erwin L. Hahn Institute for MRI, University Duisburg-Essen, Essen, Germany, ²Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Germany, ³German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, ⁴Neuropsychology and Psychopharmacology, Maastricht University, Netherlands,⁵Radiology, Maastricht University MC, Netherlands, ⁶Adv. MRI Section, LFMI, NINDS, National Institutes of Health, MD, United States, ⁷Highfield and Hybrid MR Imaging, University Hospital Essen, Germany, ⁸Medical Physics in Radiology, German Cancer Research Center (DKFZ), Germany

Three well-established B1 mapping techniques (AFI, BSS, and DREAM) were evaluated regarding quality and reproducibility for large cross-section mapping in vivo at 7 Tesla. B1 maps were obtained in 6 healthy volunteers with two standard RF shims applied (CP+ and CP2+) using an eight-channel transmit/receive RF body coil. Deviation between scan and rescan for mean B1 and maximum B1, as well as the mean ratio of zero to nonzero values within the maps were assessed to compare the methods and to rate their applicability for functional imaging of the hip joints at 7T.

Rüdiger Stirnberg¹, Daniel Brenner¹, and Tony Stöcker^1,2
1German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, ²Department of Physics and Astronomy, University of Bonn, Bonn, Germany

The use of a unique steady-state flip angle for optimal signal shaping in 3DREAM, a volumetric single-shot B₁ mapping method, is investigated. While free-induction-decay signals are shaped constantly when employing the proposed flip angle, the predictable exponential decay of stimulated echoes is shown to act as a 2D Gaussian filter on the stimulated echo image due to a spiral-out k-space view-ordering. This filter can be well estimated and employed post-hoc to the free-induction-decay image to equalize resolution properties. This is shown to improve derived B₁ maps. The findings are transferable to conventional, slice-selective DREAM B₁ mapping.

Andreas Lesch¹, Andreas Petrovic¹, and Rudolf Stollberger^1
1Department for Medical Engineering, Graz University of Technology, Graz, Styria, Austria

This work describes time varying systematic errors of the Bloch-Siegert B1 mapping technique resulting from system instabilities and describes a robust implementation. The drift in resonance frequency due to hardware imperfections is shown for different preloads of the scanner and the corresponding measurement errors are determined for the conventional Bloch-Siegert acquisition and comparison with a reference method. By appropriate implementation the adverse influences can be suppressed.

Caroline Le Ster^1,2, Giulio Gambarota¹, Eric Brillet³, Olivier Beuf⁴, and Hervé Saint-Jalmes^1,5
1INSERM UMR 1099, Université de Rennes 1, Rennes, France, ²Siemens Healthcare, Saint-Denis, France, ³Department of Imaging, Rennes University Hospital, Rennes, France, ⁴Université de Lyon, CREATIS, CNRS UMR 5220, INSERM U1044, INSA-Lyon, Université Lyon 1, Villeurbanne, France, ⁵Centre Eugène Marquis, CRLCC, Rennes, France

Quantitative MR methods often require the knowledge of the local B1 field. Many B1 mapping methods have been developed, but few of them are suited to measure low flip angles. Here we use the Low Angle Method1 coupled with EPI (LAM EPI) to acquire multislice B1 maps on a phantom and in vivo on the brain and the abdomen (breathold acquisition). We compare our results to the reference B1 mapping Double Angle Method² (DAM) and the classical LAM method (LAM FLASH).

Eamon Doyle^1,2, Jonothan Chia³, Krishna Nayak^1,4, and John C Wood^1,2
1Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, ²Cardiology, Children's Hospital of Los Angeles, Los Angeles, CA, United States,³Philips Healthcare, Cleveland, OH, United States, ⁴Electrical Engineering, University of Southern California, Los Angeles, CA, United States

Phase-based B1+ mapping methods fail in high susceptibility tissues with rapid decay. We propose using spin echo-based magnitude B1+ mapping techniques to map iron-loaded tissues incompatible with other methods.

Mohammad Mehdi Khalighi¹, Gaohong Wu², and Qin Liu^2
1Applied Science Lab, GE Healthcare, Menlo Park, CA, United States, ²MR Engineering, GE Healthcare, Waukesha, WI, United States

The adiabatic Bloch-Siegert B1 mapping is optimized for single channel transmit and channel combination transmit methods. Phantom comparison shows that single channel transmit has a better SNR for B1+ magnitude and phase. The two methods are also compared in parallel transmit application with a 3-spoke pTx RF pulse design where it is shown that single channel transmit method results in a better homogeneity. It is concluded that unlike DAM and AFI, the single channel transmit method is preferred over channel combination for B-S B1 mapping.

Douglas A C Kelley^1
1Neuro Apps and Workflow, GE Healthcare, San Francisco, CA, United States

At 7T and higher field strengths, mapping the B1 field is generally required to optimize the transmitter gain for a particular application. Characterizing these maps -- both determining whether the map is "normal" and identifying the relevant gain setting -- requires a simple, fast, and robust statistical analysis. The Kolmogorov-Smirnov test is here shown to be a good candidate for such a test.

Federico D Pineda¹, Milica Medved¹, Xiaobing Fan¹, and Gregory Karczmar^1
1Radiology, The University of Chicago, Chicago, IL, United States

Accurate measurements of the B1 field are critical in DCEMRI of the breast as small errors in the flip angle can lead to significant bias in measurements of T1, contrast media concentration, and pharmacokinetic parameters. We present a novel method for mapping the B1 field across the breast. This method is based on the fact that the T1 of fat in the breast can be accurately measured and has very low inter/intra-patient variability. The reference T1 in each image voxel is used to calculate local B1. Then the full B1 map is estimated. These maps were used to produce improved estimates of native T1

RF Pulse Design

Wednesday 3 June 2015

Laura Dupas¹, Aurélien Massire¹, Alexis Amadon¹, Alexandre Vignaud¹, and Nicolas Boulant^1
1NeuroSpin, CEA, Saclay, Ile de France, France

The simultaneous optimization of 2-spoke trajectories and RF-waveforms in the Magnitude Least Squares problem in parallel transmission is performed under explicit SAR and power constraints for axial slices of the human brain at 7 T. After making the observation that only the vector between the 2 spokes is relevant for the MLS cost-function, an active-set algorithm optimizes both the RF-weights and this vector starting from a large set of initial k-space candidates. A final normalized root mean square error of less than 2 % is systematically returned for 4 volunteers’ datasets. Bloch simulations and in-vivo T2*-weighted images validate the approach.

Alexis Amadon¹, Laura Dupas¹, Alexandre Vignaud¹, and Nicolas Boulant^1
1I2BM / NeuroSpin / UNIRS, CEA, Gif-sur-Yvette, France, France

The spoke pair k-space trajectory is known to be efficient to compensate for B1 inhomogenity in tailored slice selection at ultra high field. In the Magnitude Least Squares problem, the parameter of interest is the distance between the 2 spokes. The optimal distance is analysed across different human brains and slices at 7T. It seems to be fairly constant and its inverse is shown to roughly match the RF wavelength in the human brain.

Yunduo Li¹, Shuo Chen¹, Zechen Zhou¹, Rui Li¹, and Chun Yuan^1,2
1Center for Biomedical Imaging Research, Beijing, Beijing, China, ²Department of Radiology, University of Washington, Seattle, Washington, United States

In this study, we improved the design of velocity-selective RF pulse to compensate signal dropout due to off-resonance effect and evaluated its feasibility in carotid vessel wall imaging. Simulation and in-vivo studies showed that the new VS pulse can partially compensate the signal dropout in static tissue, which results in better image quality and contrast. As improved VS pulse widened the pass band of VS profile, which sacrificed blood suppression efficiency in case of slow blood flow causing image artifacts (Fig.4). A trade-off between off-resonance compensation and blood suppression efficiency should be considered to relief off-resonance with efficient blood suppression.

Denis Kokorin¹, Jürgen Hennig¹, and Maxim Zaitsev^1
1Department of Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany

The concept of parallel excitation allows for substantial shortening of multidimensional pulses designed for selection of arbitrarily-defined regions of interests. In this work, the use of accelerated 2D pulses is investigated experimentally for excitation of slice profiles with a limited FOV. The excitation trajectories used were based on EPI traversals in transmit k-space and were undersampled by skipping the PE lines. The feasibility of PEX was tested in a phantom on a 3T MRI system with 8 RF channels. The advantages and disadvantages of different Cartesian encoding schemes are compared based on the experimental data obtained.

Martina Flöser^1,2, Andreas Bitz¹, Sören Jost², Stephan Orzada², Marcel Gratz², Oliver Kraff², and Mark Ladd^1,2
1Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany, ²Erwin L. Hahn Institute for MRI, University Duisburg-Essen, Essen, Germany

Dynamic RF shimming can mitigate B₁⁺ inhomogeneities that hamper body imaging at high field strength. A large number of independent transmit channels offers more degrees of freedom but makes the system complex and expensive. Therefore, we propose to combine several transmit elements to adaptive elements that are excited by fully modulated Tx channels. The amplitude and phase weightings within an adaptive element, but not the pulse shape, are variable. We present a method to optimize the amplitude and phase weightings and evaluate the performance of the hybrid shimming strategy based on simulated B1 maps.

Laura Dupas¹, Alexis Amadon¹, Aurélien Massire¹, Alexandre Vignaud¹, and Nicolas Boulant^1
1NeuroSpin, CEA, Saclay, Ile de France, France

The impact of 2-spoke k-space placement is investigated for different pulse performance metrics and algorithms (under strict SAR and power constraints): the Least Squares problem with the Active-Set (AS) algorithm, the Magnitude Least Squares (MLS) problem with the Variable-Exchange method and the MLS problem with AS. After demonstrating that the MLS cost-function depends only on the vector between the two spokes in the small tip angle approximation, numerical simulations with in vivo data acquired at 7T on a human brain show that RF pulse performance is excellent and robust over surprisingly large regions of k-space.

Hadrien Dyvorne¹ and Priti Balchandani^1
1Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States

We propose a new adiabatic T2 preparation module that achieves B1-insensitive slice-selective T2 contrast by employing a modified slice-selective tunable-flip adiabatic low peak-power excitation (STABLE) RF pulse. We investigate the performance of STABLE T2 preparation against conventional non-selective T2 preparation for T2-weighted brain imaging at 7T.

Hong Shang^1,2, Peder E.Z. Larson^1,2, Adam B. Kerr³, Galen Reed⁴, Adam Elkhaled^1,2, Jeremy W. Gordon¹, Cornelius von Morze¹, Michael Lustig⁵, and Daniel B. Vigneron^1
1Radiology and Biomedical Imaging, UCSF, San Francisco, California, United States, ²UCSF-UC Berkeley Graduate Program in Bioengineering, San Francisco/Berkeley, California, United States, ³Electrical Engineering, Stanford University, Stanford, California, United States, ⁴HeartVista, Menlo Park, California, United States, ⁵Electrical Engineering and Computer Science, UC Berkeley, Berkeley, California, United States

A framework for general RF pulse design is developed based on convex optimization. It can create an RF pulse with multiband magnitude profile, arbitrary phase profile and generalized flip angle. Spectral profile sparsity is exploited to further optimize pulse characteristics such as duration, transition width and SAR with flexible trade-off among them. Designs for specialized excitation RF pulses for balanced SSFP C-13 MRI and a dualband saturation RF pulse for H-1 MR spectroscopy were developed and tested.

Oleksandr Khegai¹, Jiun-Jie Wang², Steffen J Glaser³, and Florian Wiesinger^4
1Healthy Aging Research Center, Chang Gung University, Taipei, Taiwan, ²Department of Medical Imaging and Radiological Sciences, Chang Gung University, Taipei, Taiwan,³Department of Chemistry, Technische Universität München, Munich, Germany, ⁴Diagnostics and Biomedical Technologies Lab, GE Global Research Europe, Munich, Germany

In this work we demonstrate the feasibility of the optimal control pulse sequence optimization, which allows to design as an energetically favorable RF pulse sequence that generates stable echo train with a high total or predefined MR signal at acquisition time points and brings the magnetization to the desired final distribution for certain CS offset and B1 field inhomogeneity.

Zhiyong Zhang^1,2 and Lucio Frydman^1
1Chemical Physics Department, Weizmann Institute of Science, Rehovot, Israel, ²Department of Electronic Science, Xiamen University, Xiamen, Fujian, China

Fully refocused SPatiotemporal ENcoding (SPEN) imaging has recently been shown to provide a robust alternative to EPI –erasing T₂* and chemical-shift effects from a single-scan image. Despite this cancelling of shift effects, we show how utilizing a polychromatic refocusing pulse can lead to a Fourier-encoded fully refocused SPEN sequence which can map chemical shifts with high temporal efficiency while maintaining the robustness of fully-refocused SPEN. These SPEN spectroscopic images are decoded by a Fourier transform procedure into resonance-specific images. The usefulness of this PC-SPEN approach is demonstrated on a metabolite phantom and in vivo water-fat separation imaging at 7 T.

Christoph Stefan Aigner¹, Christian Clason², Armin Rund³, and Rudolf Stollberger^1
1Institute of Medical Engineering, Graz University of Technology, Graz, Austria, ²Faculty of Mathematics, University of Duisburg-Essen, Essen, Germany, ³Institute for Mathematics and Scientific Computing, University of Graz, Graz, Austria

Radio frequency (RF) pulses are essential in MRI to excite and alter magnetization. We present a flexible approach based on optimal control of the full time-dependent Bloch equation for joint optimization of RF pulse and slice selective gradient. A globally convergent trust-region Newton method with exact derivatives via adjoint calculus allows the efficient computation of optimal pulses. The results are validated on a 3T scanner and demonstrate the ability to generate optimized waveforms for large flip angles with highly reduced RF power.

Shaihan J Malik^1,2, Anthony N Price², 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

The modulation required to produce multi-band RF pulses can lead to unacceptably high peak RF power. This can be alleviated by optimising the phase of each simultaneously excited slice. The result is typically a rapid amplitude and phase modulation. Faithful reproduction of rapid phase modulation can be error prone, especially on systems requiring frequency modulation to be defined by the pulse designer. These issues can be sidestepped if pulses are amplitude modulated (AM) only; here we identify optimal slice phases that reduce peak RF power using strictly AM MB pulses and compare them with the more general solutions.

A Alhamud¹, Jay Moore², Neal Derman¹, Ernesta Meintjes¹, and Marcin Jankiewicz^1
1Human Biology,MRC/UCT Medical Imaging Research Unit, University of Cape Town, Cape Town, Western Cape, South Africa, ²Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States

Excitation part of 2D FLASH multi-slice sequence is replaced by a train of slice-selective SLR-pulses with built-in |B₁⁺|-insensitivity. Performance of the re-designed sequence, measured in SNR gain over sequence with a standard excitation waveform, is verified in-vivo experiment in human head at 3 Tesla. The gain is substantial (up to 60% more SNR is observed) in inhomogeneity-challenged areas even when slice-profile of the standard waveform favors it over the optimized alternative . This modification to FLASH sequence represents a simple way in which image quality can be improved in various MR modalities (including fMRI, DTI and spectroscopy) in case where quality is corrupted byRF and static field inhomogeneities.

Indrajit Saha¹ and Rakesh Kumar Gupta^2
1Philips Healthcare, Philips India Ltd, Gurgaon, Haryana, India, ²fortis memorial research institute, Gurgaon, India

ewly developed MRI scan acceleration method RATE showed promises to speed-up MRI scans while preserving acquisition SNR through simultaneous acquisitions of distinct k-space phase encodes. We present our ongoing work of implementation of aliased k-space acceleration techniques in 3D MRI acquisitions with the goal of accelerating 3D dynamic MRI acquisitions. Our preliminary implementation of RATE to accelerate 3D-FFE sequence was able to produce two fold of reduction in scan time. Our ongoing work will involve developing strategy to achieve even higher acceleration factors for 3D T1 acquisitions including testing the performance of the sequence on healthy subjects.

Chanhee Lee¹, Soon Ho Yoon², Jin Mo Goo², and Jang-Yeon Park^1
1Biomedical Engineering, IBS Center for Neuroscience Imaging Research, Sungkyunkwan University, Suwon, Gyeonggi, Korea, ²Radiology, Seoul National University College of Medicine, Seoul, Korea

In this study, we show that image contrast in UTE imaging can also be affected by changing RF pulse type and its duration, not solely by changing TR, and FA, due to the magnetization-transfer (MT) effect. This was demonstrated by simulation and human brain imaging.

Hao Sun¹, Jeffrey A. Fessler¹, Douglas C. Noll², and Jon-Fredrik Nielsen^2
1Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, Michigan, United States, ²Biomedical Engineering, University of Michigan, Ann Arbor, Michigan, United States

Steady-state gradient echo sequences are generally not suitable for reduced field of view (rFOV) imaging using 3D tailored excitation pulses, due to the long duration of conventional 3D RF pulses. Recently, a method for joint RF pulse and continuous k-space trajectory design was proposed that produces tailored RF pulses capable of exciting a non-smooth 3D pattern even with a single RF coil and short pulse duration (e.g., 4 ms). However, it is unknown whether this RF pulse is sufficiently accurate for rFOV steady-state imaging since even small residual flip angles outside the rFOV can lead to relatively large steady-state signal. In this work, we evaluate using the joint RF pulse design method in for Inner Volume excitation (IVex) in three steady state sequences: RF-spoiled gradient echo (SPGR), balanced SSFP (bSSFP), and the recently proposed Small-Tip Fast Recovery (STFR) sequence. STFR can produce similar tissue signal and contrast as bSSFP, however we hypothesized that the outer-volume signal for IVex-STFR and IVex-bSSFP may differ.

Multi-Band MRI

Wednesday 3 June 2015

JaeJin Cho¹, Dongchan Kim¹, Hyunseok Seo¹, and HyunWook Park^1
1Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Chungcheong, Korea

In conventional MRI, metallic implants generate severe geometric artifacts so that images cannot be used in diagnosis. In recent years, 3D-MSI methods such as SEMAC and MAVRIC have significantly improved image quality near metallic implants. However, 3D-MSI needs very long imaging time to acquire several spectral bins for covering whole distorted magnetic field. Multiband imaging technique could be one of the solutions for shortening the imaging time of 3D-MSI. Multiband imaging acquires the signal of multiple spectral bins simultaneously so that the number of excitations decreases. In this paper, multiband 3D-MSI method will be proposed for accelerating the imaging speed.

Yajun Ma¹, Bing Wu², Wentao Liu¹, Weinan Tang¹, and Jia-Hong Gao^1
1Center for MRI, Peking University, Beijing, Beijing, China, ²GE Healthcare MR Research China, Beijing, China

we propose a RF design that may achieve both multi-band as well as SPSP selection to meet the needs of fast imaging and fat saturation while improving the time-bandwidth efficiency compared to conventional SPSP approach.

Changheun Oh¹, Dongchan Kim², and HyunWook Park^2
1Korea advanced institute of science and technology, Daejeon, Daejeon, Korea, ²Korea advanced institute of science and technology, Daejeon, Korea

In multi-band imaging technique, we propose a new multi-band imaging technique by using the RF pulse modulation with random phase to make uncorrelated aliasing pattern. To make the controlled aliasing for multi-band imaging, we make the RF pulse modulated with random phases. The phases are increasing by Golden Ratio, and the sets of phases are segmented by Fibonacci numbers and randomly reordered. Randomly distributed aliasing pattern can be reduced by using CS reconstruction. In this paper, experimental results with acceleration factor 6 are shown.

Hiroshi Toyoda¹, Naoya Yuzuriha², Sosuke Yoshinaga², and Hiroaki Terasawa^2
1Center for Information and Neural Networks, National Institute of Information and Communications Technology, Suita, Osaka, Japan, ²Department of Structural BioImaging, Kumamoto University Graduate school of Pharmaceutical Sciences, Kumamoto, Japan

The purpose of this study was to reduce ghost artifacts due to phase correction error and to reduce slice leakage artifacts in dual-band EPI, using pre-scans with a half-sized phase encoding (PE) blips technique. A phantom and in vivo rat brains were scanned using a custom single-shot dual-band 2D-EPI sequence with blipped-controlled aliasing (CAIPI) on a 7T animal scanner. The pre-scans with half-sized PE blips were useful to achieve accurate phase correction in EPI, and to estimate coil sensitivity profiles for each slice, resulting in the reduction of slice leakage artifacts in dual-band EPI with CAIPI.

Jong-Min Kim¹, Junyong Park², Chulhyun Lee², and Chang-Hyun Oh^1
1Electronic and information engineering, Korea University, Seongbuk-Gu, Seoul, Korea, ²The MRI Team, Korea Basic Science Institute, Chungchungbuk-Do, Korea

A new two-step approach to simultaneously acquire multi-slice MR images (SMI) is proposed. Simultaneously excited slices are simultaneously encoded by using Hadamard matrix (HE)1 and parallel encoding method such as SENSE2. As previously reported, HE and SENSE for SMI is useful in many applications3, 4. However these method has low temporal resolution and/or high g-factors. The proposed Hadamard and Sensitivity encoding (H-SENSE) for SMI seems to be useful to overcome these limitations by optimally choosing slice combinations for SENSE coding. Furthermore, an optimal Hadamard coding combination for better SNR with g-factor reduction.

Weiran Deng¹, Kyoko Fujimoto¹, and V. Andrew Stenger^1
1University of Hawaii JABSOM, Honolulu, HI, United States

The prolonged scan time in radial imaging can be reduced using a Simultaneous Multi-Slice (SMS) CAIPIRINHA-like acquisition. A reconstruction based on the framework of GRAPPA for SMS radial acquisition is presented. This reconstruction method is computationally fast and practical for online implementation on the MRI scanner.

Franciszek Hennel¹, Aline Seuwen¹, Constantin von Deuster¹, and Klaas P. Pruessmann^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

Separation of slices simultaneously acquired in multi-band EPI is problematic when the “N/2 ghost” needs different correction parameters for different slices. We present a straightforward manner of combining the slice-dependent ghost correction with multiband SENSE. The method inverts the sensitivity encoding equations for separately reconstructed even- and odd-echo data. It achieves a significant complexity reduction compared to the previously proposed k-space-based approach.

Peter Jan Koopmans¹, Benedikt A Poser², and Felix A Breuer^3
1FMRIB Centre, University of Oxford, Oxford, United Kingdom, ²Faculty of Psychology and Neuroscience, Maastricht University, Maastricht, Netherlands, ³Research Center Magnetic Resonance Bavaria, Wurzburg, Germany

2D-SENSE-GRAPPA is presented as a method to reconstruct in-plane accelerated simultaneous multislice data in a single step rather than sequentially unaliasing each dimension. Following a recent adaptation of Sice-GRAPPA by Setsompop et al, the method uses different kernels for the odd and even lines of k-space to ensure robustness against EPI N/2 ghosting. Reconstruction quality is very similar to Sice-GRAPPA but owing to its one-step approach, 2D-SENSE-GRAPPA has the potential to reduce calculation times which are becoming a problem in fMRI and DTI.

Mengye Lyu^1,2, Yilong Liu^1,2, Victor B. Xie^1,2, ErPeng Dai³, Hua Guo³, and Ed X. Wu^1,2
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, HKSAR, China, ²Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, HKSAR, China, ³Center for Biomedical Imaging Research, Tsinghua University, Beijing, Beijing, China

PROPELLER MRI is widely used nowadays for motion correction. One major drawback of PROPELLER MRI is the long scan time. Previous studies have investigated in-plane acceleration (SENSE and GRAPPA) in PROPELLER1,5. However, multi-band (MB) simultaneous multi-slice acquisition, without SNR penalty proportional to square root of acceleration ratio, can be a more suitable solution for accelerating PROPELLER. In addition, MB acquisition combining PINS RF pulses2 can reduce RF pulse power deposition, which is particularly useful for alleviating the SAR issue in FSE-PROPELLER at high field. A considerably low g-factor is possible in MB PROPELLER because its rotating phase encoding directions can consequently lead to a well-conditioned unwrapping problem. Another advantage of MB PROPELLER is that 3D coil sensitivity maps (CSMs) can be directly estimated from the oversampled k-space center, without acquiring additional calibration data.

Huisu Yoon¹, Dong-wook Lee¹, Juyoung Lee¹, Seung Hong Choi², Sung-Hong Park¹, and Jong Chul Ye^1
1Dept. of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Daejeon, Korea, ²Dept. of Radiology, Seoul National University College of Medicine, Seoul, Korea

The applications of dynamic compressed sensing MRI include cardiac imaging, fMRI, angiography, and perfusion imaging. Recently, by exploiting the diversities in coil sensitivity maps across the z-slice, simultaneous multislice imaging (SMS) or multi-band imaging (MB) have been extensively investigated for accelerated acquisition in brain imaging studies. By synergistically combining the two approaches, we propose a dynamic compressed sensing multi-band MR imaging technique for further acceleration in 3-D + t MR acquisition.

Parallel Imaging

Wednesday 3 June 2015

Wan Kim¹ and Yihang Zhou^1
1The State University of New York at Buffalo, Buffalo, NY, United States

We present a new iterative method using Wiener filter. In contrast to the conventional GRAPPA where only the auto calibration signals (ACS) are used to find the convolution weights, our proposed method iteratively updates the convolution weights using both the acquired and reconstructed data from previous iterations in the entire k-space. To avoid error propagation, the method applies adaptive Wiener filter on the reconstructed data. Experimental results demonstrate that even with a smaller number of ACS lines the proposed method improves the SNR when compared to GRAPPA.

Zhang Qiong¹, Sun Zhi guo¹, and Liu Wei^1
1Siemens, ShenZhen, GuangDong, China

The CAIPIRINHA (controlled aliasing in parallel imaging results in higher acceleration) technique has been successfully applied to 3D parallel imaging [1]. By shifting the phase encoding strategy, CAIPIRINHA can provide an optimal phase encoding pattern with minimum g-factor, which leads to higher SNR in comparison to the traditional GRAPPA. In this work, we demonstrate the feasibility of using CAIPIRIHNA technique for potential scan time reduction in single-slab 3D TSE (SPACE) imaging.

Mengye Lyu^1,2, Victor B. Xie^1,2, Patrick P. Gao^1,2, Yilong Liu^1,2, and Ed X. Wu^1,2
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, Hong Kong, China, ²Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, Hong Kong, China

Inter-slice shift is one of the most important factors for reducing g-factor in multi-band (MB) simultaneous multi-slice acquisition. Previous study1 has shown that the optimal distance of slice shift can deviate away from intuitive choices such as FOV/Nslice However, optimizing shift pattern is time-consuming because iterative algorithms have to be used. In this study, we propose a fast g-factor estimation model that can approximately estimate g-factor versus slice shift distance with closed form formula, with input of slice number and one parameter quantifying the coil sensitivity in slice direction.

Zhang Qiong¹ and Sun Zhi guo^1
1Siemens, Shen Zhen, Guang Dong, China

The Caipirinha concept has been successfully transferred in 3D parallel imaging [1]. By shifting the phase encoding strategy, CAIPIRINHA can provide the optimal phase encoding pattern with lowest g-factor; therefore gain higher SNR comparing with traditional GRRAPA technology. In this work, with specific tricks, we explorer the potential speed benefits of Caipirinha applying on 3D variable flip angle Tse.

Allan Raventos¹, Tao Zhang¹, and John M. Pauly^1
1Electrical Engineering, Stanford University, Stanford, California, United States

Coil compression methods combine parallel MRI data from large coil arrays into few virtual coils, and therefore significantly speed up the reconstruction. Coil compression is usually achieved by singular value decomposition, where the number of virtual coils can be determined by thresholding the singular values. However, the thresholds have to be manually tuned for different datasets or coil geometries. Here, a new approach based on noise variance estimation is proposed to automatically select the number of virtual coils. The proposed method is validated on datasets from different coil geometries.

Cishen Zhang¹ and Ifat-Al Baqee^1
1Swinburne University of Technology, Hawthorn, Victoria, Australia

The proposed direct convex optimization approach to parallel magnetic resonance imaging (pMRI) is to provide a computational algorithm for efficient processing of undersampled k-space data and accurate reconstruction of magnetic resonance images. It is based on an analysis that the magnitude image to be solved is contained in a convex hull in the solution space of the image function and the sensitivity encoded image functions. This analysis enables a novel formulation of the pMRI reconstruction problem into direct convex optimization.

Qiyuan Tian¹, Enhao Gong¹, Christoph W.U. Leuze², John Pauly¹, and Jennifer McNab^2
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States

We studied how subject motion between the acquisition of reference data and under-sampled images can affect the quality of parallel imaging reconstruction.

S. L. Talagala¹, J. E. Sarlls¹, and S. J. Inati^2
1NMRF/NINDS, National Institutes of Health, Bethesda, MD, United States, ²FMRIF/NIMH, National Institutes of Health, Bethesda, MD, United States

Recent work indicates that the temporal SNR of GRAPPA EPI data can be significantly compromised when using an EPI based GRAPPA calibration scan and that a FLASH based calibration scan can be used to correct this problem. In this work, we verify this observation with simulated data and show that phase inconsistencies in the calibration scan can lead to high g-factor and hence, lower temporal SNR in the GRAPPA EPI data. The simulations also show that the detrimental effect of calibration scan phase error is more prominent at higher SNR.

Shanshan Wang^1,2, Xi Peng¹, Jianbo Liu¹, Yuanyuan Liu¹, Pei Dong², and Dong Liang^1
1Paul C. Lauterbur Research Centre for Biomedical Imaging, Chinese Academy of Sciences, Shenzhen, GuangDong, China, ²School of Information Technologies, University of Sydney, Sydney, New South Wales, Australia

This work proposes a dictionary learning (DL) based sensitivity encoding (SENSE) approach to accurately reconstruct parallel MR images. Specifically, we regularizes the targeted image with sparse representation over an adaptive learned dictionary and formulates the reconstruction as an L2-DL minimization problem. A "divide and conquer" strategy is used to solve the proposed formulation by addressing two subproblems i.e. dictionary learning and image updating. Meanwhile, k-space data is updated as well to add more fine details back. Experimental results show that the proposed method improves the reconstruction accuracy in terms of detail preserving and outperforms the state-of-the-art SparseSENSE based approach.

Andrew T Curtis¹ and Christopher K Anand^1
1Computing and Software, McMaster University, Hamilton, Ontario, Canada

A new metric for assessing k-space sampling patterns is presented, which analyzes the smallest singular values (SSV) of the image reconstruction linear operator. The SSV is described along with an efficient means of calculation. It is compared to the gold-standard g-factor, and ranks candidate sampling patterns very similarly. SSV can assess patterns in seconds to minutes, allowing for several interesting applications. We describe one application: assessing random sampling distributions for incoherent aliasing – statistics on uniform and poisson-disk sampling patterns are easily computed over thousands of random patterns, and interesting trends arise!

Zechen Zhou¹, Jinnan Wang^2,3, Niranjan Balu³, Rui Li¹, and Chun Yuan^1,3
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, ²Philips Research North America, Briarcliff Manor, NY, United States, ³Vascular Imaging Lab, Department of Radiology, University of Washington, Seattle, WA, United States

Parallel Imaging (PI) has been widely used for MR imaging acceleration in clinical applications. However, current subspace based PI methods may not provide accurate reconstruction when it comes to spatially variant correlations due to the varying signal-to-noise characteristics. In this work, we developed a Self-supporting Tailored k-space Estimation for Parallel imaging reconstruction (STEP) technique to further improve the subspace PI reconstruction and the proposed algorithm has demonstrated its performance of reduced noise amplification, less aliasing artifacts and better structure preservation when compared to the existing PI algorithms.

Ren He¹, Jingyuan Lyu¹, and Leslie Ying^2
1Department of Electrical Engineering, University at Buffalo, Buffalo, NY, United States, ²Department of Electrical Engineering, Department of Biomedical Engineering, University at Buffalo, Buffalo, NY, United States

A novel 3D-PITA method is proposed for volumetric auto-calibrated parallel imaging. The method introduces a transition region between the ACS region and the highly reduced outer region [3], where the sampling pattern is specially designed to have a lower reduction factor than the outer region. We then perform a two-step calibration/reconstruction, one step with nonlinear GRAPPA and the other with GRAPPA, to obtain the final full k-space data. Experimental results demonstrate the proposed method is able to achieve high reconstruction quality at reduction factors higher than 5 and is superior to the conventional 3D GRAPPA.

Yuxin Hu¹, Tao Zhang², Kui Ying³, and John M. Pauly^2
1Biomedical Engineering, Tsinghua University, Beijing, Beijing, China, ²Electrical Engineering, Stanford University, CA, United States, ³Engineering Physics, Tsinghua University, China

This work generalized direct virtual coil (DVC) using SPIRiT kernel and generalized DVC can be used for arbitrary sampling pattern. In the generalized DVC, SPIRiT kernel and coil combination kernel are combined. Sensitivity map could be also estimated from coil combination kernel. Virtual coil data can be got through convolution between the combined kernel and the under sampled kspace data from all source coils. The results of the generalized DVC are nearly the same as SPIRiT and ESPIRiT and are better than those of DVC.

Wyger Brink¹, Maarten J Versluis^1,2, Johannes M Peeters², Peter Börnert^1,2, and Andrew Webb^1
1Radiology, Leiden University Medical Center, Leiden, Netherlands, ²Philips Healthcare, Best, Netherlands

The use of dielectric pads substantially improves transmit homogeneity in the thighs. However, it compromises the receive homogeneity of the body coil. If uncorrected, this can degrade image quality when using parallel imaging techniques which rely on body coil information for calibration purposes. A correction procedure using prior knowledge of the body coil reception profile confirms this effect and restores image uniformity.

Kevin Wen-Kai Tsai^1,2 and Fa-Hsuan Lin^3
1Department of Biomedical Engineering and Computational Science, Aalto University School of Science, Espoo, Finland, ²Brain Research Unit (BRU), Low Temperature Laboratory, Aalto University School of Science, Espoo, Finland, ³Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan

The under-determined inverse problem solved in InI coursed the source localization uncertainty. We proposed an anatomically constrained InI to mitigate this problem by using the high spatial resolution anatomic image as the prior information in reconstruction. Without and with the anatomical constrain in reconstruction, the spatial resolution was improved from 1.9 to 1.2 pixels in the simulation, and also has a better source localization in simulated activation at the thalamus. We conclude that anatomically constrained InI provide the better source localization accuracy than InI.

Ayano Enomoto¹ and Hiroshi Hirata^1
1Division of Bioengineering and Bioinformatics, Hokkaido University, Sapporo, Hokkaido, Japan

The purpose of this study was to improve the signal-to-noise ratio (SNR) in the parallel EPR detection scheme using multiple channels and to demonstrate the feasibility of in vivo EPR imaging with parallel EPR detection. This improvement in sensitivity leads to a decrease in the image acquisition time in three-dimensional EPR imaging. EPR imaging of 3-carbamoyl-2,2,5,5-tetramethylpyrrolidine-1-oxyl (3-CP) in a subject mouse was demonstrated with a surface coil array and a parallel detection scheme. In addition to in vivo mouse imaging, tests were also performed with phantoms filled with nitroxyl radical solutions.

Yudong Zhu^1
1Zhu Consulting, Scarsdale, NY, United States

In this work we show that one can directly identify a parallel MRI signal structure based on imaging physics, elucidate relevant rank/dimensionality with physical parameters, practice reconstruction by finding spectra/images that conform to both the signal structure and acquired spectra samples, and improve SNR by emphasizing conformity to the signal structure.

Justin P. Haldar^1
1Electrical Engineering, University of Southern California, Los Angeles, CA, United States

Low-rank modeling of local k-space neighborhoods (LORAKS) is a recent framework for constrained MRI. While LORAKS is powerful, flexible, and enables the simultaneous use of support, phase, and parallel imaging constraints, previous implementations depended on the use of time-consuming low-rank matrix completion algorithms. In this work, we show that fast LORAKS reconstructions are possible if the sampling scheme contains an autocalibration region. Results are shown with real data to demonstrate the advantages of the proposed approach relative to previous LORAKS methods. The approach can also be used as a powerful alternative to autocalibrated parallel imaging methods like SPIRiT and PRUNO.

Jingyuan Lyu¹, Yihang Zhou¹, Ukash Nakarmi¹, Chao Shi¹, and Leslie Ying^1,2
1Department of Electrical Engineering, State University of New York at Buffalo, Buffalo, NY, United States, ²Department of Biomedical Engineering, State University of New York at Buffalo, Buffalo, NY, United States

The linear model cannot describe the relationship between the missing and acquired k-space data in GRAPPA. Here we propose a more general nonlinear framework for auto-calibrated parallel imaging. In this framework, kernel tricks are employed to represent the general nonlinear relationship between acquired and unacquired k-space data without increasing the computational complexity. Identification of the nonlinear relationship is still performed by solving linear equations. We name the proposed method Kernel-based NonLinear (KerNL) method. Experimental results demonstrate that the proposed method is able to improve both image quality and computation efficiency at high reduction factors, compared with GRAPPA and nonlinear GRAPPA.

Vivek Athalye¹, Michael Lustig¹, and Martin Uecker^1
1Electrical Engineering and Computer Sciences, University of California, Berkeley, Berkeley, CA, United States

We show that parallel imaging can be formulated as an approximation of vector-valued functions in a Reproducing Kernel Hilbert Space (RKHS). This formulation provides new theoretical insights into sampling and reconstruction in k-space. In particular, we derive local bounds for the approximation error and noise amplification maps in k-space. These new metrics complement the existing g-factor maps and explain the effect of different sampling schemes on reconstruction quality. This is demonstrated for several sampling patterns using numerical experiments.

Takayuki YAMAMOTO¹, Koji FUJIMOTO¹, Tomohisa OKADA¹, Yasutaka FUSHIMI¹, Akira YAMAMOTO¹, Aurelien F. STALDER², Yutaka NATSUAKI³, Michaela SCHMIDT², and Kaori TOGASHI^1
1Diagnostic Imaging and Nuclear Medicine, Graduate School of Medicine, Kyoto University, Kyoto, Kyoto, Japan, ²Siemens Healthcare, Erlangen, Germany, ³Siemens Medical Solutions USA, Inc, Pennsylvania, United States

Due to the relatively lower CNR, application of sparse MRI techniques to non-contrast enhanced Time-of-Flight (TOF) MRA of the brain is still challenging and hence previous reports are limited. In this study, the sparse MRI technique to TOF-MRA (sparse TOF) and conventional parallel imaging (GRAPPA) was performed for 9 healthy volunteers at 3.0T MRI scanner. Two radiologists made a subjective assessment of overall impression and the visualization of distal segments of the arteries. The result indicates that sparse TOF-MRA achieved higher acceleration while maintaining visual quality and can thus be used in a clinical setting.

Karl-Heinz Herrmann¹, Anusch Mheryan², Martin Stenzel², Hans-Joachim Mentzel², Ulf Teichgräber², and Jürgen R Reichenbach^1
1Medical Physics Group, Institute of Diagnostic and Interventional Radiology, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany, ²Institute of Diagnostic and Interventional Radiology, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany

Especially in pediatric radiology locating foreign objects with CT is not desirable due to ionising radiation. A radial 3D ultra short echotime (UTE) sequence with 0.66mm isotropic resolution, both with and without the use of a long T2 tissue suppression pulse, was evaluated for wooden and glass foreign objects in a phantom (pig feet). Especially the UTE sequence with long T2 suppression seems ideal as a search modality for wood due to the excellent contrast while CT images provide very poor contrast for biological material. In the case of glass, CT image quality is cleary superior, but the UTE sequence should still be able to provide the exact location of a glass splinter in most cases.

Simon Konstandin¹ and Matthias Günther^1,2
1MR-Imaging and Spectroscopy, Faculty 01 (Physics/Electrical Engineering), University of Bremen, Bremen, Germany, ²Fraunhofer MEVIS, Bremen, Germany

Most sodium MR studies are performed using 3D ultra-short echo time (UTE) sequences with isotropic resolution. For some applications, however, anisotropic resolution is preferred to reduce partial-volume effects that hamper sodium quantification. No analytical formula could be found for the azimuth angle to acquire k-space uniformly on an ellipsoidal surface using the golden ratio (GR). In this work, a 3D acquisition scheme is presented to achieve high SNR efficiency for UTE imaging with anisotropic resolution and retrospective gating using the GR. This SNR gain can be invested in higher resolutions and/or measurement time, which is important for sodium MRI.

Ethan M Johnson¹, Urvi Vyas², Kim Butts Pauly², and John M Pauly^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States

This work describes a method for creating short-T2-selectivity in excitation without preparatory pulses and demonstrates its use with UTE imaging. The method described is also appropriate for controlling short-T2 contrast in ZTE imaging. Bone structure imaged using a multi-echo UTE sequence with multiple T2-selective excitation pulses to highlight short-T2 components. The structures represented in MRI are verified by comparison to CT images of the same subjects.

Ziyue Wu¹ and Krishna S. Nayak^1
1University of Southern California, Los Angeles, CA, United States

Radial sampling techniques are often used in dynamic MRI because they are robust to flow and motion, support short echo times, and have a diffuse aliasing pattern. One drawback is that standard implementations do not support anisotropic field-of-view (FOV). Larson et al. has provided a simple and intuitive scheme for supporting anisotropic FOV in static radial imaging. In this work,We extend that approach and demonstrate a simple solution to enable 2D anisotropic FOV with GA radial imaging, which can significantly reduce imaging times in many scenarios (abdomen, spine, etc.) where the object dimensions are anisotropic, while still allowing arbitrary temporal window reconstruction.

Naoharu Kobayashi¹, Luning Wang¹, and Michael Garwood^1
1Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota, Minneapolis, MN, United States

We introduce a novel PETRA technique with gradient modulation (GM) which enables higher readout bandwidth while keeping the missing center k-space region small. GM enables use of higher readout bandwidth while keeping excitation bandwidth relatively low, which reduces off-resonance blurring in images. The proposed sequence was evaluated by comparing PETRA techniques with and without GM for an equine knee sample in a human 7T scanner. GM-PETRA showed significantly less off-resonance blurring compared to the conventional PETRA technique under the same excitation bandwidth condition. GM can improve the PETRA image quality without any special hardware modification of clinical scanners.

Fei Han¹, Ziwu Zhou¹, Stanislas Rapacchi¹, Paul Finn¹, and Peng Hu^1
1Radiology, David Geffen School of Medicine at UCLA, Los Angeles, CA, United States

Golden angle radial reordering (GA) could provide a near-uniform k-space within a single reconstruction window with arbitrary position and duration. However, when applied with ECG-gated segmented acquisition where a single reconstruction window breaks into several temporally isolated k-space data, the k-space coverage of GA may not be as uniform as GA without ECG gating. Therefore, we sought to investigate the image artifacts caused by applying GA to ECG-gated cardiac imaging and propose a segmented GA method to address this issue.

Dallas C Turley¹ and Jim Pipe^1
1Imaging Research, Barrow Neurological Institute, Phoenix, Arizona, United States

This work presents an implementation of through-time GRAPPA in the Distributed Spirals trajectory to improve temporal resolution of 3D dynamic acquisitions.

James G Pipe^1
1Imaging Research, Barrow Neurological Institute, Phoenix, Arizona, United States

This work presents a new parallel imaging method. It is a k-space convolution method that is quite independent of the sampling geometry, and does not require training, and therefore is quite suitable for nonCartesian sampling. The method is described, and 2D and 3D examples are shown.

Robert W. Stobbe¹ and Christian Beaulieu^1
1University of Alberta, Edmonton, Alberta, Canada

A novel multi-shot 'Yarn-Ball' technique is presented which facilitates efficient 3D k-space sampling through large loops with low acceleration. The differential equations underpinning this technique are described, and the number of trajectories required to fully sampling 3D k-space are compared to standard gradient echo. Unlike gradient echo, trajectory requirements can be dramatically reduced by >100x with readout duration increase. In the context of 3D spoiled steady-state imaging, Yarn-Ball allows the generation of much higher resolution images than is possible with standard gradient echo for the same TR and scan duration.

Maria Engel¹, Nadia Benkhedah¹, and Armin M. Nagel^1
1Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany

This study developed a 3-D density adapted spiral sequence, which handles the high dependency on the slewrate limit of the MRI scanners gradient system. It was found to attain a radial fraction of up to 0.21, leading to a time saving of almost 80% compared to commonly used radial projection sequences. ²³Na-MRI of head was conducted using a 7-Tesla whole body MR system.

Zhiqiang Li¹, Dinghui Wang¹, John P Karis², and James G Pipe^1
1Imaging Research, Barrow Neurological Institute, Phoenix, AZ, United States, ²Neuroradiology, Barrow Neurological Institute, Phoenix, AZ, United States

T2-weighted TSE is a routine clinical tool for neuroimaging. TSE generates slightly different contrast than conventional SE. TSE can also incorporate the mDixon technique to provide fat/water imaging, but at the cost of prolonged scan time. In this project, we propose a spiral SE sequence for enhanced T2 contrast and fast scan speed. Preliminary results showed increased T2 contrast with spiral SE in tissues with increased ion deposition. It’s also demonstrated that the SNR of spiral SE is close to TSE even its scan time is shorter. Therefore, spiral SE is a potential alternative to TSE for T2-weighted imaging.

Jinil Park¹, Tae-Hoon Shin², and Jang-Yeon Park^1
1Biomedical Engineering, IBS Center for Neuroscience Imaging Research, Sungkyunkwan University, Suwon, Gyungki-do, Korea, ²Diagnostic Radiology and Nuclear Medicine, University of Maryland, Baltimore, Maryland, United States

In this study, we propose a modified version of Wong's method that is able to almost maintain the uniformity of k-space coverage even in the event that imax increases, providing an analytic form for the coordinates of all views in terms of the readout gradient strengths in the x-, y-, and z- axes. Our suggestion was validated by simulation and phantom experiments.

Claudiu Schirda¹, Tiejun Zhao², Ovidiu Andronesi³, James Mountz¹, Fernando Boada¹, and Hoby Hetherington^1
1Radiology, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States, ²Siemens Medical Solutions, Pittsburgh, PA, United States, ³Radiology, Massachusetts General Hospital, Boston, MA, United States

The rosette trajectory design flexibility is used to decrease the demands on scanner gradient system for the Rosette Spectroscopic Imaging (RSI) sequence. The reduced gradient readout strength and slew rate results (in addition to decreased Eddy currents and decreased frequency drift) in decreased acoustic noise and decreased scanner vibrations, which improves patient comfort. A 2D RSI acquisition with in-plane resolution of 8mm and a 20x20x12 3D RSI acquisition with nominal isotropic resolution of 8mm and effective voxel size of 0.55cc are demonstrated in-vivo with slew rate of 45mT/m/ms and gradient of 5.8mT/m in 32 sec and 5-10 mins, respectively.

Signe Johanna Vannesjo¹, Nadine N Graedel², Lars Kasper¹, Simon Gross¹, Christoph Barmet^1,3, and Klaas P Pruessmann^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland, ²FMRIB Centre, University of Oxford, Oxford, United Kingdom, ³Skope Magnetic Resonance Technologies, Zurich, Switzerland

Spiral imaging theoretically yields more time-efficient k-space coverage and increased robustness against flow and motion compared to Cartesian acquisitions. In practice however, the use of spiral acquisitions has been hampered by gradient system limitations causing deviations to the sampling trajectories. It has recently been proposed to perform image reconstruction based on trajectories predicted by a linear time-invariant model of the gradient system. Here, the method is shown to yield high-quality reconstructions of single-shot spiral images. We furthermore show high reproducibility of the gradient system characterization over years. The method holds promise to yield spiral imaging feasible on standard MR systems.

Lyu Li¹, Xiaodong Ma¹, Pascal Spincemaille², Yi Wang^2,3, Huijun Chen¹, and Hua Guo^1
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, ²Radiology, Weill Cornell Medical College, New York, United States, ³Biomedical Engineering, Cornell University, New York, United States

High resolution dynamic volumetric MRI is crucial for the accurate clinical diagnosis. Several studies have proposed some data acquisition and image reconstruction strategies such as TRACER and GRASP. The reconstruction of TRACER is sensitive to motion so breath hold is needed. GRASP is very motion insensitive, but its data acquisition efficiency is not high enough. In this study, we aim to investigate a method combining golden angle variable density spiral data acquisition with sparse and parallel imaging reconstruction, which could achieve high spatial and temporal resolution when free breathing.

Wenwen Jiang¹, Frank Ong², Roland Henry³, Michael Lustig², and Peder E.Z. Larson^3
1Bioengineering, UC Berkeley/UCSF, Berkeley, CA - California, United States, ²EECS, UC Berkeley, Berkeley, California, United States, ³Radiology and Biomedical Imaging, UCSF, San Francisco, CA - California, United States

Ultrashort echo time (UTE) imaging has shown promise for imaging tissues with ultrashort T2 values. UTE with a 3D radial trajectory requires π times the number of phase-encodings as the Cartesian counterpart, requiring long scan times. Practically, ~π times undersampling can be used for acceleration without noticeable streak aliasing artifacts with gridding reconstruction. However, the undersampling-induced aliasing results in noise-like artifacts. We propose using L1-ESPIRiT on 3D UTE for effectively removing streak and noise-like undersampling artifacts, and allowing for high acceleration factors with robustness. The in vivo applications for brain imaging shows the benefits of L1-ESPIRiT for 3D UTE.

Dmitry Kurzhunov¹, Robert Borowiak^1,2, Philipp Wagner¹, Marco Reisert¹, and Michael Bock^1
1Department of Radiology · Medical Physics, University Medical Center Freiburg, Freiburg, Baden-Württemberg, Germany, ²German Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), Heidelberg, Baden-Württemberg, Germany

In this work iterative proton-constrained reconstruction of ¹⁷O MR data set using an anisotropic non-homogeneous diffusion operator is proposed. It was tested on a ¹⁷O brain phantom and on a data set obtained from a dynamic ¹⁷O MRI experiment. ¹⁷O images show a higher SNR and superior quality compared to Kaiser-Bessel re-gridding method. Localized CMRO₂ quantification with ¹H constrained in a human frontal lobe and in a brain phantom show most precise CMRO₂ values in GM and WM. For GM it is in good agreement with literature ¹⁵O-PET data.

KC Erb¹, Ganesh Adluru¹, Srikant Kamesh Iyer¹, Devavrat Likhite¹, John A Roberts¹, and Edward DiBella^1
1UCAIR, University of Utah, Salt Lake City, Utah, United States

Pre-reconstruction cartesian-interpolation (gridding) algorithms are used to interpolate non-cartesian data onto a cartesian grid for simple cartesian reconstruction. In this work, we compare 3 such interpolation methods to the Non-Uniform Fast Fourier Transform (NUFFT). Two of the methods are simple single-coil interpolators (nearest neighbor and 3-point interpolation) and the other is a newer multi-coil method called GRAPPA Operator Regridding (GROG). We demonstrate that GROG is able to produce images which are quantitatively more similar to NUFFT reconstructed images than the other single-coil methods. We also demonstrate that GROG continues to outperform the other methods, even on highly undersampled datasets.

Boris Mailhe¹, Qiu Wang¹, Robert Grimm², Marcel Dominik Nickel², Kai Tobias Block³, Hersh Chandarana³, and Mariappan S. Nadar^1
1Imaging and Computer Vision, Siemens Corporation, Corporate Technology, Princeton, NJ, United States, ²MR Application & Workflow Development, Siemens Healthcare, Erlangen, Germany, ³Department of Radiology, New York University School of Medicine, New York, NY, United States

Density compensation is a mandatory step for direct reconstruction of radial MRI data. We interpret density compensation as a left-hand-side preconditioner of the measurement operator. We propose an alternative formulation as a right-hand-side preconditioner compatible with regularized iterative reconstruction. In the case of under-sampled radial trajectories, we show that a ramp filter overemphasizes high frequencies. Instead, we calibrate the preconditioner offline. We show that preconditioning accelerates the reconstruction and improves the sharpness of the reconstructed images.

Encoding & Reconstruction

Wednesday 3 June 2015

Esther Raithel¹, Gaurav Thawait², Shivani Ahlawat², Shadpour Demehri², Zhang Qiong³, and Jan Fritz^2
1Siemens AG, Healthcare Sector, Erlangen, Bavaria, Germany, ²Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Maryland, United States, ³Siemens AG, Guang Dong, China

While 2D TSE MRI is standard in musculoskeletal MRI, similar performance was shown for 3D TSE-type sequences (SPACE) before; however, long acquisition times may limit its clinical use. In contrast to one-dimensional acceleration, a 2x2 CAIPIRINHA pattern can yield substantial acceleration of 3D data acquisition. We demonstrate the implementation of a 4-fold CAIPIRINHA acceleration pattern into a 3D SPACE sequence with a resultant 50-60% decrease of the acquisition time when compared to a standard 2-fold accelerated SPACE. A 5-min CAIPIRINHA SPACE sequence can produce near identical image quality in the knee as 11-min SPACE and 10-min three plane 2D TSE sequences.

Peter James Ross¹ and David J. Lurie^1
1Aberdeen Biomedical Imaging Centre, University of Aberdeen, Aberdeen, Aberdeen City, United Kingdom

In this work we describe a new method of field-cycling relaxometric imaging based on the view-sharing technique known as keyhole imaging. We have employed this technique to generate images containing a new form of endogenous contrast based on the variation of R1 with the external magnetic field. Relaxometric imaging results are presented from a phantom study and a human volunteer in order to demonstrate the accuracy and practicality of the method.

Deqing Chen¹ and Weiguo Zhang^1
1Shanghai United Imaging Healthcare Co. Ltd., Shanghai, Shanghai, China

Ambiguity in global polarity can lead to wrong final image contrasts in PSIR. Current methods for determining polarity either rely on a priori assumptions or require reference data that add significant scan time, making them susceptible to errors under diverse clinical circumstances. Here we report an efficient, reliable and fully automatic method for determining the global polarity for PSIR imaging. The method makes no assumptions and requires practically no extra time.

Jia Guo¹, Yongchuan Lai¹, Xiaocheng Wei¹, Nan Cao¹, and Bing Wu^1
1GE Healthcare, Beijing, China

This method makes landmark very simple, and provides coverage knowledge of floating coil. The overall scan was just over 10 seconds, and all the information were gathered and processed during the move-in of table, no extra time was added to the scan process and when the operator returns to the operator console, the graphical information is already available on the screen for placing localizer. In this way, improper coverage of localizer scan, unnecessary body contacts with patients and blind knowledge of the location of the coil are all avoided. This method was also silent in nature that ensures patient¡¯s comfort.

Dimitris Mitsouras¹, Onur Afacan², Robert V Mulkern³, and Dana H Brooks^4
1Radiology, BWH/Harvard Medical School, Boston, MA, United States, ²Children's Hospital Boston, MA, United States, ³Childrens' Hospital Boston, MA, United States,⁴Northeastern University, Boston, MA, United States

We systematically compared undersampled variable density sampling schemes in conjunction with linear and compressed sensing (CS) reconstruction methods. Simulated and experimental results showed trajectory choice had minimal effect on carefully-defined SNR and RMSE metrics compared to reconstruction method, and that while CS outperformed linear reconstruction, there is tradeoff in terms of noise-driven uncertainty of individual image values that is spatially related to the transform domain chosen. The CS transform domain and reconstruction parameter selection is likely important to avoid uncertainty in clinically-relevant image features; for example, minimum-L1 reconstruction should be avoided when signal hyper-intensities in homogenous hypo-intense backgrounds are relevant.

Sulaiman A Al Hasani¹, Gary F Egan², and Jingxin Zhang^3
1Electrical and Computer Systems Engineering, Monash University, clayton, VIC, Australia, ²Monash Biomedical Imaging, Monash University, VIC, Australia, ³School of Software and Electrical Engineering, Swinburne University of Technology, VIC, Australia

The reconstruction quality of CS is highly dependent on the level of signal (image) sparsity as well as the level of incoherent In conventional MRI, Fourier encoding concentrates the energy of MR signal in the center of k-space, this limits the incoherent sampling and hence hinders the performance of CS reconstruction. In this work, we propose a practical encoding scheme, based on the spread spectrum analysis of Chirp modulated Fourier sensing matrix, to enhance the incoherent sampling of multi-receive MRI. The proposed method outperforms Fourier encoding in preserving image quality at high acceleration factors.

Satoshi Ito¹, Shungo Yasaka¹, and Yoshifumi Yamada^1
1Utsunomiya University, Utsunomiya, Tochigi, Japan

Sparse MRI has been introduced to reduce the acquisition time and raw data size by randomly undersampling the k-space data.　However, the image quality depends on the randomness or signal trajectory in k-space even if the reduction factor of signal is the same. In other words, the best signal trajectory depends on the object to be imaged, however, it is impossible in general to know the best signal trajectory for unknown image data. In this paper, we propose a novel image reconstruction technique in which undersampled signal at equal interval is adopted in image reconstruction. To reduce the aliasing artifact due to equally-spaced undersampled signal, we used the phase-scrambler in the acquisition.

Ehsan Hamtaei^1,2, Saifeng Liu³, Yongquan Ye², Dongmei Wu⁴, and E. Mark Haacke^1,2
1MR Innovations Inc., Detroit, MI, United States, ²Radiology, Wayne State University, Detroit, MI, United States, ³MRI Institute of Biomedical Research, Ontario, Canada,⁴East China Normal University, Shanghai, China

A new approach is proposed for improving the Partial Fourier reconstruction in asymmetrically collected double echo gradient echo data with two reverse polarity echoes. Comparison is made with conventional Partial Fourier reconstruction (POCS). The high resolution phase estimate obtained using the proposed method, helps POCS converge more efficiently.

Somaie Salajeghe¹, Paul Babyn², Jonathan C. Sharp³, and Gordon E. Sarty^1
1Division of Biomedical Engineering, University of Saskatchewan, Saskatoon, SK, Canada, ²Medical Imaging, University of Saskatchewan, Saskatoon, SK, Canada,³Department of Oncology, University of Alberta, Edmonton, AB, Canada

Transmit array spatial encoding (TRASE) uses spatial RF phase gradients in place of main field gradients to encode image information. When the RF gradients are linear the image may be reconstructed using the Fourier transform. Since linear gradients may be somewhat expensive to achieve, we investigated the possibility of using least squares to reconstruct data from non-linear TRASE. Our results indicate that this is a feasible approach.

Haifeng Wang¹, R. Todd Constable¹, and Gigi Galiana^1
1Yale University, New Haven, CT, United States

Nonlinear spatial encoding magnetic (SEM) fields have been studied to reduce the number of echoes needed to reconstruct a high quality image, but optimal schemes are still unknown. Previously, we showed that adding a rotating nonlinear field of modest amplitude, which we call the FRONSAC (Fast ROtary Nonlinear Spatial Acquisition) imaging, greatly improved the reconstructions obtained from highly undersampled conventional linear trajectories. However, since the ultimate goal is to acquire these highly undersampled trajectories in a single short TR, still lower amplitude FRONSAC gradients are desirable. FRONSAC creates undersampling artifacts that are relatively incoherent and well suited to CS reconstruction. Compressed sensing (CS) is a sparsity-promiting convex algorithm to reconstruct images from highly undersampled datasets. In this paper, we present a hybrid, CS-FRONSAC, which combines these two methods. The simulation results illustrate that the proposed method improves incoherence between the sensing and sparse domains, and it ultimately improves image quality compared with results recovered by the Kaczmarz algorithm. The resulting improvement allows us to consider FRONSAC gradients with lower amplitudes and frequencies, lowering hardware demands as well as dB/dt burden.

Jonathan I. Tamir¹, Weitian Chen², Peng Lai², Martin Uecker¹, and Michael Lustig^1
1Electrical Engineering and Computer Sciences, University of California, Berkeley, Berkeley, CA, United States, ²Global Applied Science Laboratory, GE Healthcare, Menlo Park, CA, United States

Scan efficiency plays a major role in 3D Fast Spin Echo because of the tradeoff between image blurring and SNR. The number of echo trains during a scan is limited by the TR interval, which is often much longer than the prescribed echo train length (ETL). Here we explore the benefit of using the extra time in the TR to increase the ETL. We show that a model-based, subspace-constrained reconstruction is able to use the additional data to increase SNR. The approach has a fixed SNR cost independent of ETL compared to a reconstruction where the relaxation parameters are known.

Aymeric Stamm¹, Onur Afacan², Benoit Scherrer², Jolene M Singh¹, and Simon K Warfield^1
1Computational Radiology Laboratory, Department of Radiology, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States, ²Boston Children's Hospital, Harvard Medical School, Boston, MA, United States

High resolution imaging has an important role to play in the detection of malformations of cortical development, and can dramatically improve the prospect of a surgical cure in epilepsy, or in the detection of brain lesions. We propose a novel, unified reconstruction strategy that overcomes prior limitations and show that the strategy can be applied to imaging the stria of Gennari in vivo in a reasonable amount of time on a conventional 3T magnet. Our proposed technology enables image reconstruction from inter-session and intra-session k-space data, that overcomes intra-session and inter-session phase variation to enable arbitrarily high SNR imaging.

Emre Kopanoglu¹, Haifeng Wang¹, Yuqing Wan¹, Dana C. Peters¹, Gigi Galiana¹, and Robert Todd Constable^1
1Diagnostic Radiology, Yale University, New Haven, Connecticut, United States

Nonlinear gradient fields encode spatial information along multiple directions simultaneously. Using a receiver-array, the spatial information along different directions can be recovered. O-Space imaging exploits this idea to perform accelerated data acquisitions. In standard O-Space imaging, the full field-of-view is encoded, which results in acquisition of redundant data when the region-of-interest is a sub-volume. In this study, we propose a novel O-Space implementation that focuses the encoding effort to the region-of-interest. With a typical O-Space acquisition, the acceleration factor may exceed the number of receiver coils whereas the proposed method can further reduce the imaging time by up-to 25%.

Weitian Chen¹, Rob Peters², Suchandrima Banerjee¹, Misung Han³, Roland Krug³, Garry Gold⁴, and Yuval Zur^5
1Global Applied Science laboratory, General Electric, Menlo Park, CA - California, United States, ²Global Applied Science laboratory, General Electric, Waukesha, WI, United States, ³Department of Radiology and Biomedical Imaging, UCSF, San Francisco, CA - California, United States, ⁴Radiology, Stanford University, Palo Alto, CA - California, United States, ⁵Healthcare Magnetic Resonance, General Electric, Haifa, Israel

Fast spin echo (FSE) imaging plays a central role in clinical MRI. When combined with flip angle modulation, very long echo train can be used without excessive blurring, which makes 3D FSE imaging feasible in clinical setting. FSE requires CPMG condition. However, this condition can be violated in a number of applications due to system imperfection and result in image artifacts. We recently proposed a non-CPMG FSE acquisition based on a phase cycling method. However, this approach doubles scan time. Here we discussed the methods to reduce the scan time of this approach.

Lin Chen¹, Congbo Cai², Shuhui Cai¹, and Zhong Chen^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²Department of Communication Engineering, Xiamen University, Xiamen, Fujian, China

Reducing the number of measurements required by Nyquist sampling theorem is one way to accelerate MRI acquisition at the cost of introducing aliasing artifacts. Various approaches have been proposed to eliminate aliasing artifacts, such as parallel imaging and non-Cartesian sampling. However, the higher hardware requirement of parallel imaging and complicated trajectory design and image reconstruction of non-Cartesian sampling limit their wide applications. In this abstract, we propose an MRI approach based on quadratic phase encoding, which can accelerate acquistion with efficient aliasing artifacts suppresion using a single receiver coil and uniform sampling.

Seong Dae Yun¹ and N. Jon Shah^1,2
1Institute of Neuroscience and Medicine, Medical Imaging Physics (INM-4), Forschungszentrum Juelich, Juelich, Germany, ²Faculty of Medicine, Department of Neurology, JARA, RWTH Aachen University, Aachen, Germany

EPIK had been previously validated both at 1.5T and 3T. The method was shown to provide a higher temporal resolution and less image distortions than single-shot EPI whilst maintaining comparable performance for the detection of BOLD-based signals. This work aims to perform high-resolution visual fMRI based on EPIK and evaluate its performance in direct comparison to comparable EPI; each method was accelerated with parallel imaging and partial Fourier techniques for resolution improvement. The obtained results showed that EPIK outperformed single-shot EPI in terms of imaging resolution and capability of identifying relatively small functional regions such as LGN and SC.

Vipul R Sheth¹, Jacopo Annese¹, Hongda Shao¹, Qun He¹, Jody Corey-Bloom², Graeme M Bydder¹, and Jiang Du^1
1Radiology, University of California, San Diego, CA, United States, ²Neurosciences, University of California, San Diego, CA, United States

An inversion recovery ultrashort echo time (IR-UTE) MRI pulse sequence is used to detect myelin lesions and iron deposition in multiple sclerosis patients. The mechanism by which the IR-UTE pulse sequence detects myelin and iron deposition is explored.

Kerrin J Pine¹, Gareth R Davies¹, and David J Lurie^1
1Aberdeen Biomedical Imaging Centre, University of Aberdeen, Aberdeen, Scotland, United Kingdom

In field-cycling MRI, the normally stable main magnetic field B₀ is manipulated during an examination to explore the field strength dependency of relaxation rate. Clinical adoption relies on finding alternatives to T₁ dispersion imaging which is associated with lengthy scan times. A pulse sequence is described for rapid measurement of T₁dispersion with one-dimensional projection. Results are presented demonstrating the 1D spatial variation of dispersion in a phantom. The sequence could be used to rapidly locate pathology of known dispersion for further inspection by other methods.

Image Processing and Analysis

Wednesday 3 June 2015

Diego Cantor-Rivera¹, Terry M. Peters², and Ali R. Khan^2
1Biomedical Engineering Graduate Program, Western University, London, ON, Canada, ²Medical Biophysics, Western University, London, ON, Canada

This work presents a novel multivariate asymmetry analysis for investigating focal structural abnormalities. The novel method uses multi-parametric imaging data non-rigidly registered to a symmetric template to estimate asymmetry measures using locally-sampled cumulative distribution functions (Kolmogorov-Smirnov test). We applied it to investigate structural abnormalities in temporal lobe epilepsy using quantitative relaxometry, diffusion tensor imaging, and voxel-based morphometry. Whole brain Mahalanobis distance maps were employed in a support vector machine classification to show that the use of asymmetry significantly improves discrimination between temporal lobe epilepsy patients and healthy controls.

Shuo Han¹ and Daniel A. Herzka^1
1Department of Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, Maryland, United States

Analytical phantoms with closed form solutions for their Fourier Transforms (FT) enable accurate and arbitrary sampling of k-space. Though most phantoms are limited to simple shapes such as rectangles or ellipsoids, we have recently presented an expression for the 3D FT of a polyhedron of uniform intensity. This work provides the mathematical framework that extends the original polyhedral FT to account for linear gradients in intensity. A computationally efficient implementation is demonstrated.

Xiaofei Sun¹, Lin Shi^2,3, Yishan Luo¹, Winnie CW Chu¹, and Defeng Wang^1,4
1Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Shatin, NT, Hong Kong, ²Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Shatin, NT, Hong Kong, ³Chow Yuk Ho Technology Centre for Innovative Medicine, The Chinese University of Hong Kong, Shatin, NT, Hong Kong,⁴Department of Biomedical Engineering and Shun Hing Institute of Advanced Engineering, The Chinese University of Hong Kong, Shatin, NT, Hong Kong

Intensity of similar tissues on brain MRIs is often inhomogeneous because of the various acquisitions. It is problematic since the analysis of MR images (registration, segmentation and volumes statistics) may depend on the hypothesis that corresponding anatomical locations have a similar intensity level. In this study, a new hybrid approach based on Gaussian mixture model and histogram matching to normalize for intensity differences on MR images is presented. This method does not require spatial alignment. The effectiveness of intensity normalization is validated on real data, and the results show that intensity normalization significantly improves the accuracy of tissues segmentation results.

Guillaume Bonnier^1,2, Jean-Philippe Thiran², Gunnar Krueger^1,2, Tobias Kober^1,2, Bénédicte Mortamet^1,2, Cristina Granziera^1,3, and Alexis Roche^1,2
1Siemens ACIT – CHUV Radiology, Siemens Healthcare IM BM PI & Department of Radiology CHUV, Lausanne, Vaud, Switzerland, ²LTS5, École Polytechnique Fédérale de Lausanne, Lausanne, Vaud, Switzerland, ³Department of Clinical Neurosciences, Laboratoire de recherche en neuroimagerie and Neuroimmunology Unit, Lausanne, Vaud, Switzerland

Gray-matter (GM) segmentation in brain MRI is challenged by the presence of more than one tissue type in a voxel, an effect known as partial volume. This problem is particularly evident in regions such as the lateral part of the central nuclei (CN) or in the cerebellum. In this study, we used GM concentration maps showing good qualitative results compared with histological data to improve GM segmentation and quantitative analysis in the CN and cerebellum. Our results showed that GM analysis using concentration maps has higher sensitivity than binary masks to detect T1 changes.

Chenguang Peng¹, Huayu Zhang¹, Jinchao Wu¹, Xingfeng Shao^1,2, Yingmao Chen³, Quanzheng Li⁴, Georges El Fakhr⁴, and Kui Ying^1
1Key Laboratory of Particle and Radiation Imaging, Ministry of Education, Department of Engineering, Beijing, China, ²Department of Bioengineering, UCLA, California, United States, ³Department of Nuclear Medicine, The general hospital of Chinese People's Liberation, Beijing, China, Beijing, China, ⁴Department of Radiology, Division of Nuclear Medicine and Molecular Imaging, Harvard Medical School, Boston, United States

With the implement of PET-MRI, simultaneously MRI and PET image acquisition becomes real. Researchers have been wondering how to use high resolution MRI image to improve the PET image quality. In this work, we proposed a iterative residual based deconvolution method using MRI information for PET partial volume effect correction which is one of the major causes for PET low spatial resolution. Proposed method perform well in lesion contrast enhancement without introducing additional noise to image. Both phantom simulation and in-vivo result is shown in this work.

Mary C. Kociuba¹ and Daniel B. Rowe^1,2
1Department of Mathematics, Statistics, and Computer Science, Marquette University, Milwaukee, Wisconsin, United States, ²Department of Biophysics, Medical College of Wisconsin, Milwaukee, Wisconsin, United States

To remove variability, from image acquisition artifacts and unwanted physiological sources, inherent within an acquired functional MRI (fMRI) signal, the data is preprocessed before statistical analysis is performed. Although, it is well known that preprocessing operations modify a voxel’s temporal spectrum and induce correlations. In this study, spatial correlations are described in terms of overlapping temporal frequency content between voxels. Identifying the location and extent of the induced correlations is the first step to compensate for unwanted variability, leading to more accurate interpretations from the data.

Ernesto Cuartas-M¹, Angel Torrado-C^2,3, Juan A Hernandez-T^2,3, José Ángel Pineda⁴, Eva Manzanedo-S², and German Castellanos-D^1
1Universidad Nacional de Colombia, Manizales, Caldas, Colombia, ²Medical Image Analysis and Biometry Lab, Rey Juan Carlos University, Madrid, Spain, ³Madrid-MIT M+Vision Consortium, Madrid, Spain, ⁴Centre for Biomedical Technology-U.P.M, Pozuelo de Alarcón, Spain

We analyzed the influence of neglect important tissue in the EEG/MEG forward problem using an anisotropic finite difference method. Specifically, we study the EEG/MEG dipole source estimation in the presence of an anisotropic blood vessels model. We used patient specific head models MRI-based, together with DWI tensor imaging to calculate anisotropic tensors in the white matter. The results shows that neglect the anisotropic blood vessels may induce significant errors in the source estimation for deep brain areas. It could potentially be an important drawback for source localization in focal temporal epilepsy.

Chenguang Peng¹, Jinchao Wu¹, Xingfeng Shao^1,2, Yingmao Chen³, Quanzheng Li⁴, Georges El Fakhr⁴, and Kui Ying^1
1Key Laboratory of Particle and Radiation Imaging, Ministry of Education, Department of Engineering, Beijing, China, ²Department of Bioengineering, UCLA, California, United States, ³Department of Nuclear Medicine, The general hospital of Chinese People's Liberation, Beijing, China, Beijing, China, ⁴Department of Radiology, Division of Nuclear Medicine and Molecular Imaging, Harvard Medical School, Boston, United States

PET image has relatively low resolution, and partial volume effect (PVE) is one of the major reason for this. Hybrid PET-MRI system offers us a great opportunity to synthesis PET and MRI, and MRI can provide PET with high resolution anatomical image which is a great supplementary information to correct PVE. In traditional method, point spread function is usually used to described PVE and is regarded as a spatial invariant constant which is not true in real PET system. In this work, we proposed a new method that considered spatial variant point spread function.

Taejoon Eo¹, Jinseong Jang², and Dosik Hwang^2
1Yonsei University, Seoul, Seoul, Korea, ²Yonsei University, Seoul, Korea

This study demonstrates that the accurate T2* weighted images can be obtained from highly undersampled Cartesian MGRE data by using WEST. While conventional method such as model-based fitting or CS are work with the reduction factor up to 5, the WEST can increase the reduction factor up to 16 in Cartesian trajectory. This method will be well applied for other various multi-contrast imaging such as DTI, SWI, etc.

Michael Amann¹, Simon Pezold², Yvonne Naegelin³, Ketut Fundana², Michaela Andelova³, Katrin Weier³, Christoph Stippich⁴, Ludwig Kappos³, Philippe Cattin², and Till Sprenger^1
1Neurology/Neuroradiology, University Hospital Basel, Basel, BS, Switzerland, ²Medical Image Analysis Center (MIAC), University of Basel, Basel, BS, Switzerland, ³Neurology, University Hospital Basel, Basel, BS, Switzerland, ⁴Neuroradiology, University Hospital Basel, Basel, BS, Switzerland

The "cord image analyzer (cordial)" pre-segments the spinal cord (SC) based on the continuous max-flow-approach, combined with a cross-sectional similarity prior. The SC surface is reconstructed by locating SC boundary based on image intensities. The cervical SC volume (CSCV) is defined by cutting planes perpendicular to the SC centerline. Reliability of cordial was assessed on controls with coefficients-of-variation of less than 1%. Applicability to clinical data was tested on a cohort of 48 MS patients. Multiple regression analysis revealed significant relation between CSCV loss and EDSS change.

Xiang Feng¹, Andreas Deistung¹, Jesper Hagemeier², Michael Dwyer², Robert Zivadinov^2,3, Juergen R. Reichenbach¹, and Ferdinand Schweser^2,3
1Medical Physics Group, Institute of Diagnostic and Interventional Radiology, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany, ²Buffalo Neuroimaging Analysis Center, Dept. of Neurology, School of Medicine and Biomedical Sciences, State University of New York at Buffalo, Buffalo, NY, United States, ³MRI Molecular and Translational Imaging Center, Buffalo CTRC, State University of New York at Buffalo, Buffalo, NY, United States

In this study we show that the linear registration associated to the standard FIRST pipeline fails in the case of abnormal anatomy, potentially introducing a serious bias to clinical studies. To overcome this issue, we present an improved framework for FIRST segmentation that incorporates both a hybrid contrast generation and a non-linear registration.

Christian Scharfenberger¹, Dorothy Lui¹, Farzad Khalvati², Alexander Wong¹, and Masoom Haider^2,3
1Systems Design Engineering, University of Waterloo, Waterloo, Ontario, Canada, ²Department of Medical Imaging, University of Toronto, Toronto, Ontario, Canada, ³Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada

The contouring and segmentation of the prostate gland is an important task in computer-aided prostate cancer screening using MRI. To assist medical professionals with the segmentation process, we propose a novel user-guided approach to prostate segmentation in MR images. The approach optimizes the energy components of a modified Decoupled Active Contour framework based on a Hidden Markov Model and a Rician likelihood to explicitly consider user guidance and textural and anatomical priors. Extensive experiments based on 10 patient cases and a variety of evaluation metrics showed that our approach provides a significant improvement over an existing semi-automatic segmentation approach.

Prashant Nair¹, Rajagopal K V¹, Rolla Narayana², Indrajit Saha³, and Satish M^1
1KMCH Hospital, Manipal University, Manipal, India, ²Philips Healthcare, Philips India Ltd, Bangalore, India, ³Philips Healthcare, Philips India Ltd, Gurgaon, Haryana, India

The purpose of our study was to compare 3D SHINKEI based MRN with the DWIBS neurography in the anatomical visualization of the brachial plexus at 1.5T. MRN image quality generated by 3D SHINKEI and DWIBS in five healthy volunteers was studied and was scored by two radiologists. While both the techniques were scored similar in visualization of the root, SHINKEI based MRN scored higher in depicting the trunk and cord area of brachial plexus with respect to DWIBS neurogrphy. Our ongoing work will include further optimization of SHINKEI based MRN at 1.5 T to improve brachial plexus visualization.

Novel Computing Frameworks

Wednesday 3 June 2015

Kelvin Layton¹, Stefan Kroboth¹, Jochen Leupold¹, Huijun Yu¹, Feng Jia¹, Sebastian Littin¹, Tony Stöcker², and Maxim Zaitsev^1
1Medical Physics, University Medical Center Freiburg, Freiburg, BW, Germany, ²German Center for Neurodegenerative Diseases, Bonn, NRW, Germany

This work presents a sequence programming environment that is vendor-independent and supports rapid sequence prototyping through simulation and acquisition. The environment provides a drag-and-drop programming interface that allows researchers to run sequences on any hardware platform. Central to this work, is a new hardware-independent sequence file format that can easily be converted into hardware-dependent instructions for execution on an MR scanner. This is demonstrated by using a single sequence file to obtain experimental data using scanners from two different MR manufacturers. The improved workflow dramatically reduces sequence development time, provides new teaching opportunities and promotes vendor-independence across institutions.

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

Combining compressed sensing (CS) MRI with parallel imaging can reduce the scan time and/or improve reconstruction quality. However, the iterative reconstruction algorithm required by compressive sensing is time-consuming. Several groups have reported using graphics processing units (GPUs) to accelerate CS reconstruction. However, none has been applied to CS-MRI with parallel imaging. This paper presents a method that uses an alternating direction algorithm and GPUs for CS reconstruction from parallel receive channels, which is particularly suitable for large array data. Experiments show that it takes less than a second to reconstruct a 128×128×16 3-D image from 8-channel data, which is more than 20 times faster than a quad-core, high-end commodity CPU.

Martin Uecker¹, Frank Ong¹, Jonathan I Tamir¹, Dara Bahri¹, Patrick Virtue¹, Joseph Y Cheng², Tao Zhang², and Michael Lustig^1
1Electrical Engineering and Computer Sciences, University of California, Berkeley, Berkeley, CA, United States, ²Department of Radiology, Stanford University, Stanford, United States

The high complexity of advanced reconstruction algorithms poses challenges for development and application of new reconstruction methods. Here, we present the Berkeley Advanced Reconstruction Toolbox, a framework for iterative image reconstruction. It consists of a programming library and a toolbox of command-line programs. The library provides common numerical operations and important algorithms including generic implementations of several iterative optimization algorithms. It supports parallel computation using multiple CPUs and GPUs. The command-line tools provide direct access to a wide range of functionality from basic operations on multi-dimensional arrays to complete implementations of advanced calibration and reconstruction algorithms for MRI.

Kyunghyun Sung^1,2, Di Wu³, Fei Han^1,2, Ziwu Zhou^1,2, Peng Hu^1,2, Holden Wu^1,2, Alex Bui^1,2, and Jason Cong^3
1Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, ²Bioengineering, University of California, Los Angeles, Los Angeles, CA, United States, ³Computer Science, University of California, Los Angeles, Los Angeles, CA, United States

Compressed sensing (CS) is an emerging technique that can reduce the imaging time at the cost of increased reconstruction runtime (i.e. time to reconstruct MRI images from raw k-space data). New algorithmic and acceleration approaches are needed to translate CS-based MRI methods into clinical practice. We propose to develop a new low cost CS-based MRI reconstruction platform using customized CPU accelerated implementation.

Eric A. Borisch¹, Joshua D. Trzasko¹, Adam T. Froemming², Roger C. Grimm¹, Akira Kawashima², Armando Manduca¹, Phillip M. Young², and Stephen J. Riederer^1
1Mayo Clinic, Rochester, MN, United States, ²Radiology, Mayo Clinic, Rochester, MN, United States

A description of the successful implementation and integration into clinical workflow of a faster-than-acquisition sparse sampling reconstruction is presented, with DCE perfusion imaging of the prostate as an example application. A 4D time series with 55 timeframes is acquired with 2D SENSE acceleration in under 6 minutes, with the reconstruction completing in under 5 minutes; a reconstruction time of less than 5 seconds per 3D volume. Improved SNR and retained sharpness are observed relative to a traditional SENSE reconstruction of the same acquired data.

Fei Han¹, Ziwu Zhou¹, Kyunghyun Sung¹, J Paul Finn¹, and Peng Hu^1
1Radiology, David Geffen School of Medicine at UCLA, Los Angeles, CA, United States

Despite tremendous recent advances in non-linear image reconstruction methods, their clinical utility has been significantly held back by the limited computational power and development flexibility provided by the MRI system manufacturer. We developed a MR image reconstruction framework in which any custom made algorithms can be implemented on custom built computer nodes connected to the MR scanner such that k-space data is sent to the nodes for calculation and reconstructed images are sent back to the scanner system as part of the default pipeline. In this work, we demonstrate the benefit of this framework on 12 pediatric congenital heart disease patients who underwent cardiac MRI using a highly accelerated 4D cardiac phase-resolved contrast-enhanced MRA sequence.

Image Quality Assessment

Wednesday 3 June 2015

Robin Antony Birkeland Bugge¹, Atle Bjørnerud¹, Wibeke Nordhøy¹, and Øystein Bech Gadmar^1
1Intervention Center, Oslo University Hospital, Oslo, Oslo, Norway

Quality assurance is essential both for MRI clinic and for research. Yet, the extent to which QA is performed varies significantly. The Norwegian Radiation Protection Authority made recommendations in 2005 that urged all clinics and centers using MRI to have at their disposal a quality assurance methodology that is independent of the MRI vendor. The MRI_QAP is an open source software designed to address these issues, expand on the existing procedures, remove subjective assessments and ensure consistent reports. The software consist of various seperated modules that perform different tasks; Geometric distortion, Signal to noise ratio, uniformity and dynamic stability analysis.

Brian Hanna¹, Naoharu Kobayashi¹, Djaudat Idiyatullin¹, Curtis Andrew Corum¹, Brad Weegman¹, Jinjin Zhang¹, and Michael Garwood^1
1Radiology, University of Minnesota, Minneapolis, MN, United States

A study of image quality metrics and calibration for MRI. We compared reconstruction methods and measured instrument calibration using low cost Lego phantoms with a short T2 relaxation time.

Kunio Nakamura¹, Vladimir S. Fonov¹, Nicolas Guizard¹, Sridar Narayanan¹, Douglas L. Arnold¹, and D. Louis Collins^1
1Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada

We have developed simulate MRI datasets with simulated atrophy that allows accuracy evaluation of image analysis methods in detecting volume changes in whole brain, gray matter, and white matter. We have simulated 3 types of atrophy (central, peripheral, and combination). We applied standard techniques as well as our in-house method to detect brain volume changes and quantitatively compared these methods. The results showed that SIENA in FSL tended to overestimate central brain atrophy and that tensor-based morphometry developed locally performed well in the combination of atrophy. The dataset will be publicly available as a part of BrainWeb.

Thomas Küstner^1,2, Parnia Bahar², Christian Würslin¹, Sergios Gatidis¹, Petros Martirosian³, Nina Schwenzer¹, Holger Schmidt¹, and Bin Yang^2
1Department of Radiology, University Hospital of Tübingen, Tübingen, Baden-Württemberg, Germany, ²Institute of Signal Processing and System Theory, University of Stuttgart, Stuttgart, Baden-Württemberg, Germany, ³Diagnostic and Interventional Radiology, University Hospital of Tübingen, Tübingen, Baden-Württemberg, Germany

A reliable and meaningful image quality assessment can be very demanding, especially when there is no reference or gold-standard available. Evaluation mainly depends on human observers, but due to the huge amount of acquired data, this task can be very time-consuming and costly. Hence an automatic evaluation is desired. We therefore propose a robust, accurate and flexible automatic evaluation system which is based on a machine-learning approach to evaluate certain diagnostic questions dependent on the chosen application and trained input data. Our framework achieves a test accuracy of 91.2% and hence can be used for automatic quality classification.

Bénédicte Maréchal^1,2, Stephan Kannengiesser³, Kaely Thostenson⁴, Peter Kollasch⁵, Pavel Falkovskyi^1,2, Jean-Philippe Thiran², Reto Meuli⁶, Matt A. Bernstein⁴, and Gunnar Krueger^1,2
1Siemens ACIT – CHUV Radiology, Siemens Healthcare IM BM PI & Department of Radiology CHUV, Lausanne, Switzerland, ²LTS5, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, ³Siemens Healthcare, Erlangen, Germany, ⁴Department of Radiology, Mayo Clinic, Rochester, MN, United States, ⁵Siemens Healthcare, MN, United States,⁶CHUV Radiology, Lausanne, Switzerland

Automated quality assessment of MRI is of great importance to derive reliable diagnostic information. In this work, a synthetic noise-based method is proposed which allows automated data quality classification. Only a “prescan” measurement of noise and a single image acquisition are required. The validation based on 764 head scans confirms the robustness and reliability of the method. As integrated as a prototype in online image reconstruction, it can greatly improve clinical workflow as MR technologist is provided with immediate feedback and can potentially repeat low-quality scans within the same session.

Marcel Gratz^1,2, Maximilian Völker², Sören Johst², Mark E Ladd^2,3, and Harald H Quick^1,2
1High Field and Hybrid MR Imaging, University Hospital Essen, Essen, Germany, ²Erwin L. Hahn Institute for Magnetic Resonance Imaging, University of Duisburg-Essen, Essen, Germany, ³Medical Physics in Radiology, German Cancer Research Center, Heidelberg, Germany

An image-based approach to quantify the performance and long-term stability of parallel transmit (pTx) MR systems is presented that uses a specific target excitation. Semi-automatic analysis is performed with a custom-written Matlab software that automatically segments the acquired image and aims to obtain reference points as well as characteristic image properties that are compared to target parameters. Thus, the system performance can be tracked with multiple parameters quantitatively over time and compared to other pTx systems without further requirements. Moreover, fluctuations and drifts of the transmit power as well as timing problems may be detected.

Li Sze Chow¹, Raveendran Paramesran¹, and Martyn Paley^2
1Electrical Engineering, University of Malaya, Kuala Lumpur, Wilayah Persekutuan, Malaysia, ²Academic Radiology, University of Sheffield, Sheffield, South Yorkshire, United Kingdom

A new No-Reference Image Quality Assessment (NR-IQA) model called Blind/referenceless image spatial quality evaluator (BRISQUE) uses scene statistics of locally normalized luminance coefficients to quantify possible losses of “naturalness” in the image. Structural Similarity Index (SSIM) is one of the most commonly used Full-Reference IQA (FR-IQA) method. This study employed the BRISQUE and SSIM methods in comparison in assessing the quality for four types of MR optic nerve images. It was verified that BRISQUE is more appropriate than SSIM in evaluating the quality of MR images; however modification will be required to re-train the SVM library.

Ian Marshall¹, Gabriel Rilling¹, Yuehui Tao², Chaoran Du¹, Samarth Varma¹, Dominic Job¹, Andrew Farrall¹, and Mike Davies^1
1University of Edinburgh, Edinburgh, United Kingdom, ²University of Oxford, Oxford, United Kingdom

Compressed Sensing (CS) has been shown to provide significant speed up of MRI with sufficient accuracy when judged by 'image error' against known ground truths. To be clinically useful, the images must also be of diagnostic quality and acceptable to radiologists. Very few reports have considered this crucial issue. In this study of 3D brain scanning, we found that radiologists rated 4-times and 6-times undersampled data reconstructed by CS higher than conventional least squares reconstruction. However, subtle artefacts made interpretation of deep brain structures difficult and caused significant differences in measurement of brain tissue volumes.

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

Multi-component T2 relaxation is widely used to measure the myelin water fraction, but it can also be applied to measure the total water content (TWC). This work is the first to estimate theoretical errors in T2-based TWC measurement with simulations, and determine the impact of factors including the signal to noise ratio, flip angle inaccuracies (B1+ inhomogeneity), and Rician noise on the accuracy of TWC estimation. Simulations demonstrated that TWC could be measured with T2 to a high degree of accuracy (<3%), even in the presence of B1+ inhomogeneity and Rician noise, and suggestions are given to improve accuracy further.

Dictionary-Based Reconstruction

Wednesday 3 June 2015

Zhitao Li¹, Benjamin Paul Berman², Diego R Martin³, Maria I Altbach³, and Ali Bilgin^1,4
1Electrical and Computer Engineering, University of Arizona, Tucson, Arizona, United States, ²Applied Mathematics, University of Arizona, Tucson, Arizona, United States,³Department of Medical Imaging, University of Arizona, Tucson, Arizona, United States, ⁴Biomedical Engineering, University of Arizona, Tucson, Arizona, United States

Accurate parameter estimation in MR Fingerprinting (MRF) requires large dictionaries with many atoms, each with thousands of time points. Storage of such dictionaries require large memory and the matching process becomes increasingly demanding with increasing dictionary size. We propose a Tree Structured Vector Quantizer based clustering approach for MRF dictionary design. The proposed approach allows significant reduction in dictionary dimensions and can enable clinically relevant reconstruction accuracy and time which is a major bottleneck for clinical usefulness of MRF.

Kanwal K Bhatia¹, Anthony N Price^2,3, Joseph V Hajnal^2,3, and Daniel Rueckert^1
1Biomedical Image Analysis Group, Imperial College London, London, United Kingdom, ²Centre for the Developing Brain, Kings College London, London, United Kingdom,³Biomedical Engineering Department, Kings College London, London, United Kingdom

This work describes an algorithm for fast dynamic MRI reconstruction based on random sampling and manifold-based interpolation. Results are demonstrated on both cardiac real-time and cardiac cine MRI, showing how the proposed algorithm performs at varying undersampling rates. The algorithm is fast, reconstructing 200 frames of real-time MRI in under 60 seconds, and can also be used as an initialisation to speed up convergence of more complex compressed sensing strategies.

Zhifang Zhan¹, Yunsong Liu¹, Jian-Feng Cai², Di Guo³, Jing Ye¹, Zhong Chen¹, and Xiaobo Qu^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²Department of Mathematics, University of Iowa, Iowa City, Iowa, United States, ³School of Computer and Information Engineering, Xiamen University of Technology, Xiamen, Fujian, China

Compressed sensing (CS) exploit the sparsity of magnetic resonance (MR) images to realize accurate reconstruction from the undersampled k-space data. Recently, there has been a growing interest in the study of adaptive sparse representation of MR images to achieve better reconstructions. However, most adaptive dictionary training processes are based on all the image patches and usually time-consuming. In this work, we proposed a fast dictionaries learning method that takes advantage of the geometric directions in classified similar patches. Experiments on T2-weighted brain image data show our proposed method improve the reconstruction quality both on reducing artifacts and minimizing reconstruction errors.

Benjamin Paul Berman¹, Mahesh Bharath Keerthivasan², Zhitao Li², Diego R. Martin³, Maria I. Altbach³, and Ali Bilgin^2,4
1Program in Applied Mathematics, University of Arizona, Tucson, Arizona, United States, ²Electrical & Computer Engineering, University of Arizona, Tucson, Arizona, United States, ³Medical Imaging, University of Arizona, Tucson, Arizona, United States, ⁴Biomedical Engineering, University of Arizona, Tucson, Arizona, United States

A non-Cartesian and multi-channel method of dictionary learning and compressed sensing reconstruction leads to improved T2 parameter mapping. The imaging problem is constrained to have a sparse representation within a dictionary. The principal components of the T2 decay are reconstructed, and the addition of the dictionary constraint leads to a reduction in noise and artifacts.

Jinhong Huang^1,2, Xiaohui Liu¹, Wufan Chen¹, and Yanqiu Feng^1
1Guangdong Provincial Key Laborary of Medical Image Processing, School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China, ²School of Mathemtics and Computer Science, Gannan Normal University, Ganzhou, Jiangxi, China

Image reconstruction employing adaptive sparsifying transform has demonstrated promising performance in compressed sensing magnetic resonance imaging. However, conventional overcomplete dictionary learning based methods are computationally expensive. In this work, we present a novel sparsity-promoting orthogonal dictionary updating method (SPODU) for efficient image reconstruction from highly undersampled MRI data. To further improve reconstruction, the deterministic annealing like strategy is combined into the algorithm. Experimental results demonstrate that the proposed SPODU algorithm is more efficient and accurate than the dictionary learning based method which using K-SVD for sparse coding, and thus has potential application in practice.

Chao Shi¹, Yihang Zhou¹, Yanhua Wang¹, Dong Liang², Xiaojuan Li³, and Leslie Ying^1,4
1Electrical Engineering, University at Buffalo, SUNY, Buffalo, New York, United States, ²Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institute of Advanced Technology, Shenzhen, Guangdong, China, ³Department of Radiology and Biomedical Imaging, University of California at San Francisco, San Francisco, California, United States, ⁴Biomedical Engineering, University at Buffalo, SUNY, Buffalo, New York, United States

We propose a novel method to accelerate parameter mapping in this abstract. We model MR parameter mapping as a problem to recover a parametric manifold. An iterative algorithm is proposed to recover the manifold from undersampled data based on the parametric model. The method iteratively alternates between reconstructing the image series and recovering the parameters. Both simulation and experimental results demonstrate the potential of highly accelerated parameter mapping by the proposed method.

Imaging Near Metal

Wednesday 3 June 2015

Valentina Taviani¹, Daniel Litwiller², Kevin M. Koch³, and Brian A. Hargreaves^1
1Radiology, Stanford University, Stanford, CA, United States, ²GE Healthcare, Rochester, MN, United States, ³Biophysics and Radiology, Medical College of Wisconsin, Milwaukee, WI, United States

Outer Volume Suppression (OVS) using high-bandwidth quadratic-phase RF pulses was used for reduced-FOV 2D Multispectral Imaging (MSI) near metal. OVS gradient reversal and a modified acquisition strategy were used to minimise unintentional saturation of off-resonance near metal. The technique was demonstrated in a phantom consisting of a metal implant embedded in agar. Clinical applications of the proposed method for spine imaging and MR-guided FUS (Focused Ultrasound Surgery) planning were presented. The ability to reduce the encoded FOV translated in shorter echo train lengths, i.e. reduced blurring, at the expense of slightly reduced SNR.

Jetse van Gorp¹, Chris Bakker^1,2, Job Bouwman¹, Jouke Smink³, Frank Zijlstra¹, and Peter Seevinck^1
1Image Sciences Institute, University Medical Center Utrecht, Utrecht, Utrecht, Netherlands, ²Department of Radiology, University Medical Center Utrecht, Utrecht, Utrecht, Netherlands, ³Philips, Best, Noord-Brabant, Netherlands

In this work a compressed sensing (CS) accelerated 3D phase encoded turbo-spin echo (3D-PE-TSE) sequence was used to obtain geometrically undistorted images in the presence of titanium in a clinically acceptable time frame (<10min). By ordering the acquired echo’s in spherical layers with an equal number of sampling points it was possible to combine TSE and CS acceleration to achieve a factor 60 scan time acceleration compared to a 3D-PE-SE scan. Phantom and in vivo results confirm that the presented method can be used for investigations where spatial accuracy is required.

Theresa Bachschmidt^1,2, Johanna Schöpfer³, Stephan Biber², Peter Jakob¹, and Mathias Nittka^2
1Department of Experimental Physics 5, University of Würzburg, Würzburg, Germany, ²Magnetic Resonance, Siemens AG, Erlangen, Germany, ³Corporate Technology, Siemens AG, Erlangen, Germany

Spinal fusion is a common surgery, which often requires follow-up treatments. However, the compensation of metal-induced through-plane distortion is time-consuming and its application in clinical routine is limited. Distortion through-plane is even enhanced at 3T compared to 1.5T and demands reduction before it is corrected; it scales inversely with the bandwidth of the RF pulse. Local transmit coils feature higher peak B1, hence enable higher RF bandwidths, than whole-body birdcage coils. This work presents the capability of a prototype Tx/Rx spine coil with respect to minimizing through-plane distortion. RF bandwidths of 3.3 kHz can decrease time-consuming artifact compensation significantly.

Bragi Sveinsson¹, Valentina Taviani¹, Garry Gold¹, and Brian Hargreaves^1
1Radiology, Stanford University, Stanford, CA, United States

Multi-spectral Imaging (MSI) allows imaging close to metal implants at the cost of increased scan time. Previous work has demonstrated how MSI scan time can be shortened by hexagonally undersampling ky/kz-space and removing the resulting aliased replicas using a zeroing mask. This required the extra step of the operator providing the approximate location of the implant to the reconstruction software. In this work, we demonstrate how the detection of the implant location can be fully automated by hexagonally sampling adjacent slices in a complementary manner and combining the bins to get an estimate of the slice profile with minimal aliasing.

Jonathan Paul Ashmore¹, Mathias Nittka², Lyndall Blakeway³, Steve Connor^1,3, and Geoff Charles-Edwards^3
1Neuroradiology, King’s College Hospital NHS Foundation Trust, London, London, United Kingdom, ²Siemens Healthcare, Erlangen, Germany, ³Guy’s & St Thomas’ NHS Foundation Trust, London, United Kingdom

A growing number of MR conditional cochlear implants can now be scanned safely without requiring temporary surgical removal of the internal magnet. However, with the magnet in situ the resultant images have large areas of signal drop-out and distortion around the device which may obscure nearby structures of interest, even with spin echo sequences and large receiver bandwidths. The aim of this work was to evaluate two new options available on a clinical MRI scanner for reducing metal artefacts, View Angle Tilting (VAT) and Slice Encoding for Metal Artifact Correction (SEMAC)

Dongchan Kim¹, JaeJin Cho¹, Kinam Kwon¹, and HyunWook Park^1
1Electrical engineering, KAIST, Daejeon, Yuseong-Gu, Korea

MRI is one of the most powerful imaging modalities for clinical diagnosis. However, the use of MRI is limited for the patients who have metallic implants, because metallic implants causes severe field inhomogeneity, which causes the geometrical distortion in image and the slice profile. To resolve these problems, slice encoding for metal artifact correction (SEMAC) technique was proposed. But SEMAC is not practical technique due to the long imaging time. In this work, we propose a sensitivity information based metal artifact correction (MAC) technique to reduce the imaging time for multi-contrast and time-series imaging.

Kevin M Koch^1
1Biophysics and Radiology, Medical College of Wisconsin, Milwaukee, WI, United States

3D Multi-Spectral Imaging (3D-MSI) techniques, such as MAVRIC, SEMAC, and MAVRIC SL collect multiple independently spatial-encoded images at discrete Larmor frequency offsets. Prospective knowledge of the range and location of off-resonant spins can aid in optimizing the efficiency of these sequences. Here we present a concept of 3D-MSI calibration scanning that can provide such information through an acquisition on the order of 1 minute. In addition, we demonstrate a practical method to automatically determine off-resonance frequency ranges determination using such calibration data. For many implant cases of lower magnetic susceptibility materials, such a calibration and off-resonance range computation can substantially improve the efficiency of 3D-MSI techniques.

shun qi¹, Ying Liu¹, Langlang Gao¹, Panli Zuo², Mathias Nittka³, and Hong Yin^1
1Xijing Hospital, Fourth Military Medical University, xian, shaanxi, China, ²Siemens Healthcare, MR Collaborations NE Asia, shaanxi, China, ³Siemens Healthcare, Germany, Germany

MR images with WARP sequences significantly reduced metal-related artifacts and improved delineation of the prosthesis and periprosthetic region therefore increased the diagnostic sensitivity in patients with clinical abnormities.

Daehyun Yoon¹ and Brian A Hargreaves^1
1Radiology, Stanford University, Palo Alto, CA, United States

A denoising algorithm to improve complex summation of spectral images for Slice Encoding for Metal Artifact Correction (SEMAC) sequence is presented. In SEMAC, multiple spectral images are collected, and combined together to image spins with a huge resonance frequency variation around metallic implants. The complex summation has not been often used for combining these spectral images because of a serious SNR degradation even though its image sharpness around the metal is better than other combination methods. Here we introduce a new image combination algorithm to improve the SNR for the complex summation to provide both sharpness and high SNR.

Laura J. King¹, Philip J. Bones¹, and Rick P. Millane^1
1Department of Electrical and Computer Engineering, University of Canterbury, Christchurch, New Zealand

The three-point Dixon technique can be used to successfully suppress fat near metal implants if the phase shift due to the B0 field inhomogeneities can be estimated accurately. This research describes a new technique where the phase unwrapping is guided by prior knowledge of the implant. Using an approximate model of the implant location and orientation, the phase unwrapping is simplified to estimating the difference between the model and true phase shift. The model is rotated and translated to minimise the phase unwrapping required. Initial results obtained from simulation are shown.

Andrew M. Huettner¹, Andrew S. Nencka¹, L.Tugan Muftuler², and Kevin M. Koch^3
1Biophysics, The Medical College of Wisconsin, Milwaukee, Wisconsin, United States, ²Neurosurgery, The Medical College of Wisconsin, Milwaukee, Wisconsin, United States,³Biophysics and Radiology, The Medical College of Wisconsin, Milwaukee, Wisconsin, United States

This work demonstrates nonlinear constrained numerical optimization for RF pulse design applied to a pulse sequence for multi-spectral imaging around metal implants. By reducing the peak B1 of the refocusing pulse though numerical optimization, the maximum tip angle permitted on the scanner, based on hardware and RF safety limits, was increased from 140 degrees to 180 degrees.

Xeni Deligianni^1,2, Thomas Egelhof², Thorsten Wischer², Reinhard Elke², and Oliver Bieri^1
1Radiology, Division of Radiological Physics, University of Basel Hospital, Basel, NA, Switzerland, ²Merian Iselin Klinik, Basel, NA, Switzerland

Magnetic resonance imaging (MRI) in the vicinity of metal hardware is stll characterized by long scan times for the acquisition of composite images and low fluid sensitivity. Here we propose an optimized composite fast steady state free precession sequence (COFIsp) as a fast, clinically feasible alternative for acquisition of fluid-sensitive and distortion-free images in the presence of metal. For distortion correction additional prephasing gradients and optimized phase encoding orientation were investigated. The sequence allows good artifact correction and high fluid contrast which is very important for the visualization of inflammation.

Elastography

Wednesday 3 June 2015

Shantanu Warhadpande¹, William Kenyhercz², Priyanka Illapani², Brian Raterman³, Joshua Dowell³, Michael Go³, Patrick Vaccaro³, Jean Starr³, Richard White³, and Arunark Kolipaka^3
1The Ohio State University College of Medicine, Columbus, OH, United States, ²The Ohio State University, Columbus, OH, United States, ³The Ohio State University Wexner Medical Center, OH, United States

It is known that changes in Abdominal Aortic Aneurysms (AAA) wall stiffness (WS) can reflect extra-cellular matrix integrity, a key factor in the pathophysiological development of AAA and the risk for rupture. Therefore, our study applied Magnetic Resonance Elastography (MRE) to determine aortic WS in 10 AAA patients. Our hypothesis was that there would be no relationship between aortic WS and AAA diameter and our results supported the hypothesis with a poor linear correlation of R2 =0.0051. In the future, a more personalized approach to AAA management using MRE-derived WS might be warranted to better assess the risk for rupture.

Joshua Trzasko¹, Kevin Glaser¹, Arvin Arani¹, Armando Manduca¹, David Lake¹, Phillip Rossman¹, Shivaram Poigai Arunachalam¹, Kiaran McGee¹, Richard Ehman¹, and Philip Araoz^1
1Mayo Clinic, Rochester, MN, United States

In magnetic resonance elastography (MRE), mechanically induced motion is estimated from a time-encoded series of phase-contrast images and used to generate quantitative spatial maps of tissue stiffness. Like most dynamic/parametric applications, MRE has flexibility regarding acquisition parameter assignment, particularly with respect to motion encoding gradients (MEG). In this work, we derive the Cramer-Rao Lower Bound (CRLB) for the complex harmonic signal — which completely describes mechanically induced motion in single-frequency, steady-state MRE — and use this to define performance limits of experimental MRE setups. This can serve as an objective tool for developing and comparing different protocols. Using this bound, we then identify minimum number of data samples needed for complex harmonic estimation to be well-posed.

Armando Manduca¹, David S Lake¹, Khang T Huynh¹, Rehman S Eon¹, Elizabeth M Annoni¹, and Richard L Ehman^1
1Physiology and Biomedical Engineering, Mayo Clinic, Rochester, MN, United States

In magnetic resonance elastography, the calculated stiffness values are affected by noise, which is amplified by the inversion process. It would be useful in practice to establish that, beyond some SNR threshold, stiffness calculations are accurate within some confidence limit. A variety of SNR measures have been proposed for MRE, including variations of displacement SNR and octahedral shear strain SNR. We demonstrate here that the proper SNR measure depends on the inversion algorithm used, and, more precisely, on the order of derivatives in the inversion process, and show that Laplacian-based SNR measures are required for commonly used Helmholtz inversions.

P Kennedy¹, L MacGregor², E Barnhill¹, A Cooper¹, L Hiscox¹, C Brown³, J Braun⁴, I Sack⁴, E van Beek¹, A Hunter², CL Johnson⁵, and N Roberts^1
1Clinical Research Imaging Centre (CRIC), University of Edinburgh, Edinburgh, United Kingdom, ²School of Sport, University of Stirling, Stirling, United Kingdom, ³The Mentholatum Company Ltd., Glasgow, United Kingdom, ⁴Department of Radiology, Charité - Universitätsmedizin Berlin, Berlin, Germany, ⁵Beckman Institute, University of Illinois at Urbana-Champaign, Urbana, IL, United States

Exercise Induced Muscle Damage (EIMD) is associated with force reduction, inflammation and pain after exercise and is often accompanied by Delayed Onset Muscle Soreness (DOMS). In this study significant variation in muscle stiffness, measured using Magnetic Resonance Elastography (MRE), and force output are determined following an eccentric exercise protocol. Subjective pain assessment revealed significant pain increase following damage indicating the presence of DOMS. The rectus femoris muscle group is primarily affected with localised oedema present on T2 weighted imaging.

Anthony Joseph Romano¹, Jing Guo², Michael Scheel², Sebastian Hirsch², Juergen Braun³, and Ingolf Sack^2
1Physcial Acoustics, Naval Research Laboratory, Washington, DC, United States, ²Radiology, Charite-Universitatsmedizin, Berlin, Germany, ³Medical Informatics, Charite-Universitatsmedizin, Berlin, Germany

Waveguide Elastography combines Magnetic Resonance Elastography, Diffusion Tensor Imaging and Orthotropic inversions for the analysis of the anisotropic properties of white matter. Previously, this method was applied to the Cortico-spinal tracts of healthy human volunteers and to patients diagnosed with Amyotrophic Lateral Sclerosis. Here we apply this approach to the classification of the anisotropic properties of the 12 major white matter tracts in the human brain. This method is intended as a new modality to assess white matter structure and health by means of the evaluation of the anisotropic elasticity tensor of nerve fibers.

Kevin Glaser¹, Jun Chen¹, and Richard Ehman^1
1Radiology, Mayo Clinic, Rochester, MN, United States

MR elastography (MRE) is a technique for noninvasively assessing hepatic fibrosis that has a high diagnostic accuracy and is safer and less expensive than biopsy. Typically, MRE requires multiple breath holds of 10-20 seconds, which may not be possible for some patients and can add several minutes to clinical liver MRI exams. In this work, we demonstrate an implementation of 2D hepatic MRE that can be performed in a single 3.2-second breath hold or while free breathing that can significantly reduce the number and duration of breath holds required for performing MRE and improve clinical patient throughput.

Roger Grimm¹, Eric Stinson¹, and Richard Ehman^1
1Mayo Clinic, Rochester, MN, United States

Elastography sampling with gradient echo sequences can lead to long scan times. Scan efficiency is limited by gradient heating concerns and the need to synchronize the TR to the period of motion excitation. These inefficiencies can be ameliorated by using a phase advancing TR. This work uses a phase advancing TR and acquires the minimum 7 samples to provide complex sampling of the wave field along all 3 axes. The proposed method uses a half-NEX like reconstruction to combine the conjugate and non-conjugate samples.

Jing Guo¹, Sebastian Hirsch¹, Jürgen Braun², and Ingolf Sack^3
1Radiology, Charité - Universitätsmedizin Berlin, Berlin, Berlin, Germany, ²Department of Medical Informatics, Charité - Universitätsmedizin Berlin, Berlin, Germany,³Radiology, Charité - Universitätsmedizin Berlin, Berlin, Germany

An anisotropic MR elastography method is introduced which relies on shear fields in incompressible transversely-isotropic media. The method provides three independent moduli, μ₁₂, μ₁₃ and E₃, the shear moduli in the planes of isotropy and symmetry as well as the Young's modulus parallel to the fiber axis, respectively. The method is demonstrated on the soleus in ten healthy volunteers and reveals that 1) E₃>μ₁₃>μ₁₂ and 2) elasticity parameters along the muscle fiber axis are higher in the right leg compared to the left leg which is most likely resulting from the right leg dominance of soleus function in our group.

Arvin Arani¹, Ondrej Slezak¹, Nikoo Fattahi¹, Kevin J Glaser¹, Joel Felmlee¹, Armando Manduca², Clifford R. Jack¹, Richard L. Ehman¹, and John Huston III^1
1Radiology, Mayo Clinic, Rochester, Minnesota, United States, ²Physiology and Biomedical Engineering, Mayo Clinic, Rochester, Minnesota, United States

Several groups have investigated the use of MR elastography in neurological diseases, using a single vibration frequency or the use of multi-frequency data to acquire information about the viscoelastic properties of brain tissues. Although the precision, repeatability, and sensitivity to changes in these viscoelastic parameters has been reported in phantoms and in vivo, the accuracy of either the single-frequency or multi-frequency stiffness measurements has not been thoroughly evaluated experimentally. The objective of this work was to evaluate the accuracy of both single-frequency and multi-frequency brain MRE in a geometrically accurate brain stiffness mimicking phantom and compare it with mechanical testing.

Samuel Patz^1,2, Katharina Schregel³, Iga Muradyan^1,2, Angelos Kyriazis^1,2, Jens Wuerfel^3,4, Srini Mukundan^1,2, and Ralph Sinkus^5
1Brigham & Women's Hospital, Boston, MA, United States, ²Harvard Medical School, Boston, MA, United States, ³Institute of Neuroradiology, University Medicine Goettingen, Goettingen, Germany, ⁴NeuroCure, Charité University Medicine, Berlin, Germany, ⁵Imaging Sciences & Biomedical Engineering, Kings College, London, United Kingdom

Using Magnetic Resonance Elastography (MRE), we observed changes in mouse brain cortical stiffness associated with changes in the amplitude of an externally applied acoustic stimulus.

Jules Nelissen^1,2, Larry de Graaf¹, Tom Schreurs^1,2, Willeke Traa³, Kevin Moerman⁴, Cees Oomens³, Aart Nederveen⁴, Klaas Nicolay¹, and Gustav Strijkers^1,2
1Biomedical NMR, Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, Netherlands, ²Biomedical Engineering and Physics, Academic Medical Center, Amsterdam, Netherlands, ³Soft Tissue Biomechanics and Engineering, Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, Netherlands, ⁴Department of Radiology, Academic Medical Center, Amsterdam, Netherlands

In order to study the role of tissue mechanical properties with MR Elastography (MRE) in the etiology of pressure ulcer related deep tissue injury a MRE actuator was added to a previously proposed MR compatible indentation setup for sustained tibialis anterior (TA) muscle loading in rats. We expect that the use of this novel device will provide new insights in the etiology of pressure ulcer related deep tissue injury.

Tomokazu Numano¹, Yoshihiko Kawabata², Kazuyuki Mizuhara³, Toshikatsu Washio⁴, Junichi Hata⁵, and Kazuhiro Homma^4
1Radiological Sciences, Tokyo Metropolitan University, Arakawa-ku, Tokyo, Japan, ²Takashima Seisakusho Co., Ltd., Tokyo, Japan, ³Tokyo Denki University, Tokyo, Japan,⁴National Institute of Advanced Industrial Science and Technology (AIST), Ibaraki, Japan, ⁵Graduate School of Medicine Keio University, Tokyo, Japan

In this work we developed a new retrofit MR Elastography (MRE) system consist of a gradient-echo type multiecho MR sequence and a wirelessly synchronized pneumatic vibration system. In the previous work (ISMRM2014, 1686), we introduced the wireless TR synchronization system with a dipole antenna tuned for the RF excitation frequency to make any electrical wiring from the MRI electronics unnecessary. However, the dipole antenna connected to the cable through a hole in the RF shield room, may introduce RF noise into the shield room and cause MR noises. In this study, we replaced the cable by an optical fiber, to prevent the external RF noise enter the magnet room.

Ria Mazumder^1,2, Renee Miller³, Haodan Jiang⁴, Bradley D. Clymer¹, Richard D. White^2,5, Alistair Young³, Anthony Romano⁶, and Arunark Kolipaka^2,5
1Department of Electrical and Computer Engineering, The Ohio State University, Columbus, OH, United States, ²Department of Radiology, The Ohio State University College of Medicine, Columbus, OH, United States, ³Department of Anatomy with Radiology, The University of Auckland, New Zealand, ⁴Department of Research and Development, Ohio Supercomputer Center, OH, United States, ⁵Department of Internal Medicine-Division of Cardiology, The Ohio State University College of Medicine, OH, United States, ⁶Naval Research Laboratory, DC, United States

Anisotropic mechanical properties of tissues vary in response to different pathological conditions; hence, the development of a technique for non-invasive anisotropic stiffness quantification is expected to have diagnostic and prognostic significance. Recently, waveguide magnetic resonance elastography (MRE) has been used to non-invasively estimate anisotropic stiffness of biological tissues by measuring tissue deformation as a result of external perturbation in directions specified by fiber orientation. In this study, we simulate fibers using finite element modeling in a cylindrical rod to validate the stiffness measurements obtained using waveguide MRE. Our results show that the technique can successfully estimate anisotropic stiffness in an orthotropic material.

Ondrej Holub¹, Simon Lambert¹, Katharina Schregel², Lynne Bilston³, Samuel Patz^4,5, and Ralph Sinkus^1
1Imaging Sciences and Biomedical Engineering, King's College London, London, London, United Kingdom, ²University Medicine Goettingen, Institute of Neuroradiology, Goettingen, Goettingen, Germany, ³University of New South Wales, Neuroscience Research Australia, Sydney, New South Wales, Australia, ⁴Brigham and Women's Hospital, Radiology, Boston, Massachusetts, United States, ⁵Harvard Medical School, Radiology, Boston, Massachusetts, United States

External finger tapping stimulus was successfully correlated to changes of the material properties in-situ. Work presents a novel imaging modality base on Magnetic Resonance Elastography (MRE) to provide additional information to fMRI assessment.

Mapping Magnetism using Magnetoencephalography

Wednesday 3 June 2015

Richard Bowtell¹, Mobeen Ali¹, Jason Medica¹, Ingrid Vella¹, and Mattthew Brookes^1
1School of Physics and Astronomy, University of Nottingham, Nottingham, United Kingdom

ahvcf'phaqwephf

Multi-Scale Motion

Wednesday 3 June 2015

Samuel Fielden¹, Li Zhao¹, Max Wintermark², and Craig Meyer^1,3
1Biomedical Engineering, University of Virginia, Charlottesville, Virginia, United States, ²Radiology, Stanford University, Palo Alto, California, United States, ³Radiology, University of Virginia, Charlottesville, Virginia, United States

The major disadvantage for ASL is low SNR and low spatial resolution of the resulting images. The hypothesis of this work is that the SNR and spatial resolution of perfusion images acquired with ASL can be improved by incorporating side information from high-SNR anatomical images into iterative reconstructions of the data. Here, we use the non-local means filter, trained on high-SNR anatomical images, to denoise and sharpen the ASL reconstruction results. We have tested this method in a simulated numerical phantom and with in-vivo data and found that it improves SNR and reduces error.

Thomas Lindner¹, Ulf Jensen-Kondering¹, Fritz Wodarg¹, Olav Jansen¹, and Michael Helle^2
1Department of Radiology and Neuroradiology, UKSH, Kiel, Germany, ²Philips Research, Hamburg, Germany

In this study, we present a method for dynamic artery-selective non-contrast enhanced magnetic resonance angiography. This method is based on the inflow effect of arterial blood. Two individual cylinder pulses are placed over two major brain feeding arteries, resulting in the image acquisition of the third remaining vessel. Over a time-course of 1450ms, six individual temporal phases with a frame-rate of 200ms can be acquired, while the total acquisition time is kept to five minutes. This method was successfully applied in healthy volunteers, visualizing even the distal parts of the individual flow territories of the intracranial arteries.

Swati Rane¹, Pratik Talati², Manus Donahue^3,4, and Stephan Heckers^2
1Radiology and Radiological Sciences, Vanderbilt University Institute of Imaging Science, Nashville, TN, United States, ²Psychiatry, Vanderbilt University, Nashville, TN, United States, ³Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, United States, ⁴Neurology, Vanderbilt University, Nashville, TN, United States

In this work we assessed the intra- and inter-scan reproducibility of iVASO at multiple inversion times (TI) of blood in the cortex and the hippocampus. iVASO reproducibility was higher at shorter TIs for the hippocampus compared to the cortex. Reproducibility was assessed using intra-class correlation analysis and found to be similar to that of arterial spin labeling approaches.

Yinghua Zhu¹, Yi Guo¹, Sajan Goud Lingala¹, R. Marc Lebel², Meng Law¹, and Krishna Nayak^1
1University of Southern California, Los Angeles, CA, United States, ²GE Healthcare, Calgary, Canada

The proposed GoLden Angle CartesIan Encoded Randomization (GLACIER) sampling scheme combines two existing techniques, Poisson ellipsoid pseudorandom undersampling and golden angle (GA) Cartesian sampling. GLACIER randomizes ky-kz phase encode along golden angle radials with designed sampling probability. Constrained reconstruction results of GLACIER are compared with two conventional methods in retrospective studies. Normalized root-mean-square error is used as an objective image quality metric. GLACIER algorithm is fast and allows online sampling pattern generation. GLACIER shows comparable results with Poisson ellipsoid and improved quality over conventional GA method in retrospective studies.

Jason Kraig Mendes¹, Scott McNally², and Dennis L Parker^1
1Radiology, University of Utah, Salt Lake City, Utah, United States, ²Clinical Radiology, University of Utah, Salt Lake City, Utah, United States

Dynamic contrast enhanced (DCE) imaging is useful in evaluating the functional status of a vascular system. Demands for high temporal and spatial resolution often lead to low signal to noise in the measurements. A novel sequence is presented that calculates concentration from bot T1 and T2 simultaneously to increase signal to noise.

Ulysse Gimenez¹, Antoine Triquet¹, and Hana Lahrech^1
1Clinatec, CEA, Grenoble, Rhones-Alpes, France

Super Resolution Reconstruction is performed on 2D EPI distorted diffusion images acquired in vivo on brain mouse at 9.4T. At this field, the feasibility to correct severe distortions with Reverse Gradient method is proved. Based on three orthogonal “low resolution” DTI acquisition (120*120*360 µm3), SRR is successfully performed allowing the calculation of 3D isotropic high spatial resolution DTI (120*120*120 µm3). Thus, high spatial resolution DTI parameter maps can be visualized. Fibertracking is also computed and fine white matter structures such as the fornix can be rebuilt.

Sen Ma¹, Xiaodong Ma², and Hua Guo^2
1Department of Electronic Engineering, Tsinghua University, Beijing, China, ²Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China

This paper proposes an effective joint reconstruction method to accelerate diffusion tensor imaging acquisition, combining the low-rank structure and sparsity constraints of the correlated diffusion-weighted images. We show that by jointly enforcing low-rank and sparsity constraints, we can achieve high reduction factor of diffusion tensor imaging acquisition while maintaining rather accurate reconstruction result.

Steven Baete^1,2, Tiejun Zhao³, and Fernando Emilio Boada^1,2
1Center for Advanced Imaging Innovation and Research (CAI2R), NYU School of Medicine, New York, NY, United States, ²Center for Biomedical Imaging, Dept. of Radiology, NYU School of Medicine, New York, NY, United States, ³Siemens Healthcare, Siemens Medical Solutions USA, Inc., New York, NY, United States

Diffusion Spectrum Imaging (DSI) has become a powerful tool for non-invasive imaging of white matter brain architecture. Unfortunately, widespread clinical implementation has been hampered by long acquisition times. DSI approaches sampling q-space in a radial fashion have been shown to be more efficient and accurate. When combined with a multi-echo train, RDSI provides significant throughput improvements over conventional DSI. We demonstrate a further acceleration of multi-echo RDSI by extending this technique with Simultaneous Multi-Slice excitation. The combined acceleration allows for whole brain fully sampled RDSI in under 7 min.

Eric Kenneth Gibbons¹, Shreyas Vasanawala², John Mark Pauly³, and Adam Bruce Kerr^3
1Department of Bioengineering, Stanford University, Stanford, California, United States, ²Department of Radiology, California, United States, ³Department of Electrical Engineering, Stanford University, California, United States

Here we present an modified non-CPMG FSE approach applied to abdominal DWI. We find that by modifying mixing crushers and use ESPIRiT/SENSE reconstruction we are able to achieve nearly artifact and distortion-free images that are often associated with body DWI images while also maintaining an appropriate echo train length.

Akira Yamamoto¹, Koji Fujimoto¹, Yasutaka Fushimi¹, Tomohisa Okada¹, Kei Sano², Toshiyuki Tanaka², and Kaori Togashi^1
1Department of Diagnostic Imaging and Nuclear Medicine, Graduate School of Medicine, Kyoto University, Kyoto, Kyoto, Japan, ²Department of Systems Science, Graduate School of Informatics, Kyoto University, Kyoto, Kyoto, Japan

Three different regularization methods, L1-norm, wavelet, and total variation in NESTA method for undersampled TOF-MRA image reconstruction were evaluated. In qualitative visual analysis, subtle but distinct difference was noted among them. In quantitative analysis, L1-norm showed the largest vessel-brain-ratio and more than 30 % undersampled data seemed sufficient for TOF-MRA reconstruction. Undersampled data less than 30 % showed visible image degradation. In conclusion, NESTA method can be used for TOF-MRA undersampled data reconstruction and L1-norm should be a choice for regularization method.

Koji Fujimoto¹, Takayuki Yamamoto¹, Thai Akasaka¹, Tomohisa Okada¹, Yasutaka Fushimi¹, Akira Yamamoto¹, Toshiyuki Tanaka², Kei Sano², Masayuki Ohzeki², and Kaori Togashi^1
1Diagnostic Imaging and Nuclear Medicine, Graduate School of Medicine, Kyoto University, Kyoto, Kyoto, Japan, ²Department of Systems Science, Graduate School of Informatics, Kyoto University, Kyoto, Kyoto, Japan

Reports on applying Compressed Sensing (CS) to TOF-MRA is still limited, probably because of difficulty due to a relatively lower SNR, and is hence challenging. In this work, we propose a simple and practical method to select a good regularization parameter applicable to TOF-MRA image reconstruction. We performed CS with 4x accelerated data at 3.0T by the FCSA algorithm with varying weights for Wavelet and Total Variation penalty. Among 6 different quantitative measures, the image selected by the highest SSIM value by using a masked MIP image was considered best by a clinical radiologist’s evaluation.

Thai Akasaka¹, Koji Fujimoto¹, Takayuki Yamamoto¹, Tomohisa Okada¹, Yasutaka Fushimi¹, Akira Yamamoto¹, Toshiyuki Tanaka², Kei Sano², Masayuki Ohzeki², and Kaori Togashi^1
1Diagnostic Imaging and Nuclear Medicine, Kyoto University Graduate School of Medicine, Kyoto, Kyoto, Japan, ²Kyoto University Graduate School of Informatics, Kyoto, Japan

The application of compressed sensing (CS) to Time-of-Flight MRA is usually 2D-based, although exploitation of 3D sparsity should perform better in principle. We performed CS with 4x accelerated data at 3.0T by the FCSA algorithm in both a 2D and 3D Wavelet and Total Variation reconstruction scheme. Among different results with varying weights for each priors, the best 2D and 3D-based images were subjectively selected. Images were evaluated by two radiologists, and visualization of clinically relevant structures were slightly superior for the best 3D-based image compared with best 2D-based image. Similarity measures were also in agreement with these results.

David Manuel Grodzki¹, Aurelien F Stalder¹, Yutaka Natsuaki², Julie Roesch³, and Bjoern Heismann^1,4
1Magnetic Resonance, Siemens Healthcare, Erlangen, Bavaria, Germany, ²Siemens Healthcare USA, Los Angeles, California, United States, ³Neuroradiology, University of Erlangen, Erlangen, Bavaria, Germany, ⁴Friedrich-Alexander-University of Erlangen-Nuremberg, Pattern Recognition Lab, Germany

During an MRI scan, the patient is exposed to high acoustic-noise levels caused by fast-switching gradients. By minimizing gradient activity, acoustic-noise can be significantly decreased. For flow-compensation gradients however, optimization of gradients also needs to consider the M1 moments of the gradient events. In this work, we present an automated gradient conversion algorithm that does not only smooth gradient shapes but also saves the M1 moment during flow compensation. Increased noise reduction is demonstrated for TOF and SWI scans, without sacrificing diagnostic image quality.

Joel Daouk¹, Roger Bouzerar^1,2, and Olivier Baledent^1,2
1BioFlow Image, University of Picardie Jules Verne, Amiens, Picardie, France, ²Medical Image Processing, CHU Amiens, Picardie, France

MR vascular images are obtained after the acquisition of several cardiac cycles followed by an average of these measures. In this case, we do not have the true cardiac information and respiratory influence is lost. The aim of the present study was to apply a near real time imaging to follow physiological changes in CSF, arterial and venous flows. Moreover, we expose the principle of an automatic Fourier-based method to extract respiratory and cardiac frequencies in fast MR images and to assess the contribution of the two physiological phenomena to the signal obtained during MR acquisitions in various fluids.

Yi-Hang Tung¹, Yun-An Huang², Edzer L. Wu², Wan-Ting Zhao², Tzi-Dar Chiueh², and Jyh-Horng Chen^2
1National Taiwan University, Taipei, Taiwan, Taiwan, ²National Taiwan University, Taiwan, Taiwan

Non-invasive self-gated IntraGate FLASH sequence (IG-FALSH) was used to perform cardiac imaging by reducing motion artifacts. However, IG sequence requires large repetitions to reduce the motion artifact and hence was hard to examine high spatial or temporal resolution cardiac imaging. In this study, we demonstrate the capability of implementing W=2 acceleration Single Carrier Wideband MRI on IG-FLASH sequence. Moreover, by trading time reduction for higher spatial resolution, the fine structures such as valve, papillary muscles and chordae tendineae could be observed more clearly with this new technique.

Motion Correction Head

Wednesday 3 June 2015

Thomas Siegert¹, Enrico Reimer¹, Roland Müller¹, Robert Turner¹, Harald Möller¹, and Jessica Schulz^1
1Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Sachsen, Germany

Time-of-flight (ToF) cameras have the potential to deliver real-time position information for prospective motion correction without special optical markers like retro-reflective spheres, checkerboard patterns or Moiré markers. In this abstract, we are reporting our initial experiments with a ToF camera within the bore of a 7T MR scanner. The modifications to make the camera MR-compatible as well as residual challenges are described and discussed.

Nicolai Spicher¹, Markus Kukuk¹, Mark E. Ladd^2,3, and Stefan Maderwald^2
1University of Applied Sciences and Arts Dortmund, Dortmund, Germany, ²Erwin L. Hahn Institute for Magnetic Resonance Imaging, University Duisburg-Essen, Essen, Germany, ³Division of Medical Physics in Radiology, German Cancer Research Center, Heidelberg, Germany

A video of the forehead is used to offer an alternative to contact-based triggering hardware. The video is obtained from a camera installed above the subjects head in a 7T MR bore; triggers are computed in real-time by an algorithm based on phase information of mean pixel intensities. During nonenhanced MR angiography of the lower extremities of one volunteer, triggers from the vendor-provided pulse oximeter and from the algorithm were saved. Results suggest that the video-based approach is principally able to reproduce the triggers obtained by pulse oximetry but suffers in case of the head motion.

Benjamin Knowles¹, Thomas Lange¹, Aditya Singh², Michael Herbst², and Maxim Zaitsev^1
1Medical Physics, Univeristy Medical Centre Freiburg, Freiburg, Germany, ²John A Burns School of Medicine, University of Hawaii, Honolulu, Hawaii, United States

For accurate prospective motion correction using markers, the marker must be rigidly coupled to the motion of the imaged object. However, patient comfort and ease of adhesion are important factors, especially in a clinical setting. In this study, dual-marker tracking is used to assess drifts and non-coupled motion of markers adhered to the forehead or nose. This is achieved by tracking the relative pose of the marker compared to a reference marker fixated to a mouthpiece, assumed perfectly coupled to the skull. Drifts were measured to be small, but squinting causes large decoupling of marker motion with the skull.

Uten Yarach¹, Daniel Stucht¹, Frank Godenschweger¹, and Oliver Speck^1
1Department of Biomedical Magnetic Resonance, Otto-von-Guericke University, Magdeburg, Sachsen-Anhalt, Germany

Patient motion during an MRI of the brain can result in non-diagnostic image quality. Even with perfect prospective motion tracking and correction, the varying coil sensitivity and gradient non-linearity can cause significant artifacts that cannot be corrected prospectively. Recently, a model-based MR image reconstruction via iterative solver was employed to minimize the sensitivity misalignment due to coil or physiological movement. In this study, we extended the mentioned model by gradient warp correction to reconstruct the prospective Mo-Co MR data. The result shows the improvement that is free from both artifacts after a few iterations of the proposed technique.

Joelle E Sarlls¹, Francois Lalonde², Dan Rettmann³, Ajit Shankaranarayanan⁴, Vinai Roopchansingh⁵, and S. Lalith Talagala^1
1NMRF/NINDS, National Institues of Health, Bethesda, MD, United States, ²NIMH, National Institutes of Health, Bethesda, MD, United States, ³GE Healthcare, Rochester, NY, United States, ⁴GE Healthcare, Menlo Park, CA, United States, ⁵FMIRF/NIMH, National Institutes of Health, Bethesda, MD, United States

Prospective motion correction techniques, such as PROMO, are increasingly available to improve the quality of high resolution anatomical MRI. PROMO consists of two parts, FOV updating and reacquisition of k-space data when excessive motion is detected. Reacquisition of data adds extra scan time, which is undesirable for pediatric or patient studies where scan time is limited. The contribution of each part of PROMO in reducing errors of cortical surface reconstruction measures was investigated. This study shows that utilizing either part of PROMO improves accuracy of measures, but full PROMO provides the greatest accuracy.

Burak Erem^1,2, Onur Afacan^1,2, Ali Gholipour^1,2, and Simon K Warfield^1,2
1Department of Radiology, Boston Children's Hospital, Boston, MA, United States, ²Harvard Medical School, Boston, MA, United States

Prospective motion correction of MRI steers the imaging field of view (FOV) to image as if the subject was not moving. All existing prospective motion correction techniques to date will perform the action of steering the FOV using motion measurements made in the past, because all motion measurements are available after some delay. Moreover, all of these existing techniques assume that this delay is infinitesimal. However, recent work demonstrates that this assumption is wrong and leads to poor motion compensation, especially when motions are rapid. Prediction of motion parameters shortly into the future has been proposed to reduce errors due to measurement delays and improve motion compensation. Here we assess how dynamic prediction compares to PROMO and static motion estimation.

Nils Noorman¹, Fredy Visser^1,2, Peter R. Luijten¹, and Jaco J.M. Zwanenburg^1
1Department of Radiology, University Medical Center Utrecht, Utrecht, Utrecht, Netherlands, ²Philips Healthcare, Best, Netherlands

Cardiac-driven volumetric strain pulsations in the brain are small, but relevant for normal physiology and disease. These strain rates can be obtained from phase contrast MRI with low encoding gradients, but the corresponding long TEs reduce the SNR. This work determines the optimal TE for phase contrast MRI with very low encoding velocities (< 1 cm/s) and the potential gain in SNR at 7T (with shorter T2*) compared to 1.5T, in 6 subjects. Despite signal attenuation due to shorter T2* at 7T, we obtained a 5-fold increase in SNR between 1.5T and 7T.

Daniel Papp¹, Martina F. Callaghan¹, Craig Buckley², Heiko Meyer³, and Nikolaus Weiskopf^1
1Wellcome Trust Centre For Neuroimaging, UCL Institute of Neurology, London, United Kingdom, ²SIEMENS PLC (Healthcare Division), United Kingdom, ³SIEMENS Healthcare AG, Germany

Quantitative imaging methods often rely on multiple scans and are therefore susceptible to inter-scan motion. In the presence of rapidly varying coil sensitivities, volunteer movement between scans will result in position-specific signal intensity modulation, which cannot be corrected for by rigid body motion correction and leads to error in the quantitative maps. Here we demonstrate the impact of this artefact on quantitative R1 maps and propose and validate a correction method based on dynamically updating coil-sensitivity maps. Our approach can be readily implemented and generalized to other multi-scan acquisitions.

Alexandru V. Avram¹, Joelle E. Sarlls², Cibu P. Thomas^1,3, Vinai Roopchansingh⁴, Dan Rettmann⁵, Ajit Shankaranarayanan⁶, and Peter J. Basser^1
1Section on Tissue Biophysics and Biomimetics, NICHD, National Institutes of Health, Bethesda, MD, United States, ²NINDS, National Institutes of Health, Bethesda, MD, United States, ³The Henry Jackson Foundation, Bethes, MD, United States, ⁴NIMH/Functional MRI Facility, National Institutes of Health, Bethesda, MD, United States, ⁵ASL, GE Healthcare, Rochester, MN, United States, ⁶ASL, GE Healthcare, Menlo Park, CA, United States

Images with different tissue contrasts acquired with the MP2RAGE pulse sequence, although inherently co-registered, can suffer from imaging artifacts due to subject motion especially in studies involving non-compliant subjects such as pediatric or elderly patient populations. We implement a PROMO-enabled 3D MP2RAGE pulse sequence, evaluate its ability to correct for subject motion, and illustrate how it can be used for multi-contrast high-resolution tissue visualization and characterization in moving subject at 3T.

Alexander Loktyushin¹, Maryna Babayeva^2,3, Daniel Gallichan⁴, Gunnar Krueger^2,3, Klaus Scheffler^5,6, and Tobias Kober^2,3
1Empirical Inference, Max Planck Institute for Intelligent Systems, Tübingen, Germany, ²Siemens ACIT - CHUV Radiology, Siemens Healthcare IM BM PI, & Department of Radiology, University Hospital (CHUV), Lausanne, Switzerland, ³LTS5, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, ⁴CIBM, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, ⁵High-Field Magnetic Resonance Center, Max Planck Institute for Biological Cybernetics, Tübingen, Germany, ⁶Department for Biomedical Magnetic Resonance, University of Tübingen, Tübingen, Germany

The present study combines retrospective motion correction and GRAPPA reconstruction. We propose a technique that performs several alternations of GRAPPA interpolation and motion correction steps, suppressing the artifacts caused by motion over the course of the optimization. Motion parameters are estimated directly from the data with the aid of free induction decay navigators. The proposed algorithm does not require a priori knowledge of coil sensitivity profiles and can be applied retrospectively to data acquired with generic sequences such as MP-RAGE. The algorithm was tested on motion corrupted brain images of healthy volunteers, performing controlled head movement during the scan. Results demonstrate a significant improvement in image quality.

Lucilio Cordero-Grande^1,2, Emer Hughes^1,2, Rui Pedro A. G. Teixeira^1,2, and Joseph V. Hajnal^1,2
1Centre for the Developing Brain, King's College London, London, London, United Kingdom, ²Division of Imaging Sciences and Biomedical Engineering, King's College London, London, London, United Kingdom

A method is presented for retrospective rigid motion correction in multi-shot brain neonatal MR acquisitions. It is based on using parallel imaging so that the joint motion correction and image reconstruction problem becomes overconstrained. Additionally, no blurring is introduced in the presence of motion as no regridding is applied. The procedure has proven to significantly enhance the quality of neonatal brain images by decreasing the presence of confounding artifacts, improving the contrast, better resolving structures, and enabling the emergence of new anatomical details hindered by motion.

Uten Yarach¹, Daniel Stucht¹, Frank Godenschweger¹, and Oliver Speck^1
1Department of Biomedical Magnetic Resonance, Otto-von-Guericke University, Magdeburg, Sachsen-Anhalt, Germany

Unlike the retrospective technique, prospective motion correction (Mo-Co) can avoid spin history effects as well as k-space inconsistencies. However, in case of large motion during multi-coil acquisition, the coil sensitivities that change relative to the object are one of the most considerable residual artifact causes in prospective Mo-Co. In this study, we investigate the effect of coil sensitivity variations on parallel prospective Mo-Co data if only initial coil sensitivities are known. We also show that if a small number of central k-space can be acquired when motion occurred, they are sufficient for estimating coil sensitivities to correct abovementioned artifact.

Paul Wighton^1,2, Matthew Dylan Tisdall^1,2, Erez Nevo³, and André Dylan van der Kouwe^1
1Martinos Center for Biomedical Imaging, MGH, Charlestown, MA, United States, ²Harvard Medical School, Boston, MA, United States, ³Robin Medical, Baltimore, MD, United States

Several external tracking prospective motion correction systems require a marker or sensor be affixed to the subject's head. These systems assume that the sensor moves rigidly with respect to the subject's head. When this assumption is violated, performance of the system degrades. This abstract investigates the incorporation of position estimates from imaging data to recover from non-rigid sensor affixation.

Mathias Engström^1,2, Enrico Avventi^1,2, Magnus Mårtensson^2,3, Ola Norbeck¹, and Stefan Skare^1,2
1Dept. of Neuroradiology, Karolinska University Hospital, Stockholm, Stockholm, Sweden, ²Dept. of Clinical Neuroscience, Karolinska Institutet, Stockholm, Stockholm, Sweden, ³EMEA Research and Collaboration, GE Applied Science Laboratory, GE Healthcare, Stockholm, Sweden

Here we suggest combining collapsed fat navigation (cFatNav) with diffusion-weighted 3D multi-slab echo-planar imaging (DW 3D-MS EPI) for prospective motion correction. 3D rigid body motion was registered via three orthogonal EPI readouts, which depicted projections of the subcutaneous fat. With a readout block duration of 6.4 ms (R = 8) motion updates were received at a ~5-10 Hz update rate, at a 5% scan time increase.

Mahta Karimpoor^1,2, Zahra Faraji-Dana^1,2, and Simon James Graham^1,2
1Physical Scienses, Sunnybrook Research Institute, Toronto, Ontario, Canada, ²Medical Biophysics, University of Toronto, Toronto, Ontario, Canada

An fMRI-compatible computerized tablet system has been previously developed in house to be used for complex Neuropsychological testing such as writing. The tablet did not include visual feedback of hand position (VFHP), a human factors component that may be important for fMRI of certain patient populations. Because the user is lying in the magnet bore, they cannot view their hand during this process and must rely on their sense of proprioception. A real-time system was thus developed to provide VFHP and integrated with the tablet in an augmented reality display. Attempting to visualize the hand position during performing such complex motor movements may induce excess subject’s head motion. We assessed the influence of VFHP on head motion and the associated brain activity involving a set of handwriting tasks in young healthy adults. ​

Chung-Ki Wong¹, Vadim Zotev¹, Han Yuan¹, Masaya Misaki¹, Raquel Phillips¹, Qingfei Luo¹, and Jerzy Bodurka^1,2
1Laureate Institute for Brain Research, Tulsa, Oklahoma, United States, ²College of Engineering, University of Oklahoma, Norman, Oklahoma, United States

Head motion during fMRI impairs data quality. EEG-assisted retrospective motion correction (E-REMCOR), which utilizes EEG data to correct for head movements in fMRI on a slice-by-slice basis, was shown to be capable of substantially removing movement in fMRI datasets. To enhance E-REMCOR usability, especially for the rapidly growing interest in concurrent EEG and fMRI measurements, we developed automatic E-REMCOR (aE-REMCOR) for head motion correction. aE-REMCOR was applied to 305 fMRI scans at 3 Tesla. The average change of TSNR over the brain goes up to 24%. The largest 10 percent of TSNR improvement reaches over 43%.

Patricia Johnson^1,2, Junmin Liu¹, Trevor Wade¹, and Maria Drangova^1,2
1Imaging Research Laboratories, Robarts Research Institute, London, Ontario, Canada, ²Dept. of Medical Biophysics, Schulich School of Medicine & Dentistry, Western University, London, Ontario, Canada

Spherical navigator echo (SNAV) techniques are promising for measuring rigid body motion in all 6 degrees of freedom. In this abstract we demonstrate – for the first time - the accuracy of retrospective preRot-SNAV-correction of motion by interleaving the SNAVs within a 3D imaging pulse sequence (efgre-SNAV). We have performed a preliminary evaluation of this sequence and determined that the interleaved SNAVs perform comparably to stand alone SNAVs in measurements of rotation and translations; they agree within 1.0° and 0.43mm. We have also demonstrated successful retrospective correction of a phantom image.

Yilong Liu^1,2, Xiaodong Ma², Hua Guo², and Ed X. Wu^1,3
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, Hong Kong, China, ²Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, Beijing, China, ³Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, Hong Kong, China

Variable density spiral (VDS) sampling is advantageous for fast imaging, it also has potential in motion correction due to repeated sampling of each interleaf around the k-space center. However, for VDS, the full sampling area of each interleaf may be too small for accurate motion estimation. In this work, the navigation region is defined to cover more than the full sampling area, and we use SPIRiT to further improve navigation region. Results show that our proposed method can significantly remove the artifacts in both simulation and in vivo studies.

Motion Correction - Body

Wednesday 3 June 2015

Aaron T Hess¹, Andre JW van der Kouwe^2,3, Matthew Dylan Tisdall², Stefan Neubauer¹, and Matthew D Robson^1
1Oxford Centre for Clinical Magnetic Resonance Research (OCMR), Oxford, Ox, United Kingdom, ²Radiology, Harvard Medical School, Boston, MA, United States, ³Martinos Center, Massachusetts General Hospital, Boston, MA, United States

A novel 2D rapid gradient echo navigator is proposed for real time diaphragm navigation. At ultrahigh fields (>3T) traditional diaphragm navigators are challenging due to either large B0 gradients, limited B1+ or SAR demands. The proposed gradient echo navigator requires minimum B1+ and takes ~70 ms to acquire. The new 2D navigator is compared to a product 1D cross paired navigator at 3T, and is demonstrated to be robust and reliable at 7T. The contrast to noise is of the new 2D navigator is similar to that of the cross paired navigator, with an improved diaphragm edge sharpness.

Xiaodong Chen^1,2, Zuoquan Zhang³, Qihua Yang¹, Zebin Luo², Ziliang Cheng¹, Jiaji Mao¹, Queenie Chan⁴, Hua Guo⁵, and Biling Liang^1
1Sun Yat-Sen Memorial Hospital, Guangzhou, Guangdong, China, ²Affiliated hospital of Guangdong Medical College, Zhanjiang, Guangdong, China, ³The Fifth Affiliated Hospital of Sun Yat-Sen University, Guangdong, China, ⁴Philips Healthcare, Hong Kong, China, ⁵Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China

With the development of MRI techniques, T2* mapping is becoming the routine examination for evaluation of the iron level of the heart and liver in TM patients. Generally, the common T2* sequence is scanned with one breath hold and less attention is paid to motion artifact reduction because most adults can suspend respiration on command. However, as the recommended age to start the first cardiac MRI screening in TM patients has become younger, more and more children with TM will get T2* examination and most of them are generally less able to suspend respiration. In this study, We optimized a technique of T2* respiratory gating, developed and tested the use of respiratory gating to decrease motion artifacts in pediatric heart and liver T2* MRI examinations. From the preliminary study it suggests that, comparing to the results of breath-hold T2* scanning, the accuracy and reproducibility of respiratory gating method are good enough to be used in the clinical practice.

Francisco Contijoch¹, Yuchi Han¹, Michael Hansen², Peter Kellman², Gene Gualtieri³, Mark A Elliott¹, Sebastian Berisha¹, James J Pilla¹, Robert C Gorman¹, and Walter RT Witschey^1
1University of Pennsylvania, Philadelphia, PA, United States, ²National Institute of Health, Bethesda, Maryland, United States, ³Drexel University, Philadelphia, PA, United States

Inconsistent motion during a cine MRI can result in image artifacts, loss of signal-to-noise ratio, or poor representation of the underlying physiology because the scan does not always respond correctly or quickly, to changes in physiology. We proposed an adaptive real-time k-space sampling trajectory (ARKS) to respond to a physiologic feedback signal to reduce motion effects and ensure sampling uniformity. The most recent signals from an ECG waveform of normal subjects and patients were continuously matched to the previous signal history, new radial k-space locations were determined, and these MR signals were combined using multi-shot or single-shot radial acquisition schemes.

Nicholas Dwork¹, Daniel O'Connor², Nii Okai Addy¹, Reeve Ingle¹, John Pauly¹, and Dwight Nishimura^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Mathematics, University of California, Los Angeles, CA, United States

3D image based navigators (iNAVs) provide an opportunity to measure whole volume non-rigid beat-to-beat motion of the heart. In this work, we show that we can use optical flow to estimate motion from 3D iNAVs. The Alternating Direction Method of Multipliers is used to minimize the optical flow cost function, where total variation regularization has been imposed.

Jason Brown¹, Cihat Eldeniz¹, Wolfgang Rehwald², Brian Dale³, Hongyu An¹, and David Lalush^1
1Joint Department of Biomedical Engineering, The University of North Carolina at Chapel Hill and North Carolina State University, Chapel Hill, NC, United States, ²Siemens Healthcare, Malvern, PA, United States, ³Siemens Healthcare, Cary, NC, United States

In this application, we effectively estimated patient-specific 3D deformable motion from fast 2D MRI images. CLARET is an image registration method that has been used to relate a set of 2D images to a corresponding set of 3D images. Using CLARET to predict the 3D motion of a subject from a set of 2D projection images has the potential to be used in MRI imaging of dynamic processes. The results of the registration give a motion estimate that reduced alignment error mean and variance in 2D frames. We concluded that CLARET can be used effectively in an MRI setting.

Yuchou Chang¹, Dallas C Turley¹, and James G Pipe^1
1Imaging Research, Barrow Neurological Institute, Phoenix, Arizona, United States

A PROPELLER based 3D Fast Field Echo (FFE) sequence and reconstruction method for free-breathing data acquisition in dynamic abdominal imaging is studied. The proposed method acquires 3D blades dynamically depending on the physiology property. Two weighting functions are proposed to preferentially use the preferred data in the gridding process and reconstruction. Experimental results demonstrate that the proposed method is able to reduce respiration artifacts and sharpen images.

Isabel Dregely¹, Fei Han¹, Ziwu Zhou¹, Kyung Sung¹, Peng Hu¹, and Holden H Wu^1
1Radiological Sciences, University of California Los Angeles, Los Angeles, CA, United States

T2-weighted MRI is the central component for body cancer MRI protocols. In clinical routine T2-w MRI is often restricted to 2D multi-slice acquisitions, because 3D acquisition times are relatively long and achieving robustly good image quality is often hindered by motion corruption. The purpose of our work was to develop a 3D T2-w motion robust imaging sequence using a magnetization-prepared radial imaging technique with self-navigating motion correction. We found that bulk and respiratory motion could be reliably detected using the self-navigating radial sequence. The golden angle view ordering allowed for flexible retrospective image reconstruction, allowing removing motion-corrupted views. ESPIRiT reconstruction yielded good image quality even in under-sampled data sets.

Tao Zhang^1,2, Joseph Y Cheng^1,2, Yuxin Chen², John M Pauly², and Shreyas S Vasanawala^1
1Radiology, Stanford University, Stanford, CA, United States, ²Electrical Engineering, Stanford University, Stanford, CA, United States

Abdominal MRI is often limited by respiratory motion. Some acquisition trajectories, such as modified Cartesian and radial, can provide self-navigating signals during data acquisition. This enables retrospective motion compensation (MC). With phased array coils, navigators can be acquired for each coil element. Since navigators from different coil elements track motion in local regions that can vary significantly, averaging navigators from all coils sometimes yields inaccurate respiration estimation, and therefore ineffective MC. Here, a robust navigator processing method using coil clustering is described, implemented, and validated. The proposed method automatically selects the appropriate navigators for MC from coil arrays.

Christine Nabuurs¹, Gabriele Beck¹, Silke Hey¹, and Marko Ivancevic^2
1Clinical Excellence, Philips Healthcare, Best, NB, Netherlands, ²Clinical Science, Philips Healthcare, Best, NB, Netherlands

Inconsistent breathholds during 3D dynamic DCE MR imaging of the abdomen are a major cause of motion artifacts. In this study we compared motion sensitivity of dynamic 3D-DIXON FFE to end-breath hold motion between interleaved1 and grouped2 peripheral ky-kz viewsharing. The acquisition of the complete peripheral kykz at the end of the breath hold makes gouped viewsharing less sensitive to breathing induced motion at the end of the acquisition. It also allows for a higher temporal resolution in the early dynamics, which is especially beneficial for capturing early arterial phases of abdominal MR imaging.

Paul Kinchesh¹, Philip D Allen¹, John S Beech¹, Emmanouil Fokas¹, Stuart Gilchrist¹, Veerle Kersemans¹, Ruth Muschel¹, and Sean C Smart^1
1Department of Oncology, University of Oxford, Oxford, United Kingdom

Constant T1 weighting is difficult to achieve in 2D multi-slice MRI when used with respiratory gating or triggering. Long term averaging and retrospective gating have been used to ameliorate this, but at the expense of significantly extended scan time. This report describes a new technique that enables respiratory-gated 2D multi-slice MRI to operate at a constant, short TR (TR<3T1). The method is maximally efficient as data are acquired throughout the entire respiratory interval, and are discarded only when motions are present. The method is compatible with the addition of cardiac triggering for TR>5T1 and is demonstrated in the mouse abdomen.

Holger Eggers¹, Rafael Wiemker¹, Peter Börnert¹, and Tim Leiner^2
1Philips Research, Hamburg, Germany, ²Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands

Flow compensation is usually not applied in first-pass peripheral angiography due to stringent scan time constraints. This entails the occurrence of flow artifacts, which are potentially aggravated by the recently proposed use of Dixon- instead of subtraction-based background suppression in this application. Therefore, the effect of flow artifacts inside the vessels on the water-fat separation is studied in the present work and illustrated by examples. Moreover, an approach is proposed that relies on a vessel tree segmentation to provide a comprehensive suppression of flow artifacts in subtractionless first-pass peripheral angiography.

Patrick McDaniel¹, Borjan Gagoski², M. Dylan Tisdall^3,4, André J. W. van der Kouwe^3,4, P. Ellen Grant^2,4, Lawrence Wald^3,4, and Elfar Adalsteinsson^1,5
1Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, ²Fetal-Neonatal Neuroimaging and Developmental Science Center, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States, ³Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States,⁴Radiology, Massachusetts General Hopsital, Boston, MA, United States, ⁵Health Sciences and Technology, Harvard-MIT, Cambridge, MA, United States

The ability to perform higher quality fetal brain imaging would be facilitated by navigators that track fetal head motion. We demonstrate the feasibility of measuring this motion along six degrees of freedom using 3D volumetric navigator (vNav) acquisitions. A series of vNavs is acquired over time, and a fetal head ROI is identified within the first 3D volume to serve as a template for rigid-body registration of subsequent 3D images. Registrations are performed with FSL FLIRT, producing affine transformation matrices which provide quantitative measurements of fetal head displacement and rotation.

Jin Liu¹, Huijun Chen², Zechen Zhou², Jinnan Wang³, and Chun Yuan^1
1University of Washington, Seattle, WA, United States, ²Tsinghua University, Beijing, China, ³Philips Research North America, NY, United States

Motion artifact is a major challenge affecting the image quality of MRI. Navigator based motion correction usually need extra scan time with sequence dependence, while existing optical motion tracking systems need markers for accurate motion measurements. Therefore, we proposed a non-marker-attached optical motion tracking system based on structured light to measure the absolute distance, which has the merits of high accuracy, sequence independence and no patient interaction. This study demonstrated that the motion can be accurately identified with the proposed method and then corrected retrospectively by deleting and resynthesizing the corresponding corrupted k-space lines.

Thomas Küstner^1,2, Christian Würslin¹, Sergios Gatidis¹, Petros Martirosian³, Nina Schwenzer¹, Bin Yang², and Holger Schmidt^1
1Department of Radiology, University Hospital of Tübingen, Tübingen, Baden-Württemberg, Germany, ²Institute of Signal Processing and System Theory, University of Stuttgart, Stuttgart, Baden-Württemberg, Germany, ³Diagnostic and Interventional Radiology, University Hospital of Tübingen, Tübingen, Baden-Württemberg, Germany

In the field of oncology, simultaneous PET/MR scanners offer a great potential for improving diagnostic accuracy. However for accurate lesion detection and quantification, the induced motion artifacts in the PET image originating from a long free-breathing acquisition have to be compensated. The simultaneous acquisition allows performing a MR-based non-rigid motion correction of the PET data. Therefore it is essential to acquire a 4D MR image as accurate and fast as possible under free-breathing conditions. We propose a clinical feasible motion correction setup which uses a Compressed Sensing acquisition. In-vivo patient data substantiates the diagnostic improvement of the motion corrected PET.

Christopher M Rank¹, Thorsten Heußer¹, Marcus Brehm¹, and Marc Kachelrieß^1
1Division of Medical Physics in Radiology, German Cancer Research Center, Heidelberg, Germany

We propose a new method for PET/MR respiratory Motion Compensation (MoCo), which is based on strongly undersampled MR data. In our simulation study, we used a 3D encoded radial stack-of-stars sampling scheme with 160 radial spokes per partition and an acquisition time of 38 s. Based on iteratively reconstructed 4D MR images, high-fidelity motion vector fields were estimated and MoCo 4D PET images of a simulated breathing thorax were reconstructed. Evaluation of 8 artificial hot lesions in the lungs and upper abdomen showed a significant visual as well as a quantitative improvement in terms of SUV_mean values of MoCo 4D PET images compared to 3D and 4D gated reconstructions.