MR Fingerprinting & Quantitative Imaging

Tuesday 2 June 2015

Debra McGivney¹, Anagha Deshmane², Yun Jiang², Dan Ma², and Mark Griswold^1,2
1Radiology, Case Western Reserve University, Cleveland, Ohio, United States, ²Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States

Magnetic resonance fingerprinting (MRF) is a technique that can provide quantitative maps of tissue parameters such as T1 and T2 relaxation times through matching observed signals to a precomputed complex-valued dictionary of modeled signal evolutions. Since each dictionary entry is uniquely defined by two real parameters, specifically T1 and T2, we propose to compress the dictionary onto a real-valued manifold of three dimensions using the nonlinear dimensionality reduction technique of kernel principal component analysis. Once the compression is achieved, we explore new computational applications for MRF, namely solving the partial volume problem.

Debra McGivney¹, Anagha Deshmane², Yun Jiang², Dan Ma², and Mark Griswold^1,2
1Radiology, Case Western Reserve University, Cleveland, Ohio, United States, ²Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States

Magnetic Resonance Fingerprinting (MRF) can produce quantitative maps of tissue parameters such as T1 and T2 relaxation times by matching acquired signals to a predefined dictionary of signal evolutions. One inherent issue is that all voxels are assigned only one dictionary entry, even if they exhibit the partial volume effect. We apply a Bayesian statistical framework to solve the general partial volume problem for MRF without assigning in advance the specific dictionary entries that comprise a signal from one of these mixed voxels, rather, assumptions are made on the probability distributions of the mixed signals and their component signals.

Philippe Pouliot^1,2, Louis Gagnon³, Tina Lam⁴, Pramod Avti⁵, Michèle Desjardins¹, Ashok Kakkar⁴, Sava Sakadzic³, David Boas³, and Frédéric Lesage^1
1Electrical Engineering, Ecole Polytechnique Montreal, Montreal, QC, Canada, ²Research Centre, Montreal Heart Institute, Montreal, QC, Canada, ³Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, MA, United States, ⁴Chemistry Department, McGill University, QC, Canada, ⁵Montreal Heart Institute, QC, Canada

MR vascular fingerprinting is a novel approach to estimate cerebral blood volume, vessel radius and oxygenation. To our knowledge, this approach has not yet been fully validated. Here we implemented the sequence in mice and exploited a dictionary built on simulations of the MR signal based on realistic vasculature built on 2-photon angiograms. A dictionary for fingerprint extraction was generated by sampling along 5 parameters: hemoglobin saturation, vessel radius, capillary density, SPION concentration and magnetic field inhomogeneity. Following linearization, the dictionary eigensystem was characterized. This confirmed that all its eigenvalues are positive and distinct, and therefore all parameters studied are theoretically identifiable.

Cristoffer Cordes¹ and Matthias Günther^1,2
1Fraunhofer MEVIS, Bremen, Germany, ²MR-Imaging and Spectroscopy, University of Bremen, Bremen, Germany

In order to extract the information density of images acquired with a given protocol, data was parameter mapped (T1, T2, M0) using an objective function based on a simulated signal model, minimized with a variation of the simulated annealing algorithm. Calculating the uncertainty volumes based on an uncertainty condition of the objective function reveals a contrast that is able to rank the performance of the utilized sequences by eliminating the sequence of least preferable impact in a greedy fashion. It also reveals the voxel-wise shape of the remaining flaws. The algorithm was tested on a series of TSE acquisitions.

Taejoon Eo¹, Jinseong Jang², Minoh Kim², Dong-hyun Kim², and Dosik Hwang^2
1Yonsei University, Seoul, Seoul, Korea, ²Yonsei University, Seoul, Korea

Proposed tier-specific WEST method could sufficiently suppress the noise-like artifacts in the maps obtained by the conventional WEST. Consequently, this method enables acquisition of accurate maps from extremely undersampled Cartesian MRF data.

Mathieu Sarracanie^1,2, Ouri Cohen¹, and Matthew S Rosen^1,2
1MGH/A.A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ²Department of Physics, Harvard University, Cambridge, MA, United States

2D MR Fingerprinting has recently been shown at low magnetic field. Here, we demonstrate MRF in 3D at 6.5 mT, using an optimized set of 15 flip angles and repetition times (FA/TR), in a Cartesian acquisition of k-space with a new hybrid b-SSFP-EPI sequence. We measure quantitative parameters in 3D, and generate several image contrasts in a single acquisition (proton density, T1, T2) in less than 30 minutes. The combination of 3D MRF with low field MRI scanners has great potential to provide clinically relevant contrast with portable low cost MR scanners.

Jesse Ian Hamilton¹, Katherine L Wright¹, Yun Jiang¹, Luis Hernandez-Garcia², Dan Ma¹, Mark Griswold^1,3, and Nicole Seiberlich^1,3
1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, ²Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, ³Radiology, Case Western Reserve University, Cleveland, OH, United States

A flexible framework for MR Fingerprinting pulse sequence design is presented that includes the MRI signal encoding, gridding, and pattern recognition directly in the optimization. The method was validated in a phantom study by designing sequence for mapping T1, T2, and M0 in under 3s using a highly undersampled spiral trajectory. Parameter maps obtained with the optimized sequence have fewer artifacts and higher agreement with spin echo measurements compared to unoptimized sequences. The optimization framework is easily generalizable to other MRF applications.

Christian Anderson¹, Ying Gao¹, Chris Flask^1,2, and Lan Lu^2,3
1Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States, ²Radiology, Case Western Reserve University, Cleveland, Ohio, United States, ³Urology, Case Western Reserve University, Cleveland, Ohio, United States

Magnetic resonance fingerprinting (MRF) offers rapid simultaneous multi-parametric quantification, and also provides the potential to generate maps of other parameters. We have developed a novel scheme named "Multi-Preparation MRF" (MP-MRF) that implements adaptable magnetization preparations periodically during the dynamic MRF acquisition. Our initial simulations of the MP-MRF methodology show sensitivity to diffusion and perfusion contrast and reasonable estimates of T1, T2, and velocity in Shepp-Logan phantoms.

Te-Ming Lin¹, Su-Chin Chiu¹, Cheng-Chieh Cheng¹, Wen-Chau Wu^1,2, and Hsiao-Wen Chung^1
1Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan, ²Graduate Institute of Oncology, National Taiwan University, Taipei, Taiwan

The signal acquisition number is related to the computational complexity during signal analysis in MR fingerprinting. In this study, we develop a contour area index and demonstrate a quantitative method to evaluate the mapping precision under different signal acquisition numbers. It has potential in evaluating different RF excitation schemes in MR fingerprinting.

Nicolas Pannetier^1,2 and Norbert Schuff^1,2
1Radiology, UCSF, San Francisco, California, United States, ²VAMC, San Francisco, CA, United States

We evaluate the use of kd-tree (a space partitioning data structure) to speed-up the matching process in magnetic resonance fingerprinting. We found that, in combination with PCA reduction, the matching time can be reduced by 2 to 3 order of magnitude while preserving the accuracy. The matching time, however, increases with noise level and the PCA threshold remains a key element to tune to achieve the best performance.

Dan Ma¹, Eric Y Pierre¹, Yun Jiang¹, Kawin Setsompop², Vikas Gulani³, and Mark A Griswold^3
1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, ²A.A Martinos Center for Biomedical Engineering, MGH, Harvard Medical School, Boston, MA, United States, ³Radiology, Case Western Reserve University, Cleveland, OH, United States

The purpose of this study is to extend the 2D MR Fingerprinting (MRF) and MRF-Music framework to 3D acquisitions. Both methods were originally implemented in 2D acquisitions and have shown high scan efficiency for quantifying multiple tissue properties simultaneously. In addition to the multi-parameter quantification in MRF, the MRF-Music sequence was proposed to provide musical sounds that can dramatically improve the patients’ experience in the MR scanner. In this study, the MRF and MRF-Music sequences were implemented to achieve 3D coverage while still maintaining a high scan efficiency and providing desirable sounds. T1 and T2 values from phantom studies of the 3D slab selective MRF and MRF-Music methods showed good agreement to the values from the standard measurements. The T1, T2, off-resonance and M0 maps from 3D non-selective MRF and MRF-Music also showed promising results of achieving 3D isotropic quantitative mapping.

Florian Knoll^1,2, Martijn A Cloos^1,2, Thomas Koesters^1,2, Michael Zenge³, Ricardo Otazo^1,2, and Daniel K Sodickson^1,2
1Center for Advanced Imaging Innovation and Research (CAI2R), NYU School of Medicine, New York, NY, United States, ²Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, NY, United States, ³Siemens Medical Solutions USA, Malvern, PA, United States

Despite the extensive opportunities offered by PET-MR systems, their use is still far from routine clinical practice. While it is feasible to acquire PET data in about 5 minutes, collecting the clinically relevant variety of traditional MR contrasts requires substantially more time. This bottleneck formed by the traditional MR paradigm leads to inefficient use of the PET component. This work proposes a one-step procedure that merges the MR fingerprinting framework with the PET acquisition, and employs a dedicated multi-modality reconstruction to enable a 6 minute comprehensive PET-MR exam, which can provide the majority of clinical MR contrasts alongside quantitative parametric maps of the relaxation parameters (T1,T2) together with improved PET images.

Bernhard Strasser¹, Wolfgang Bogner¹, Peter Bär¹, Gilbert Hangel¹, Elisabeth Springer¹, Vlado Mlynarik¹, Mark A Griswold^2,3, Dan Ma², Yun Jiang², Mathias Nittka⁴, Haris Saybasili⁴, and Siegfried Trattnig^1
1MRCE, Department of Biomedical Imaging and Image-guided Therapy, University of Vienna, Vienna, Vienna, Austria, ²Department of Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States, ³Radiology, Case Western Reserve University, Cleveland, Ohio, United States, ⁴Siemens Healthcare USA, Inc., Chicago, Illinois, United States

Previously, MR Fingerprinting (MRF) has been presented as a new method for simultaneous quantitative mapping of different physical MR properties. In this study, the T1 and T2 values of MRF were compared to conventional T1- and T2-mapping methods in the brains of five volunteers at 1.5T. Each volunteer was measured five times with a TrueFISP and a FISP based spiral MRF sequence, an MP2RAGE and a multi echo spin echo sequence for conventional T1 and T2 maps, respectively. Both MRF sequences showed a similar reproducibility but seemed to slightly underestimate the T2-values in comparison to the conventional sequences.

Zhaohuan Zhang^1,2, Zhe Wang^2,3, Subashini Srinivasan^2,3, Kyunghyun Sung^2,3, and Daniel B. Ennis^2,3
1Department of Physics & Astronomy, Shanghai Jiao Tong University, Shanghai, China, ²Department of Radiological Sciences, University of California, Los Angles, CA, United States, ³Department of Bioengineering, University of California, Los Angles, CA, United States

The objective of this study was to evaluate the accuracy and precision of pseudorandom inversion recovery balanced steady-state free precession magnetic resonance fingerprinting (MRF) relaxometry (T1 and T2) estimates over a range of SNRs and the number of acquired TRs (NTR) using Bloch equation simulations. Under the condition of perfect sampling, the Bloch simulations defined a lower-bound acquisition requirement of SNR¡Ý5 and NTR¡Ý400 for accurate and precise T1 and T2 estimates when using MRF. This work also concluded that MRF provides nearly equivalent T1 and T2 estimates.

Thomas Amthor¹, Mariya Doneva¹, Peter Koken¹, Jochen Keupp¹, and Peter Börnert^1
1Philips Research Europe, Hamburg, Germany

We present a comparison of different pattern matching algorithms for tissue characterization based on Magnetic Resonance Fingerprinting. The applicability of a simple dot product approach and a number of machine learning algorithms is investigated for different parameter regimes. We find that, in many cases, machine learning algorithms can offer higher accuracy and faster matching.

Mariya Doneva¹, Thomas Amthor¹, Peter Koken¹, Jochen Keupp¹, and Peter Börnert^1
1Philips Research Europe, Hamburg, Germany

In this work we demonstrate a comprehensive accuracy analysis exemplified on a bSSFP-based MRF sequence, which allows predicting the accuracy of MRF in different parameter ranges and defining confidence areas for the performance of MRF.

Enhao Gong¹, Qiyuan Tian¹, John M Pauly¹, and Jennifer A McNab^2
1Electrical Engineering, STANFORD UNIVERSITY, Stanford, California, United States, ²Radiology, STANFORD UNIVERSITY, Stanford, California, United States

Low Signal-to-Noise Ratio (SNR), especially at high b-values, is a critical problem for Diffusion MRI (dMRI). Methods with different signal models may fail to reconstruct under-sampled data from noisy measurement. Diffusion MRI signal contains redundancy as a multi-dimensional signal in both k-space and q-space. Here we proposed a novel approach to recover signal without explicitly enforcing any physical signal model. The method is model-free but learns the multi-dimensional redundancy, including the redundancy between neighborhood voxels, different directions and low\high b-values, from training samples. A Dictionary Learning approach is used to recover under-sampled signals in q-space. Quantitative results demonstrate the method can more accurately predict high b-value signal (>3000s/mm2) from low b-value signal. Also it produces more accurate physiological metrics such as Generalized Fractional Anisotropy (GFA) and Orientation Distribution Function (ODF) that potentially help to resolve intra-voxel crossing fibers.

Andrea Vargas¹, Laurent Milot², Simon Graham¹, and Philip Beatty^1
1Medical Biophysics, University of Toronto, Toronto, Ontario, Canada, ²Sunnybrook Research Institute, Toronto, Canada

The use of variable flip angles for 3D fast spin echo sequences (3DFSE) have shown to alter contrast in T2-weighted images relative to conventional 2DFSE. While these alterations of contrast may be minimal in brain tissues, they can have a great consequences in body applications that encompass a wide range of T2 values. In this study we evaluate the performance of current methods that aim to correct T2-contrast in a cervix cancer application which has a wide range of T2 values (35 ms < T2 < 84 ms). We show that the differentiation between recurrent tumor and radiation fibrosis may be ambiguous at clinical echo times using 3DFSE.

Lili He¹, Jinghua Wang², Mark Smith³, and Nehal A. Parikh^1,4
1Center for Perinatal Research, The Research Institute at Nationwide Children's Hospital, Columbus, Ohio, United States, ²Center for Cognitive and Behavioral Brain Imaging, The Ohio State University, Columbus, Ohio, United States, ³Radiology Department, Nationwide Children's Hospital, Columbus, Ohio, United States, ⁴Department of Pediatrics, The Ohio State University College of Medicine, Columbus, Ohio, United States

Three-dimensional T1-weighted sequences such as MP-RAGE are extremely valuable to evaluate neonatal and infant brain injury/development. Yet, the lack of complete myelination and smaller head size results in comparatively lower quality images as compared to adult brains. In this study, we consider WM-GM contrast efficiency as an objective function to optimize neonatal MP-RAGE parameters under optimal k-space sampling by means of computer simulation. Quantitative analysis indicated that WM-GM contrast to noise efficiency of images acquired with our optimal parameters was 20% higher than those using parameters recommended by a published protocol; similarly, mean SNR efficiency was increased by approximately 150%.

Jonathan I. Tamir¹, Weitian Chen², Peng Lai², Martin Uecker¹, Shreyas S. Vasanawala³, 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, ³Radiology, Stanford University, Stanford, CA, United States

Fast Spin Echo (FSE) is widely used in MR imaging due to its speed and robustness to image artifacts. However, blurring due to T2 decay inhibits its use for 3D musculoskeletal imaging. By compensating for signal decay and reconstructing a time series of images, the blurring can be reduced. In this work we resample and reorder phase encodes over a longer echo train length to improve scan efficiency. We add a locally low rank constraint to improve the conditioning of the reconstruction, producing multicontrast 3D FSE images at clinically feasible scan times.

Yi-Cheng Hsu¹, Ying-Hua Chu¹, Shang-Yueh Tsai², Wen-Jui Kuo³, and Fa-Hsuan Lin^1
1Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan, ²Institute of Applied Physic, National Chengchi University, Taipei, Taiwan,³Institute of Neuroscience, National Yang Ming University, Taipei, Taiwan

A MP trueFISP sequence for brain structural imaging was implemented and tested. Compared with MP RAGE using the same acquisition time, it improves the contrast from 40% to 80% with 37.8% noise increase due to a wider readout bandwidth.

Bing Wu¹, Nan Hong², and Zhenyu Zhou^1
1GE healthcare China, Beijing, Beijing, China, ²Peking university people's hospital, Beijing, China

T1 rho acquisition is often constrained to single slice due to the long TSL needed, which makes the cross-examination with other measurements such as resting state fMRI difficult. In this work, we develop a rapid T1-rho mapping method that utilizes single-shot EPI acquisition and multi-band excitation that completes a 2mm isotropic whole brain T1 rho mapping within 5 minutes, which allows this acquisition to be added in a Parkinson disease related clinical study.

Won-Joon Do¹, Seung Hong Choi², Eung Yeop Kim³, and Sung-Hong Park^1
1Korea Advanced Institute of Science and Technology, Daejeon, Korea, ²Department of Radiology, Seoul National University College of Medicine, Seoul, Korea,³Department of Radiology, Gachon University Gil Medical Center, Incheon, Korea

Dual-echo sequence allows us to acquire 3D T1 and T2*-weighted images simultaneously. The conflicting parameter conditions of T1and T2* contrasts can be resolved by echo-specific k-space reordering schemes. However, abrupt changes in scan conditions for the echo-specific k-space reordering can cause ringing artifacts. In this study, we propose a new approach of smooth transition in the regions of abrupt changes, to suppress the artifacts. The ringing artifacts in the echo-specific k-space reordered dual-echo sequence without the smooth transition could be effectively suppressed with the proposed approach and thus the image qualities became closer to those acquired with conventional single-echo sequences.

Galen D Reed¹, Reeve Ingle¹, Ken O Johnson¹, Juan M Santos¹, Bob S Hu², and William R Overall^1
1Heartvista, Menlo Park, California, United States, ²Cardiology, Palo Alto Medical Foundation, Menlo Park, California, United States

High resolution inversion recovery imaging of myocardium within small breath hold durations is challenging due to the need for segmented acquisitions and short readout windows. By combining the efficiency of parallel spiral imaging with a 2-dimensional field-of-view reduction, we designed a sequence that acquires 1.7 mm in-plane resolution images in a 7 heartbeat breath hold. The short acquisition window enabled repeating the sequence to obtain a series of images with different inversion times. The efficacy of multiple TI imaging with and without 2D outer volume suppression was demonstrated.

Reconstruction & Processing Algorithms

Tuesday 2 June 2015

Paniz Adipour¹ and Michael R. Smith^1,2
1Electrical and Computer Engineering, University of Calgary, Calgary, Alberta, Canada, ²Radiology, University of Calgary, Calgary, Alberta, Canada

Truncation artifacts appear in DFT reconstructions through discontinuities across the ends of the data set which mathematically is cyclic in k-space. A suggestion indicates that similar position dependent distortions will be present in CS reconstructions which repeatedly use the DFT. A comparison is made between standard Wavelet and Contourlet CS reconstructions and proposed high k-space extrapolation enabled (Hi-KEE) variants of these approaches. The CS-Contourlet outperforms the common CS-Wavelet in providing a better sparse representation of contour-shaped objects and detailed textures. The Hi-KEE-CS-Contourlet is shown to outperform the CS-Contourlet by providing a better position independent resolution solution.

Edward Li¹, Mohammad Javad Shafiee¹, Audrey Chung¹, Farzad Khalvati², Alexander Wong¹, and Masoom A Haider^3
1Systems Design Engineering, University of Waterloo, Waterloo, Ontario, Canada, ²Department of Medical Imaging, University of Toronto, Toronto, Ontario, Canada, ³Sunnybrook Health Sciences Center, Toronto, Ontario, Canada

Compressive sensing reduces MRI acquisition times but requires advanced sparse reconstruction algorithm to produce high-quality MR images. We propose a novel sparse reconstruction method using a cross-domain stochastically fully-connected random field (CD-SFCRF) for improved reconstruction from compressive sensing MRI data. Peak-to-peak signal-to-noise ratio (PSNR) analysis of CD-SFCRF and other methods using a prostate training phantom demonstrate that CD-SFCRF has the highest PSNR across all under-sampling ratios of radial MRI acquisitions. A visual comparison using real patient cases illustrate that CD-SFCRF can improve fine tissue detail and contrast preservation while eliminating under-sampling artifacts.

Michael R. Smith^1,2, Jordan Woehr¹, Mathew E. MacDonald^2,3, and Paniz Adipour^1
1Electrical and Computer Engineering, University of Calgary, Calgary, Alberta, Canada, ²Radiology, University of Calgary, Calgary, Alberta, Canada, ³Seaman MR Family Research Centre, University of Calgary, Calgary, Alberta, Canada

Truncated k-space data sets provide higher temporal resolution but compromise spatial resolution during DFT reconstruction. Compressed sensing, using under-sampled data, is used to improve spatial resolution while retaining temporal resolution. Certain Fourier domain properties can produce MRI CS reconstruction with resolutions that are dependent on the position of an object in the final reconstructed image. We demonstrate this position dependent resolution and propose two approaches to overcome it: Fourier Shift (FS) and Area Specific Additional Truncation (ASAT) image resolution enhancement pre-processing techniques.

Satoshi Ito¹, Mone Shibuya¹, Kenji Ito¹, and Yoshifumi Yamada^1
1Utsunomiya University, Utsunomiya, Tochigi, Japan

It is difficults to apply CS to images with rapid spatial phase variations, since not only the magnitude but also phase regularization is required in the CS framework. An iterative MRI reconstruction with separate magnitude and phase regularization was proposed for applications where magnitude and phase maps are both of interest. Since this method requires the approximation of phase regularizer to cope with phase unwrapping problem, it is roughly 10 times slower than conventional CS and the convergence is not guaranteed. In this article we propose a novel image reconstruction scheme for CS-MRI in which phase regularizer or symmetrical sampling trajectory are not required in the rather standard CS reconstruction scheme, but highly robust to rapid phase changes. The proposed method uses multi-frame complex transforms to introduce sparseness for the complex image data.

Matthias Weigel^1
1Radiological Physics, Dept. of Radiology and Nuclear Medicine, University of Basel Hospital, Basel, Switzerland

The extended phase graph (EPG) concept is a favorite approach for the rapid quantitation of magnetization response. However, users frequently have problems to properly program the framework. One major reason may be that care has to be taken with the complex Fourier domains of the transverse magnetization and their inherent symmetry relations. The present educational abstract depicts these issues and shows how RF pulses and gradients act differently on the magnetization components. Solutions to overcome the described issues are presented and discussed. Additionally, the author provides representative EPG software demonstrating the solutions.

Pavan Poojar¹, Bikkemane Jayadev Nutandev¹, Amaresha Sridhar Konar¹, Rashmi R Rao¹, Ramesh Venkatesan², and Sairam Geethanath^1
1Medical Imaging Research Centre, Dayananda Sagar Institutions, Bangalore, Karnataka, India, ²Wipro-GE Healthcare, Bangalore, Karnataka, India

Cardiac MRI scans demands rapid acquisition of images to avoid motion artifacts. Region of interest (ROI) selected will be sparse and leads to arbitrary k-space shape. Active contour in combination with convex optimization leads to new ROI based acquisition strategy which gives arbitrary k-space trajectories and optimized gradients based on the constraints for given ROI. Retrospective studies were carried out on six cardiac datasets for different accelerations (3x, 4x, 5x and 10x) and Normalized Root Mean Square Error was calculated. Future work includes reconstruction of image using ROI Compressed Sensing.

Weiyi Chen¹, Yi Guo¹, Ziyue Wu², and Krishna S. Nayak^1,2
1Electrical Engineering, University of Southern California, Los Angeles, CA, United States, ²Biomedical Engineering, University of Southern California, Los Angeles, CA, United States

We propose a method for MRI constrained reconstruction using ADMM framework that is data-driven, and does not require manual selection of tuning parameters. We use Morozov's discrepancy principle as a criterion to iteratively determine the tuning parameter. Tests with T2w brain data show that the reconstruction quality is comparable with reconstructions using manually selected parameter.

Xiaobo Qu¹, Yunsong Liu¹, Jing Ye¹, Di Guo², Zhifang Zhan¹, and Zhong Chen^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²School of Computer and Information Engineering, Xiamen University of Technology, Xiamen, Fujian, China

Compressed sensing magnetic resonance imaging (CS-MRI) is to reconstruct MR images from undersampled k-space data by enforcing the sparsity of MR images. Patch-based nonlocal operator (PANO) is proposed as a linear operator to exploit the nonlocal self-similarity of MR images to further sparsify them. However, the original PANO is a frame and its numerical algorithm for CS-MRI problem is solved by the alternating direction minimization with continuation (ADMC). These two aspects lead the reconstruction to be time consuming. In this work, we first convert the PANO into a tight frame, and then applied the alternating direction method of multipliers (ADMM) algorithm to accelerate the image reconstruction. The empirical convergence demostrates that the new approach significantly accelerate the image reconstruction in compressed sensing MRI and can accomplish the reconstruction of one 256256 within several seconds.

Nishant Zachariah¹, Johannes M Flake², Qiu Wang³, Boris Mailhe³, Justin Romberg¹, Xiaoping Hu⁴, and Mariappan Nadar^3
1Department of Electrical and Computer Engineering, Georgia Institute of Technoloy, Atlanta, GA, United States, ²Department of Mathematics, Rutgers University, New Brunswick, NJ, United States, ³Imaging and Computer Vision, Siemens Corporate Technology, Princeton, NJ, United States, ⁴Department of Biomedical Engineering, Emory University and Georgia Institute of Technology, Atlanta, GA, United States

Increasing MR image resolution, decreasing MR instrumentation noise and reconstructing high quality MR images from under sampled measurements are open challenges. In this paper we tackle these three problems under a novel non convex framework. We show that our method out performs state of the art techniques (quantitatively and qualitatively) for image super-resolution, denoising and under sampled reconstruction. In addition, we are able to recover regions of clinical interest with greatest fidelity thereby substantially aiding the clinical diagnostic process. Our powerful generic framework lends itself to tackling additional future applications such as image in-painting and blind de-convolution.

Matthew J. Muckley¹, Douglas C. Noll¹, and Jeffrey A. Fessler^2
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, ²Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States

MRI scan times can be accelerated by combining parallel MRI with sparse models. These models give rise to optimization problems that are traditionally minimized with variable splitting algorithms that require tuning of penalty parameters. We review a new algorithm, BARISTA, that circumvents penalty parameter tuning while preserving convergence speed. We then propose a new optimized momentum update term for BARISTA that gives a theoretically-predicted factor of 2 increase in convergence speed of the cost function, terming the new algorithm OMBARISTA. Our optimization experiments agreed with the theory predictions, and we propose using OMBARISTA in place of BARISTA in general settings.

Rizwan Ahmad¹, Philip Schniter¹, and Orlando P. Simonetti^2
1Electrical and Computer Engineering, The Ohio State University, Columbus, Ohio, United States, ²Internal Medicine and Radiology, The Ohio State University, Columbus, Ohio, United States

Redundant dictionaries are routinely used to exploit rich structure in MR images. When using a redundant dictionary, however, the level of sparsity may vary across different groups of atoms, i.e., across “subdictionaries.” In this work, we propose a method, called Sparsity Adaptive Compressive Recovery (SCoRe), that adapts to the inherent level of sparsity in each subdictionary. Moreover, the proposed adaptation is data-driven and does not introduce any tuning parameters. For validation, results from digital phantom and real-time cine are presented.

Zongying Lai^1,2, Yunsong Liu¹, Di Guo³, Jing Ye¹, Zhifang Zhan¹, Zhong Chen¹, and Xiaobo Qu^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²Department of Communication Engineering, Xiamen University, Fujian, China,³School of Computer and Information Engineering, Xiamen University of Technology, Xiamen, Fujian, China

Compressed sensing MRI can speed up imaging by undersampling k-space data. However, the sparse representation of magnetic resonance images affects the quality of reconstructed images. In this work, a graph-based compressed sensing MRI image reconstruction method is proposed. This method views an image patch as a vertex on graph and reorders the pixel to be smooth by traveling this graph with shortest path. Image reconstruciong from compressively sampled data shows that the proposed reconstruction method outperforms conventional wavelets in terms of visual quality and evaluation criteria.

Zechen Zhou¹, Niranjan Balu², Rui Li¹, Jinnan Wang^2,3, and Chun Yuan^1,2
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, ²Vascular Imaging Lab, Department of Radiology, University of Washington, Seattle, WA, United States, ³Philips Research North America, Briarcliff Manor, NY, United States

Parallel Imaging (PI) and Compressed Sensing (CS) enable accelerated MR imaging. However, the actual PI-CS reconstruction performance is usually limited by noise amplification and image boundary/structure blurring particularly at high reduction factor. In this work, a Gaussian Mixture Model (GMM) was optimized to promote structured sparsity and it was further merged into the SPIRiT framework as a regularization constraint. The proposed algorithm has demonstrated its improved performance for image boundary and detail structure preservation in accelerated 3D high resolution brain imaging.

Gabriel Ramos-Llordén¹, Hilde Segers¹, Willem Jan Palenstijn¹, Arnold J. den Dekker^1,2, and Jan Sijbers^1
1iMinds Vision-Lab, University of Antwerp, Antwerp, Antwerp, Belgium, ²Delft Center for Systems and Control, Delft University of Technology, Delft, Netherlands

In MRI reconstruction, undersampled data sets lead to ill-posed reconstruction problems. To regularize these problems, prior knowledge is commonly exploited. In this work, we introduce a new type of prior knowledge, partial discreteness, where part of the image is assumed to be homogeneous and can be well represented by a constant magnitude. We introduce this prior in the common algebraic reconstruction problem and propose an iterative algorithm to approximately solve it. It combines a penalized least squares reconstruction with an internal Bayesian segmentation. Results with synthetic data demonstrate that more detailedly restored images are obtained when partial discreteness is exploited

Andres Saucedo^1,2, Stamatios Lefkimmiatis³, Stanley Osher³, and Kyunghyun Sung^1,2
1Department of Radiological Sciences, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, California, United States, ²Biomedical Physics Interdepartmental Graduate Program, University of California Los Angeles, Los Angeles, California, United States, ³Department of Mathematics, University of California Los Angeles, Los Angeles, California, United States

This study introduces a novel constrained reconstruction technique that exploits both the local correlation of image data across multiple coils and the inherent non-local self-similarity property of images. Our approach is based within a non-local total variation regularization framework. The proposed method is applicable to both compressed sensing and parallel imaging, and demonstrates substantial advantages with regard to high levels of noise.

Ryan Wen Liu¹, Lin Shi², Simon C.H. Yu¹, and Defeng Wang^1,3
1Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong, ²Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong, ³Department of Biomedical Engineering and Shun Hing Institute of Advanced Engineering, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong

In this study, we propose to combine L₀ regularized tree-structured wavelet sparsity (TsWS) and second-order total generalized variation (TGV²) to reconstruct MR image from highly undersampled k-space data. In particular, the L₀ regularized TsWS could better represent the measure of sparseness in wavelet domain. TGV² is capable of maintaining trade-offs between artefact suppression and tissue feature preservation. To achieve solution stability, the corresponding minimization problem is decomposed into several simpler subproblems. Each of these subproblems has a closed-form solution or can be efficiently solved using existing optimization algorithms. Experimental results have demonstrated the superior performance of our proposed method.

Johannes F. M. Schmidt¹ and Sebastian Kozerke^1,2
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland, ²Division of Imaging Sciences and Biomedical Engineering, King's College London, United Kingdom

A statistical approach to combine multiple denoising algorithms in MR image reconstruction to suppress reconstruction artifacts.

Xiaohui Liu¹, Jinhong Huang¹, Wufan Chen¹, and Yanqiu Feng^1
1Guangdong Provincial Key Laborary of Medical Image Processing, School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China

Undersampled MRI reconstruction techniques based on Compressed Sensing (CS) exploiting sparsity which is implicit in MR images can provide significant help in reducing the scan time during clinical period, but remains challenging due to the requirement of high reconstruction accuracy. A novel algorithm is developed and tested in vivo for solving the MR image reconstruction problem due to Nesterov¡¯s smoothing scheme and convex conic optimization.

MingJian Hong¹, MengRan Lin¹, Feng Liu², and YongXin Ge^1
1ChongQing University, ChongQing, ChongQing, China, ²ITEE, The University of Queensland, QLD, Australia

This work presented a new model by enforcing both local and global sparsity, which captures both the patch-level and global sparse structures of the anatomical images. Using a model split approach, the image reconstruction quality can be iteratively further improved. Our simulation results demonstrate that, the proposed method outperform those existing methods using only the patch-level or global sparse structure.

Yunsong Liu¹, Jian-Feng Cai², Zhifang Zhan¹, 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) has shown to be promising to accelerate magnetic resonance imaging (MRI). There are two different sparse models in CS-MRI: analysis and synthesis models with different assumptions and performance when a redundant tight frame is used. A new balance model is introduced into CS-MRI that can achieve the solutions of the analysis model, synthesis model and some in between by tuning the balancing parameter. It is found in this work that the typical balance model has a comparable performance with the analysis model in CS-MRI. Both of them achieve lower reconstructed errors than the synthesis model no matter what value the balancing parameter is. These observations are consistent for different tight frames used CS-MRI.

Florian Knoll^1,2, Martin Holler³, Thomas Koesters^1,2, and Daniel K Sodickson^1,2
1Center for Advanced Imaging Innovation and Research (CAI2R), NYU School of Medicine, New York, NY, United States, ²Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, New York, United States, ³Department of Mathematics and Scientific Computing, University of Graz, Graz, Austria

It was recently shown that simultaneously acquired data from state-of-the-art MR-PET systems can be reconstructed simultaneously using the concept of joint sparsity, yielding benefits for both MR and PET reconstructions. In this work we propose a new dedicated regularization functional for multi-modality imaging that exploits common structures of the MR and PET images. The two modalities are treated as single multi-channel images and an extension of the second order Total Generalized Variation functional for vector valued data is used as a dedicated multi-modality sparsifying transform.

Chenguang Peng¹, Rong Guo¹, Yicheng Chen¹, Yingmao Chen², Quanzheng Li³, Georges El Fakhr³, and Kui Ying^1
1Key Laboratory of Particle and Radiation Imaging, Ministry of Education, Department of Engineering, Beijing, China, ²Department of Nuclear Medicine, The general hospital of Chinese People's Liberation, Beijing, China, Beijing, China, ³Department of Radiology, Division of Nuclear Medicine and Molecular Imaging, Harvard Medical School, Boston, United States

PET is a practical medical imaging technique for brain function diagnosis. However, the low spatial resolution limits the use of PET in neurology and disease like Alzheimer's disease. With the help of MRI-PET, people can use high resolution MRI to provide anatomical information to correct partial volume effect of PET image which is a great cause for low resolution. Nevertheless, traditional partial volume effect correction method requires an accurate MRI segmentation and PVE model estimation which are not usually applicable. In this work, we proposed a method that is insensitive to PVE model estimation error and segmentation error.

Mathias Lukas¹, Anne Kluge², Jorge Cabello¹, Christine Preibisch^2,3, and Stephan Nekolla^1
1Department of Nuclear Medicine, Klinikum rechts der Isar, TU München, Munich, Germany, ²Department of Neuroradiology, Klinikum rechts der Isar, TU München, Munich, Germany, ³Department of Neurology, Klinikum rechts der Isar, TU München, Munich, Germany

Attenuation correction (AC) in quantitative PET/MR is affected by SNR and CNR of underlying MR sequences. In this work, the quality of MR data currently used for attenuation correction in PET (UTE, DIXON, MPRAGE) was observed in-vivo under changing clinical conditions over 3 months to investigate the reliability and robustness for in-house established MR based AC methods. In spite of its semi quantitative character, all sequences were found to be very invariant in SNR and CNR and can be used without any concerns.

Lishui Cheng¹, Sangtae Ahn¹, Dattesh Shanbhag², Florian Wiesinger³, Sandeep Kaushik², and Ravindra Manjeshwar^1
1GE Global Research, Niskayuna, NY, United States, ²GE Global Research, Bangalore, India, ³GE Global Research, Munich, Germany

Attenuation correction is critical to accurate PET quantitation. In PET/MR, MR-based attenuation correction (MR-AC) has challenges in bone, air, lung and implant regions. To address the problem, we combined 1) a segmentation-based MR-AC method, which works well in soft-tissue regions, and 2) a selective-update joint estimation approach, which reconstructs both attenuation and activity from PET emission data, to resolve the attenuation in the challenging regions. The method was evaluated on clinical data from a PET/MR scanner with TOF information and it was demonstrated that the method can distinguish between abdominal air and spinal implant/bone regions, otherwise hidden in MR.

Non-Cartesian, Multiband & Parallel Imaging

Tuesday 2 June 2015

Yi-Cheng Hsu¹, Ying-Hua Chu¹, and Fa-Hsuan Lin^1
1Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan

We propose an iterative k-x method to estimate weights to reconstruct missing radial sampling k-space data points using individually reconstructed coil images from the under-sampled data directly. Once missing k-space data were estimated, individually reconstructed coil images used in the last estimation were replaced by coil images in this reconstruction for the next iteration. Our method can successfully reconstruct human brain images with 2 mm spatial resolution and minimal streaking artifacts using 22 radial projections at 3T using a 32-channel head coil array.

Stephen F Cauley^1,2, Kawin Setsompop^1,2, Lawrence Wald^1,2, and Jonathan R Polimeni^1,2
1Athinoula A. Martinos Center for Biomedical Imaging, MGH/HST, Charlestown, MA, United States, ²Dept. of Radiology, Harvard Medical School, Boston, MA, United States

Regularization of parallel imaging (PI) reconstruction has a significant impact on signal-to-noise and image artifact levels. Attempts have been made to automatically determine the correct balance between stability and data consistency. We introduce effective rank as a proxy to be used for automated PI regularization. Unlike condition number, effective rank correlates with the number of dominate basis vectors that are contributing to the reconstruction. Line search algorithms can quickly sweep regularization levels to determine the appropriate parameter. We demonstrate the benefits of our approach for GRAPPA reconstruction with two classes of regularization using typical array coils and acceleration factors.

Ronald Mooiweer¹, Alessandro Sbrizzi¹, Alexander Raaijmakers¹, Cornelis A. T. van den Berg¹, Peter R. Luijten¹, and Hans Hoogduin^1
1UMC Utrecht, Utrecht, Utrecht, Netherlands

Including the rFOX in the SENSE reconstruction results in a substantial reduction of the g-factor penalty in SNR and enables highly accelerated scans. Images of good quality were obtained at a 25-fold acceleration, using a 32-channel receive coil. This is shown by calculations and measurements using 2D SSE in vivoat 7T.

Jinsuh Kim¹, Casey P Johnson², Dingxin Wang³, and Philip Zhe Sun^4
1University of Iowa, Iowa City, IA, United States, ²University of Iowa, IA, United States, ³Siemens Medical Solutions USA, Inc., Minneapolis, MN, United States, ⁴Martinos Center for Biomedical Imaging, MGH, Charlestown, MA, United States

CEST imaging generally requires long scanning time which significantly hampers clinical translation. In this work, we introduce an accelerated CEST imaging using a compressed sensing image reconstruction of radial acquisition trajectory with a golden-angle ordering scheme combined with parallel imaging technique. We tested this method in a creatine phantom and in vivo muscle before and after exercise. This work provides a proof-of-concept of new method for future clinical application.

Cihat Eldeniz¹, Wolfgang Rehwald², Brian Dale³, Yasheng Chen¹, and Hongyu An^1
1University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, ²Siemens Healthcare, Malvern, PA, United States, ³Siemens Healthcare, Cary, NC, United States

One key objective for MR/PET motion correction is to obtain 3D MRI images at multiple respiratory phases in order to derive 3D motion fields. In this study, we developed a reconstruction method that can yield multiple 3D respiratory phase images during free breathing. This method takes advantage of the redundant information provided by multiple coils and the neighboring respiratory phases to fill out the missing partitions. Since the reconstruction is performed in the k-vs-phase space, the proposed method is named as kp-GRAPPA.

Kyong Hwan Jin¹, Dongwook Lee¹, and Jong Chul Ye^1
1Dept. of Bio and Brain Engineering, KAIST, Daejeon, Daejeon, Korea

In this paper, we propose a novel parallel imaging method called PRESTO (pyramidal representation of block Hankel structure low rank matrix) that do not require any calibration data but still outperform all the existing parallel imaging methods such as GRAPPA, SAKE (irregularly sampled k-space without calibration region), etc. In multi coil k-space, we reveal that the set of k-space data from several multi coils have novel annihilation properties between different coils as well as within coils. These annihilation properties lead us to a block Hankel structured matrix whose rank should be low dimensional. Accordingly, similar to SAKE, the parallel imaging problem becomes a low rank matrix completion of missing k-space data. However, unlike the SAKE, which exploits the low rankness from all k-space data or needs to combine E-SPIRiT to reduce the complexity, we demonstrate that the low rankness needs to be exploited in a pyramidal representation of block Hankel structured matrix to improve image quality as well as to reduce the complexity.

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

Calibrationless constrained reconstruction methods employing low-rank models have attracted recent attention due to their high accuracy and sampling flexibility. Recently, k-space-based methods LORAKS and P-LORAKS were proposed for calibrationless reconstruction of images with slowly varying phase from single-channel and multi-channel parallel imaging data. For the same settings, we propose an image-domain locally low rank model to exploit slow phase variation. The model can be used to augment other image-domain constrained reconstruction models to exploit slow phase variation with little overhead.

Yu Y. Li^1
1Radiology, Imaging Research Center, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, United States

This work introduces a mathematical model that converts k-space parallel imaging into the function of a low-rank Toeplitz-like correlation matrix formed from auto- and cross-channel correlation functions. By applying a k-space variant low-rank constraint to this correlation matrix, missing data can be reconstructed in a region-by-region fashion. Imaging acceleration can be improved if a higher undersampling factor is used in those regions with a more stringent constraint. It is demonstrated that this approach permits the use of a net acceleration factor higher than the number of coil elements in the phase-encoding direction.

Andrew J Coristine^1,2, Jérôme Yerly^1,2, and Matthias Stuber^1,2
1Department of Radiology, University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, VD, Switzerland, ²CardioVascular Magnetic Resonance (CVMR) research centre, Centre for Biomedical Imaging (CIBM), Lausanne, VD, Switzerland

Two dimensional (2D) spatially selective radiofrequency (RF) pulses may be used to constrain the location from which an MR signal is obtained. T₂-Preparation, or T₂-Prep, is a magnetization preparation scheme used to improve blood/myocardium contrast. By incorporating a "pencil beam" 2D pulse into a T₂-Prep module, one may create a "2D-T2-Prep" that combines T2-weighting with the intrinsic spatial selectivity of a 2D pulse. This may be of particular benefit to parallel imaging techniques such as GRAPPA, where artefacts can originate from residual foldover signal. As the 2D-T₂-Prep suppresses signal from outside the area of interest, parallel imaging artefacts may likewise be reduced. In this abstract, we present numerical simulations, phantom validation, and in vivo MRA of the right coronary artery, demonstrating that GRAPPA accelerated images may be dramatically improved through the use of a 2D-T₂-Prep.

Nils Nothnagel¹, Rodrigo Fernandez-Jiménez², Gonzalo Lopez-Martin², Manuel Desco³, Valentin Fuster², Borja Ibañez², and Javier Sánchez-González^1
1Philips Healthcare Spain, Madrid, Spain, ²Atherothrombosis in Experimental Imaging, Centro Nacional de Investigaciones Cardiovasculares (CNIC), Madrid, Spain, ³Departamento de Bioingeniería e Ingeniería Aerospacial, Universidad Carlos III, Madrid, Spain

In this work, we developed a new image reconstruction algorithm that enable the possibility to acquire isotropic 3D cardiac cine imaging in a single breath-hold. This technique use the data redundancy to reach acceleration factors up to 13.5 allowing 3D cine acquisition in a single breath-hold. In addition the reconstruction strategy allows for total reconstruction time around 3 mins for a whole 3D data set (2.0x2.0x2.0mm3 isotropic resolution, 16 cardiac phases). In the abstract the acquisition and reconstruction strategy are presented and in-vivo results are shown in pig model.

Ali Ersoz¹ and L Tugan Muftuler^2,3
1Department of Biophysics, Medical College of Wisconsin, Milwaukee, Wisconsin, United States, ²Department of Neurosurgery, Medical College of Wisconsin, Milwaukee, Wisconsin, United States, ³Center for Imaging Research, Medical College of Wisconsin, Milwaukee, Wisconsin, United States

Conventional radial imaging requires interpolation onto Cartesian grid, which might degrade image quality. Pseudo-Polar Fourier Transform (PPFT) is a direct, exact and fast transformation between the k-space data in pseudo-polar (PP) grid and image in Cartesian grid. In this study, we incorporated GRAPPA into PPFT reconstruction and compared it with the conventional radial GRAPPA using simulations, 3T phantom experiments and 7T human experiments. Both simulation and experimental results demonstrated that PP trajectory provides images with significantly reduced reconstruction errors and sharper edge resolution compared to conventional radial trajectory even at high acceleration rates.

El-Sayed H. Ibrahim^1,2, Robert Pooley², Mellena Bridges², Joseph Cernigliaro², James Williams³, and William Haley^2
1University of Michigan, Ann Arbor, MI, United States, ²Mayo Clinic, Jacksonville, FL, United States, ³Indiana Unicersity, IN, United States

CT is established as the method of choice for imaging kidney stones, especially with dual-energy CT (DECT) that can identify uric-acid (UA) from non-UA stones. With the advent of ultra-short echo-time (UTE) MRI, adequate imaging of kidney stones becomes possible. The purpose of this work is to compare MRI versus DECT imaging of 114 kidney stones, representing different stone types and sizes, in phantom experiments using different surrounding materials and scan-setups. The results showed that MRI is capable of imaging kidney stones of different types and sizes. However, no significant differences were observed in relaxation times from different stone types.

Melisa Okanovic¹, Robert Trampel², Martin Blaimer¹, Felix Breuer¹, and Peter Michael Jakob^1,3
1MRB Research Center for Magnetic-Resonance-Bavaria, Würzburg, Bavaria, Germany, ²Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Saxony, Germany, ³Experimental Physics 5, University of Würzburg, Würzburg, Bavaria, Germany

The radial GRASE hybrid sequence is presented for high-resolution Neuro-imaging at 7T. In this sequence, the number of refocusing pulses is reduced by additional readout gradients while keeping the number of acquired echoes per excitation constant. The number of refocusing pulses and thus SAR is reduced by the additional readout gradients compared to conventional TSE sequences. Furthermore, the radial readout allows for the reconstruction of different T2-weighted images from one measurement.

Thomas Benkert¹, Philipp Ehses^2,3, Martin Blaimer¹, Peter Jakob^1,4, and Felix Breuer^1
1Research Center Magnetic Resonance Bavaria, Würzburg, Bavaria, Germany, ²Department for Neuroimaging, University of Tübingen, Tübingen, Baden-Württemberg, Germany, ³High-Field MR Center, Max Planck Institute for Biological Cybernetics, Tübingen, Baden-Württemberg, Germany, ⁴Experimental Physics 5, University of Würzburg, Bavaria, Germany

Dynamically phase-cycled radial bSSFP (DYPR-SSFP) is a recently proposed method for fast, banding-free bSSFP imaging. Based on a dynamically changing phase-increment in combination with a (quasi-) randomly sampled radial trajectory, images without banding artifacts can be obtained from one single acquisition. Up to now, the DYPR-SSFP concept has been combined with a 2D radial trajectory, yielding slight artifacts due to the applied dynamic phase-increment. Here, the combination with a 3D radial trajectory is proposed, effectively mitigating this drawback and allowing the generation of 3D banding-free bSSFP images with high isotropic resolution.

Ozan Sayin¹, Haris Saybasili², M. Muz Zviman³, Mark Griswold^4,5, Nicole Seiberlich⁵, and Daniel A. Herzka^1
1Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²Siemens Healthcare USA, Inc., Chicago, IL, United States, ³Department of Medicine, Cardiology, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ⁴Department of Radiology, Case Western Reserve University, Cleveland, OH, United States, ⁵Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States

Recently, a well-established parallel imaging technique GRAPPA, has been successfully extended to radial imaging via an improved calibration scheme that extends the calibration to the time dimension (through-time radial GRAPPA calibration), and has demonstrated high acceleration factors in real-time imaging. The current study aims to eliminate calibration scans while preserving image quality. A novel self-calibration method is proposed and validated in real-time cardiac imaging in swine and normal subjects. Parameters related to self-calibration are explored for optimal image reconstruction of high acceleration rates (R=9-12).

Giuseppe Valvano^1,2, Nicola Martini², Dante Chiappino², Luigi Landini^1,2, and Maria Filomena Santarelli^2,3
1Department of Information Engineering, University of Pisa, Pisa, PI, Italy, ²Fondazione G. Monasterio CNR-Regione Toscana, Massa, MS, Italy, ³Institute of Clinical Physiology, CNR, Pisa, PI, Italy

A new sampling strategy for Compressive Sensing cardiac cine MRI based on variable density spirals is presented. The low coherence needed for a good reconstruction in Compressive Sensing was achieved delaying the gradient waveforms starting from random positions. To further reduce the coherence we rotated with random angles each spiral. The proposed strategy was validate by means of off-line reconstruction of a cardiac cine dataset. With this approach we were able to reconstruct the dataset with good image quality up to a 5-fold acceleration.

Kie Tae Kwon¹, Adam B Kerr¹, and Dwight G Nishimura^1
1Stanford University, Stanford, CA, United States

A sliding interleaved cylinder acquisition has previously been incorporated into a steady-state free precession sequence to achieve improved artery-vein contrast in the lower extremities. In this work, we extended the sequence to acquire navigator echoes without additional scan time to allow more flexibility in selecting an RF excitation pulse. A set of in vivo experiments on healthy volunteers demonstrated that the proposed scheme allowed the use of a non-linear-phase RF excitation pulse with a shorter TR and a sharper slab profile, which improved the robustness of the sequence.

Dinghui Wang¹, Zhiqiang Li¹, and James G. Pipe^1
1Neuroimaging Research, Barrow Neurological Institute, Phoenix, Arizona, United States

3D MP-RAGE was implemented with distributed spirals (DS) to increase scan efficiency and flexibility of scan parameters. Water and fat were separated and deblurred using data collected from two interleaved TEs. Data were acquired with different readout times. Compared to Cartesian reference images, spiral images showed similar overall contrast and sharpness, and higher signal to noise ratio (SNR) despite of shorter scan times. As the spiral readout time increases, the SNR increases and the total scan time decreases. The results suggest the feasibility of 3D DS MP-RAGE with readout time up to 20ms for high-resolution anatomical neuroimaging.

Keigo Kawaji¹, Hui Wang², Sui-Cheng Wang^1,3, Akiko Tanaka⁴, Takeyoshi Ota⁴, Roberto M. Lang¹, and Amit R. Patel^1
1Medicine, Section of Cardiology, The University of Chicago, Chicago, Illinois, United States, ²Philips Medical Systems, Cleveland, Ohio, United States,³Biomedical Engineering, Northwestern University, Evanston, Illinois, United States, ⁴Surgery, The University of Chicago, Chicago, Illinois, United States

In k-space segmented radial sampling strategies, radial trajectories can be acquired repeatedly over multiple consecutive cycles. In this study, we propose a novel interleaving method for 2D or 3D stack-of-stars imaging that uses the property of prime numbers and modular arithmetic (Modular Prime Spokes: MoPS) to provide efficient coverage of k-space within any desired temporal reconstruction window. The MoPS interleaving was demonstrated in a 3-minute cardiac scan with 3D stack-of-stars covering a large >1R-R acquisition window. 15ms temporal resolution reconstructions with 2ms sliding windows were used to investigate filtering methods in the temporal frequency domain for removing streaking artifacts.

Vimal Singh¹ and David Ress^2
1Electrical Engineering, University of Texas at Austin, Austin, Texas, United States, ²Neuroscience, Baylor College of Medicine, Hosuton, Texas, United States

This work presents a simple theory that accounts for the interaction of any k-space trajectory and echo time on BOLD contrast. The theory quantifies the need for different TEs to obtain best contrast for different acquisition trajectories. It also allows comparison of BOLD contrast available from different trajectories over various echo times. The theory was tested by performing high-resolution fMRI in superior colliculus using a variety of single- and dual-echo spiral trajectories and EPI as echo time was varied. . The proposed theory shows a satisfactory fit to the empirical data.

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

In golden angle radial MRI a constant azimuthal radial profile spacing of 111.246...° guarantees a nearly uniform azimuthal profile distribution in k-space for an arbitrary number of radial profiles and was recently used in various real-time imaging methods. However, in combination with balanced SSFP sequences the large azimuthal angle increment may lead to strong image artifacts, due to the varying eddy currents introduced by the rapidly switching gradient scheme. We introduce a sequence of smaller angles (49.750…°, 32.039... °, 27.198... °, 23.628...°, ... ), based on a generalized Fibonacci sequence, that guarantee the same sampling efficiency as the golden angle.

Frank Ong¹, Martin Uecker¹, Wenwen Jiang², and Michael Lustig^1
1Electrical Engineering and Computer Sciences, University of California, Berkeley, Berkeley, California, United States, ²Bioengineering, UC Berkeley/UCSF, Berkeley, California, United States

We present a method to accelerate almost all non-Cartesian MR reconstruction methods using pruned FFT. Contrary to common belief, we show that no memory overhead is required for any oversampling factors in non-Cartesian reconstruction. For iterative methods, we also propose partial pruning to approximate the non-Cartesian Fourier Transform operator to speed up each iteration while guaranteeing convergence. We apply our proposed method on compressed sensing and parallel imaging reconstruction of in vivo datasets and show that our proposed method reduces the computation time for non-Cartesian image reconstruction with gridding and toeplitz-circulant embedding.

Thomas Boyd Martin^1,2, Yi Wang², Steen Moeller³, Kyung Sung⁴, and Danny JJ. Wang^2
1Biomedical Physics Interdepartmental Program, University of California Los Angeles, Los Angeles, California, United States, ²Neurology, University of California Los Angeles, Los Angeles, California, United States, ³Center for Magnetic Resonance Research, University of Minnesota, Minnesota, United States, ⁴Radiological Sciences, University of California Los Angeles, Los Angeles, California, United States

CAIPIRINHA is an acceleration technique that uses phase modulated multiband excitation pulses to simultaneously acquire slices. Its application in balanced-SSFP (CAIPI-bSSFP), however, has been limited because the phase modulation of CAIPIRIHNA results in shifted off-resonance profiles and subsequent banding artifacts in simultaneously excited slices. A unique case of an SSFP-FID sequence allows for removing banding artifacts while maintaining the unique bSSFP tissue contrast by averaging the bSSFP signal profile (integrated-SSFP or iSSFP). This study demonstrates the combination of CAIPIRINHA and iSSFP techniques (CAIPI-iSSFP) up to 4 times acceleration while removing the banding artifacts seen in CAIPI-bSSFP imaging.

Nathan S Artz^1,2, Curtis N Wiens¹, Matthew R Smith¹, Diego Hernando¹, Alexey Samsonov¹, and Scott B Reeder^1,3
1Department of Radiology, University of Wisconsin, Madison, WI, United States, ²Department of Radiological Sciences, Saint Jude Children's Research Hospital, Memphis, TN, United States, ³Department of Medical Physics, University of Wisconsin, Madison, WI, United States

A spectrally-resolved, fully phase-encoded (SR-FPE) 3D FSE technique can avoid artifacts near metal due to frequency-encoding, but long scan times have limited work to phantoms. The purpose of this work was to translate SR-FPE to the in-vivo setting using multiband RF excitation to accelerate imaging. In a volunteer with a total knee replacement, an interleaved tri-band SR-FPE approach successfully acquired six RF offsets in 13:19min. In a volunteer with the head of a hip prosthesis placed posterior to the knee, 24 RF offsets were acquired in 34:38min. This work demonstrates the feasibility of in-vivo, multiband SR-FPE near metallic implants.

Fat Water Separation

Tuesday 2 June 2015

Karim B Samji^1,2, Abdulmohsen Alrashed^1,2, Wael M Shabana^1,2, Matthew DF McInnes^1,2, and Nicola Schieda^1,2
1Department of Medical Imaging, The Ottawa Hospital, Ottawa, ON, Canada, ²University of Ottawa, Ottawa, ON, Canada

T1W TSE imaging is fundamental for prostate cancer staging with MRI. This study compared a modified high-resolution free breathing 2-Point Dixon GRE sequence with fat-water separation (FLEX-LAVA) to T1W TSE as a potential time saving measure. There was no difference in detection of nodal or skeletal metastases and image quality was comparable or slightly improved with FLEX-LAVA. Our results suggest that T1W TSE can be safely replaced in prostate cancer MRI examinations using a high resolution 2-Point Dixon GRE sequence therefore decreasing examination time.

Yuji Iwadate¹, Kunihiro Miyoshi², Masanori Ozaki², and Hiroyuki Kabasawa^1
1Global MR Applications and Workflow, GE Healthcare Japan, Hino, Tokyo, Japan, ²MR Engineering, GE Healthcare Japan, Tokyo, Japan

Pencil beam navigator echo is sensitive to off-resonance, and undesired excitation of subcutaneous fat is caused at 3T. We modified a previously reported 2-point Dixon reconstruction for pencil beam navigator with dual-echo acquisition. Volunteer scans were performed on a 3 T imaging system with the proposed method, and dark bands was reduced in the navigator signal, which resulted in accurate motion detection.

Wei Wang¹, Xiuzhong Yao¹, Hongmei Yan², Hua Bian², Xiaodong Zhong³, Radhouene Neji⁴, Caixia Fu⁵, Hui Liu⁶, Dehe Weng⁵, Ignacio Vallines⁶, and Mengsu Zeng^1
1Radiology Department, Zhongshan Hospital, Fudan University, Shanghai, Shanghai, China, ²Endocrinology Department, Zhongshan Hospital, Fudan University, Shanghai, China, ³MR collaborations, Siemens Healthcare, Atlanta, Georgia, United States, ⁴MR collaborations, Siemens Healthcare, Frimley, Camberley, United Kingdom, ⁵Application Department, Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, Guangdong, China, ⁶MR collaborations, Siemens Healthcare, Shanghai, China

Accurate non-invasive detection and quantification of proton density fat fraction (PDFF) as a marker for liver fat in patients with non-alcoholic fatty liver disease (NAFLD) is gaining increasing interest. The aim of this study was to evaluate both rapid single breath-hold, multi-echo, T2 corrected single-voxel spectroscopy( HISTO ), and a recently developed, multi-echo 3D gradient echo acquisitions using a hybrid multi-step fitting approach (Advanced Dixon, AD) with conventional, invasive liver biopsy as a reference for hepatic fat quantification in NAFLD patients.

Kang Wang¹, Ken-Pin Hwang², Zachary Slavens³, and Ersin Bayram^2
1Global Applications and Workflow, GE Healthcare, Madison, WI, United States, ²Global Applications and Workflow, GE Healthcare, Houston, TX, United States,³MR Engineering, GE Healthcare, Waukesha, WI, United States

Conventional 2-pt Dixon water-fat separation method requires the input source echo images to be close to in-phase (IP) and out-of-phase (OOP). This makes no assumptions about the water-percentage (or fat-percentage) of the pixel data. In this work, the assumption that in vivo pixels are either very water-dominant or fat-dominant was used, and it was experimentally demonstrated that the input echo source images can substantially deviate from the ideal IP and OOP echo times while still obtaining robust water-fat separation using the unchanged IP-OOP signal model for water-fat separation. Theoretical analysis is also given.

Dong Zhou¹, Jianwu Dong², Pascal Spincemaille¹, Ashish Raj¹, Martin Prince¹, and Yi Wang^1
1Weill Cornell Medical College, New York, NY, United States, ²Tsinghua University, Beijing, China

In this work, we present a two-point Dixon method with flexible echo times where the smoothness of the inhomogeneity field is imposed as an non-convex energy minimization problem, solved by Quadratic Pseudo-Boolean Optimization (QPBO). Robust water/fat separation is achieved in in contrast enhanced dynamic liver imaging at 1.5T.

Jedrzej Burakiewicz¹, Melissa T. Hooijmans¹, Erik H. Niks², Jan J.G.M. Verschuuren², Andrew G. Webb¹, and Hermien E. Kan^1
1Department of Radiology, Leiden University Medical Center, Leiden, Zuid Holland, Netherlands, ²Department of Neurology, Leiden University Medical Center, Leiden, Zuid Holland, Netherlands

Robust olefinic fat suppression is of great importance in diffusion tensor imaging (DTI) of skeletal muscle. Current methods may be susceptible to main field inhomogeneities or reduce image quality. We propose a novel use of combined 6- and 2- point Dixon techniques with shortened echo time to eliminate olefinic fat signal in skeletal muscle and demonstrate the feasibility of the new method in healthy volunteers.

Jan Hendrik Wülbern¹, Mariya Doneva¹, Holger Eggers¹, Christian Stehning¹, and Peter Börnert^1
1Philips Research Europe, Hamburg, Hamburg, Germany

Combined radial stack of stars k-space sampling with dual-echo readout Dixon water-fat separation is demonstrated. This is enabled by a radial phase correction, which is performed on-the-fly without requiring pre-scan calibrations relying on imaging data only. Two types of radial sampling schemes are considered: uniform angular sampling with alternating readout directions and golden angle sampling. The phase correction method preserves phase information for Dixon methods, is robust to radial undersampling, stable over long scan durations, and works for golden angle acquisitions.

Gabriele Beck¹, Alan Huang¹, Gert van Ijperen¹, Lars van Loon¹, and Marko Ivancevic^1
1Philips Healthcare, Best, Netherlands

Despite its superb fat suppression characteristics, Dixon TSE is known to be sensitive to motion artifacts. In this study we investigated a novel partial averaging approach for Dixon TSE acquisition, where the k-space center was sampled denser compared to the k-space periphery and randomization was used. Simulation, phantom and volunteer experiments demonstrate the reduced motion ghosting and show that it outperforms the longer two average scans.

Dirk Krüger¹, Silvia Lorrio González¹, and René M. Botnar^1
1Division of Imaging Sciences & Biomedical Engineering, King's College London, London, United Kingdom

The aim of this project is to improve and validate a fat saturation technique for MR imaging with magnetic nanoparticles (MNPs). Inversion Recovery with ON-resonant water suppression (IRON) has been shown to produce reliable positive contrast images with MNPs. The most common fat saturation technique in combination with IRON is a spectrally selective pre-saturation pulse (SPIR). We used a two echo Dixon method instead of SPIR to achieve fat saturation in a phantom study. The Dixon method demonstrated superior fat suppression ability compared to SPIR while achieving the same positive contrast of MNP rich areas of the phantom.

JaeJin Cho¹, Changheun Oh¹, Kinam Kwon¹, and HyunWook Park^1
1Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Chungcheong, Korea

Water-fat separation is essential technique for accurate diagnosis in many area of MRI. If a parallel imaging method is applied to water-fat separation in the spectral direction, imaging time can be reduced. However, conventional water-fat separation methods including IDEAL cannot use sensitivity difference of multichannel RF coil because water and fat images share the same FOV. In this paper, a new water-fat separation method is proposed by using the coil sensitivity map and by encoding difference of resonant frequency between water and fat.

Felix Lugauer¹, Dominik Nickel², Jens Wetzl¹, Berthold Kiefer², and Joachim Hornegger^1
1Pattern Recognition Lab, Department of Computer Science, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Siemens AG, Healthcare, Imaging & Therapy Systems, Magnetic Resonance, Erlangen, Germany

Multi-contrast water-fat separation based on the Dixon method is gaining importance in clinical routine. A combination with iterative reconstruction also addressing field inhomogeneities, relaxation and eddy current effects is, however, not straightforward as the optimization is rendered non-convex. Here we demonstrate that water-fat separation can be decoupled by first reconstructing the multiple echos using a locally low-rank regularization. This enforces a representation of the contrast images with as few chemical components as possible, assuming a low-resolution phase evolution. Both are common assumptions. The approach allows bipolar acquisitions, varying sampling patterns across contrasts and promises superior image quality over conventional reconstructions.

Johan Berglund^1
1Karolinska Institutet, Stockholm, Sweden

Several water/fat separation techniques use graph cuts to resolve the B0 field map prior to water and fat component estimation in each voxel separately. These algorithms have demonstrated robustness to severe field inhomogeneity, but the tolerance to noise has not been well examined. A graph cut based algorithm was modified to operate at multiple resolution levels in order to resolve the field map in a coarse-to-fine manner. The algorithm was evaluated on benchmark datasets with added noise to synthesize a range of SNR levels. The multi-scale approach was demonstrated to increase the tolerance to noise in the input data.

Stefan Ruschke¹, Holger Eggers², Hendrik Kooijman³, Pia M. Jungmann¹, Axel Haase⁴, Ernst J. Rummeny¹, Thomas Baum¹, and Dimitrios C. Karampinos^1
1Department of Diagnostic and Interventional Radiology, Technische Universität München, Munich, Bayern, Germany, ²Philips Research, Hamburg, Hamburg, Germany, ³Philips Healthcare, Hamburg, Hamburg, Germany, ⁴Zentralinstitut für Medizintechnik, Technische Universität München, Garching, Bayern, Germany

Chemical shift encoding-based water-fat imaging has been emerging for quantifying skeletal muscle fat content. In regions with low fat content, magnitude-based techniques have been used to overcome the sensitivity of complex-based techniques to phase errors. However, magnitude-based techniques can become unstable for certain combinations of echo times, when the chemical shift separation between water and fat is not known, e.g. due to the susceptibility-induced fat resonance shift effects in skeletal muscle. The present study aims to characterize complex-based and magnitude-based methods for water-fat separation in skeletal muscle, where both fat resonance shift and phase errors can be present.

Dian Liu¹, Jelena Curcic^1,2, Andreas Steingoetter^1,2, and Sebastian Kozerke^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland, ²Division of Gastroenterology and Hepatology, University Hospital Zurich, Zurich, Switzerland

For quantifying the dynamics of gastrointestinal fat digestion, imaging efficiency of multipoint gradient echo methods must be improved. A reconstruction of fat fraction (FF) maps for parallel MRI using signal model-based dictionaries is proposed and studied for the prospective and retrospective data acquisition of intragastric fat distribution, resulting in improved reconstruction. The performance is subject to variations due to different echo times. FF of undersampled data was found to be in good agreement with FF of the fully sampled acquisition. Underestimation occurred at higher fat fractions, which is however tolerable in terms of the relative error.

Chuanli Cheng^1,2, Chao Zou¹, Hairong Zheng¹, and Xin Liu^1
1Paul C. Lauterbur Biomedical Imaging Research Center, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, Guangdong, China, ²University of Chinese Academy of Sciences, Beijing, Beijing, China

A novel multiresolution region growing algorithm is introduced for robust fat water separation and accurate field map estimation from three-point non-equally spaced multi-echo images. The non-equally spaced TEs scheme decreases the number of false local minimum and therefore the probability of fat water swap. The seed region finding and regional growing are executed in multiple resolutions independently, which avoids the premature convergence in field search due to drastic field change in the finer resolution. The algorithm was tested on c-spine and ankle data and shown to be robust in large field inhomogeneity and disjoint areas.

Stefan Ruschke¹, Holger Eggers², Hendrik Kooijman³, Thomas Baum¹, Marcus Settles¹, Axel Haase⁴, Ernst J. Rummeny¹, and Dimitrios C. Karampinos^1
1Department of Diagnostic and Interventional Radiology, Technische Universität München, Munich, Bayern, Germany, ²Philips Research, Hamburg, Hamburg, Germany, ³Philips Healthcare, Hamburg, Hamburg, Germany, ⁴Zentralinstitut für Medizintechnik, Technische Universität München, Garching, Bayern, Germany

Phase errors are known to cause significant bias in fat fraction estimation using complex-based water-fat separation. Time-interleaved multi-echo gradient-echo (TIMGE) acquisitions (where echoes are time interleaved and acquired in multiple TRs) are highly desirable at 3 T, as they enable high spatial resolution while maintaining short echo time steps. However, their phase error behavior has not been previously addressed. The present study aims to decompose the phase errors in TIMGE and propose a methodology to correct for the effect of phase errors in TIMGE acquisitions.

M. Dominik Nickel¹, Stephan A.R. Kannengiesser¹, and Berthold Kiefer^1
1MR Applications Development, Siemens Healthcare, Erlangen, Germany

Advanced multi-gradient-echo imaging based proton density fat fraction estimation requires complex-valued fat signal dephasing factors, which are typically derived from a multi-peak fat spectral model. Several, slightly differing models have been published, but may not be universally applicable. We show how the dephasing factors can be easily derived in the time domain from well-known single-breath-hold multi-echo single-voxel STEAM spectroscopy, making individualized calibration feasible. Based on liver data acquired in one volunteer, it is shown that the dephasing factors from spectral model and spectroscopy match very well, and produce very similar results when applied to 3D multi-gradient-echo fat fraction imaging.

Abraam S Soliman^1,2 and Charles A McKenzie^1,3
1Biomedical Engineering, University of Western Ontario, London, Ontario, Canada, ²Robarts Research Institute, Imaging Research Laboratories, London, Ontario, Canada, ³Medical Biophysics, University of Western Ontario, London, Ontario, Canada

Chemical-shift encoded-water fat imaging is usually performed using multi-gradient-echo sequence. Unipolar echoes are acquired over multiple TR in order to achieve optimal echo-spacing. Although bipolar acquisitions are more efficient than the unipolar ones, they suffer from phase and magnitude errors that disrupt the fat quantification process. We have recently proposed a new bipolar sequence that can overcome these errors and provide accurate fat measurement. In this work we demonstrate the efficiency of this sequence for whole liver imaging, commonly used for the diagnosis of non-alcoholic fatty liver disease (NAFLD).

Ahmed Fahmy¹, El-Sayed H. Ibrahim², and Nael Osman^3
1Cairo University, Cairo, Egypt, ²University of Michigan, Ann Arbor, MI, United States, ³Johns Hopkins University, Baltimore, MD, United States

Stimulated-echo acquisition mode (STEAM) is a key pulse sequence in MRI in general, and in cardiac imaging in particular, as it allows for marking the modulated magnetization and saving it from rapid T2 relaxation. Speeding up the temporal-resolution of STEAM-based sequences is desirable, although it is compromised in cases when fat-suppression is applied. In this work, we present an efficient fat-suppression technique (Spectrally-Presaturated Modulation (SPM)) for STEAM-based sequences without affecting the temporal-resolution, scan-time, or SAR-level compared to other fat suppression techniques (e.g. spectral-spatial selective-pulses (SSSP) and chemical-shift selective (CHESS)), which could result in accurate parameter measurement and improved image analysis.

Xiaoke Wang¹, Diego Hernando², and Scott B. Reeder^2,3
1Biomedical Engineering, University of Wisconsin, Madison, Wisconsin, United States, ²Radiology, University of Wisconsin, Madison, Wisconsin, United States,³Medical Physics, University of Wisconsin, Madison, Wisconsin, United States

The estimation of PDFF by CSE imaging may be biased by T1 relaxation. T1 related bias can be minimized by using a small flip angle (SFA) approach. SFA results in reduced SNR and residual bias. T1 bias can also be corrected using a dual flip angle method (standard DFA), which redundantly estimates R2* and B0 twice. In this study, we propose a joint fitting of T1, B0, R2* and PDFF based on dual-flip-angle acquisition (joint DFA). Joint DFA slightly reduced noise in R2* and PDFF estimates compared with standard DFA as shown by Cramer-Rao lower bound and Monte Carlo simulation.

Davide Piccini^1,2, Peter Kellman³, Diego Hernando⁴, Simone Coppo², Gabriele Bonanno², and Matthias Stuber^2
1Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland, ²Department of Radiology, University Hospital (CHUV) and University of Lausanne (UNIL) / Center for Biomedical Imaging (CIBM), Lausanne, Switzerland, ³Laboratory of Cardiac Energetics, National Institutes of Health/NHLBI, Bethesda, Maryland, United States, ⁴Department of Radiology, University of Wisconsin-Madison, Madison, Wisconsin, United States

Respiratory self-navigation (SN), using a modified 3D radial trajectory has successfully been tested for whole-heart coronary MRA achieving 100% scan efficiency. However, radial imaging in combination with standard CHEmical Shift Selective (CHESS) pulses exhibits a compromised fat-suppression performance due to fat magnetization recovery during the acquisition. A multi-echo version of SN 3D radial MRA was combined with the VARiable PROjection method (SN-VARPRO) for Dixon-like fat-water separation with flexible echo-times and compared to the SN-CHESS acquisition in volunteers. Increased fat suppression was achieved with SN-VARPRO. A trend for increased vessel sharpness and similar visualized length was obtained with respect to SN-CHESS.

Weiyi Chen¹ and Krishna S. Nayak^1
1Electrical Engineering, University of Southern California, Los Angeles, CA, United States

We propose a fast and efficient method to determine the effects of thermal noise in water-fat separation. Field map is estimated using graph-cut algorithm prior to the least square separation. We measured coil covariance matrix with a no-RF acquisition, then derived pixel-wise standard deviation maps for separated water and fat image by concatenating the corresponding linear operators. Finally we use Taylor approximation method to assess the variability of quantitative fat fraction measurement in ROIs. The proposed method is validated against the pseudo-replica technique, and is 300 times faster.

Esther Raithel¹, Gaurav Thawait², Shadpour Demehri², Shivani Ahlawat², Heiko Meyer¹, Wesley Gilson³, and Jan Fritz^2
1Healthcare Sector, Siemens AG, Erlangen, Bavaria, Germany, ²Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ³Siemens Healthcare USA, Baltimore, MD, United States

High-spatial-resolution isotropic 3D MRI can eliminate partial volume effects and enable multi-planar and curved reconstructions, as well as interactive 3D interpretation. In the shoulder, however, 3D MRI requires time-consuming oversampling steps in phase and slice encoding directions, which can result in clinically impractical acquisition times. K-space undersampling and iterative reconstruction is a method that can yield substantial acceleration of 3D data acquisition. We show the implementation of this method into a comprehensive, 11 min clinical 3D SPACE protocol that can produce similar image qualities as a 21 min standard 2D TSE protocol.

Benjamin Leporq¹, Simon Auguste Lambert^1,2, Francois Cauchy^1,3, Imane Boucenna⁴, Pierre Colinart⁴, Maxime Ronot^1,5, Valerie Vilgrain^1,5, Valerie Paradis^1,6, and Bernard Edgar Van Beers^1,5
1Center of research on inflammation, Paris 7 University; INSERM U1044, Paris, France, ²BHF Centre of Excellence, Division of Imaging Sciences and Biomedical Engineering, King’s College London King’s Health Partners, St. Thomas’ Hospital, London, United Kingdom, ³Department of HPB and liver transplantation, Beaujon University hospital Paris Nord, Clichy, France, ⁴Matière et systèmes complexes, Paris 7 University; CNRS UMR 7057, Paris, France,⁵Department of Radiology, Beaujon University hospital Paris Nord, Clichy, France, ⁶Department of Pathology, Beaujon University hospital Paris Nord, Clichy, France

This work report a MRI method able to quantify both fat content and fatty acid composition on preclinical systems at 7.0T. MR acquisition was based on a spoiled multiple gradient echoes sequence with bipolar readout gradient. After phase unwrapping, complex images were rebuilt. By using a model including water and eight fat resonances each expressed according to ndb, nmidb and CL, parametric images such as PDFF, SFA, MUFA and PUFA were reconstructed. Results suggest that it possible to follow diet effects on fat content and fatty acid compositions in mice with this method.

Motion Correction

Tuesday 2 June 2015

Benjamin Zahneisen¹, Aditya Singh², Michael Herbst², and Thomas Ernst^2
1Stanford University, Stanford, CA, United States, ²University of Hawaii, HI, United States

One of the barriers for using Prospective Motion Correction (PMC) in the clinic is the unpredictable nature of a scan because of the direct interference with the imaging sequence. Here, we suggest using the framework of retrospective motion correction to reverse the effects of prospective motion correction (“reverse retrospective correction”) for brain scans. The most important impact will most likely be in the clinical application, where our approach guarantees that a data set can be presented whose quality is at least as good as a scan acquired without PMC.

Aditya Singh¹, Brian Keating¹, Benjamin Zahneisen¹, Michael Herbst¹, and Thomas Ernst^1
1John A. Burns School of Medicine, University of Hawaii, Honolulu, Hawaii, United States

Prospective motion correction for brain MRI using external tracking systems with skin-attached markers may suffer from errors in head tracking data introduced by changes in facial expressions, such as squinting. We demonstrate the feasibility of detecting non rigid-body motion events using single 6-DOF information, with an algorithm that is validated on motion data obtained from a trained volunteer and seven clinical subjects who performed involuntary motion. The receiver operation characteristic of the algorithm was calculated to show a mean false positive rate of 0.09, true positive rate of 0.38 and a positive predictive value of 0.86.

José P. Marques¹ and Daniel Gallichan^1
1CIBM, EPFL, Lausanne, Vaud, Switzerland

tracking Discrete Off-Resonance markers with Three Spokes (trackDOTS) allows the localization of a large number of markers (12 in the current implementation) with very high temporal resolution (24ms). Additionally, with the same acquisition, the background frequency fluctuations can be monitored. In this work we evaluate the motion tracking accuracy and precision of trackDOTS when using 12 spherical markers filled with acetic acid. Additionally we demonstrate its potential to perform dynamic shimming by showing its sensitivity to respiration induced frequency fluctuations.

Julian Maclaren¹, Murat Aksoy¹, Benjamin Zahneisen¹, and Roland Bammer^1
1Department of Radiology, Stanford University, Stanford, CA, United States

Optical prospective motion correction requires that the position and orientation of the tracking system (typically a camera) is known relative to the reference frame of the MRI scanner. This provides a challenge for systems where the camera is frequently removed and reinstalled. One such example is when the camera is mounted on the head coil, which then moves in and out of the scanner with the patient table. In this work, we use simulations and in vivo measurements to investigate how precisely the camera needs to be repositioned in order to maintain good quality prospective motion correction.

Martin Eschelbach¹, Yu-Chun Chang¹, Jonas Handwerker², Jens Anders², Anke Henning^1,3, and Klaus Scheffler^1
1High-Field Magnetic Resonance Center, Max Planck Institute for Biological Cybernetics, Tuebingen, BW, Germany, ²Institute of Microelectronics, University of Ulm, Ulm, BW, Germany, ³Institute for Biomedical Engineering, ETH Zürich, Zurich, Switzerland

In this work, ¹⁹F NMR field probes are used to track subject head motion and field fluctuations due to motion and breathing at a 9.4 T human scanner. The field probes are rigidly attached to the subject’s head via a bite bar. A custom made transmit/receive chain using a PCB and external signal processing prevents the use of scanner receive channels. With this setup, position measurements are possible with a standard deviation of 57 µm or smaller depending on the axis as well as field measurements with a standard deviation of 3 Hz.

Enhao Gong¹, Qiyuan Tian¹, Jennifer A McNab², and John Pauly^1
1Electrical Engineering, STANFORD UNIVERSITY, Stanford, California, United States, ²Radiology, STANFORD UNIVERSITY, Stanford, California, United States

Accurate and efficient motion estimation is critical in MRI studies to achieve better anatomical resolution and faithful physiological analysis. However, most of the existing quantitative motion estimation methods require great amount of extra time cost for acquisition, post-processing, extensive registration or additional hardware/equipment. Here we propose an innovative method that accurately estimates motion using the noise correlation information intrinsically existing in all multi-channel MRI signals. This approach enables both real-time and prospective motion estimation without modifying acquisition or adding equipment. The proposed method is named: Motion Estimation from Noise Intrinsic Correlation between RF Channels (MECHANICS)

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

We recently introduced the concept of a very highly accelerated fat-excitation 3D-GRE acquisition as a motion-navigator to detect very small involuntary head-motion of compliant subjects for high-resolution imaging applications. In this study we investigate the accuracy and precision of the navigator at various resolutions and acceleration factors. At the highest acceleration factor used (2mm resolution, 8x8=64 acceleration) the FatNav still gives reasonable image quality, but estimated motion parameters become biased. At matched TR per volume, a 4mm resolution, 4×4=16 acceleration gave the best compromise between bias and accuracy, resulting in a robust navigator in 288ms.

Nicolas Pannetier^1,2, Theano Stavrinos², Peter Ng², Michael Herbst^3,4, Maxim Zaitsev⁴, Karl Young¹, Gerald Matson^1,2, and Norbert Schuff^1,2
1Radiology, UCSF, San Francisco, CA, United States, ²VAMC, San Francisco, CA, United States, ³Radiology, JABSOM, Honolulu, HI, United States, ⁴Radiology, University Medical Center Freiburg, Freiburg, Germany

We provide a framework to quantitatively evaluate the image quality improvement provided by prospective motion correction (PMC) techniques while considering the intrinsic motion variations between MRI acquisitions. We tested this framework to compare 2 marker setups and we show that considering the intrinsic motion as a covariate changes the statistical significance of the comparison. This framework could be used in larger studies to compare efficiently different PMC setups and fixation markers.

Emre Kopanoglu¹, Gigi Galiana¹, and Robert Todd Constable^1
1Diagnostic Radiology, Yale University, New Haven, Connecticut, United States

Nonlinear gradient fields vary along multiple spatial directions. Therefore, spatial information along multiple directions is encoded simultaneously. By using an array of receiver coils, the spatial variation along different directions can be recovered. In this study, this encoding capability is exploited to reconstruct low-resolution images every TR, using a second-order gradient field and an 8-channel receiver-array. The reconstructed images were then used to track translational and rotational rigid in-slice motion, with sub-pixel precision. Limited to the same field strength inside the field-of-view as the linear gradient fields typically available clinically, the motion navigator lasts only 500 us, including rewinding.

Murat Aksoy¹, Julian Maclaren¹, Eric Peterson¹, and Roland Bammer^1
1Radiology, Stanford University, Stanford, CA, United States

EPI ghost correction often involves the acquisition of calibration data at the beginning of the scan with phase encoding turned off. This method works well in most cases, but does not handle the scenario where the gradient axes are rotated during the scan, such as during prospective motion correction. In this case, a more robust technique is required. In this study, we compared three different techniques in order to assess the importance of performing EPI ghost correction individually for each readout during prospective motion correction.

John W Grinstead¹, Himanshu Bhat², M. Dylan Tisdall³, Andre van der Kouwe³, William Rooney⁴, and Gerhard Laub^2
1Siemens Healthcare, Portland, USA, United States, ²Siemens Healthcare, USA, United States, ³A.A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, MA, United States, ⁴Advanced Imaging Research Center, Oregon Health & Science University, Portland, OR, United States

The MPRAGE pulse sequence is commonly used for T₁ -weighted imaging. Previous work showed that the relative signal intensity of blood in MPRAGE is dominated by the IR pulse, and by controlling the selectivity of this pulse across multiple measurements one could generate a subtraction MR angiogram (MRA) in addition to a standard T₁ MPRAGE. However, such a subtraction technique is sensitive to motion between the multiple measurements, limiting its applicability. Here we address this shortcoming with the addition of volumetric navigators for prospective motion correction, enabling simultaneous MPRAGE and non-contrast MRA even in the presence of subject motion.

Sui-Cheng Wang^1,2, Amit R. Patel², Akiko Tanaka³, Hui Wang⁴, Xiang Zhu⁵, Dianwen Zhang⁶, Takeyoshi Ota³, Roberto M. Lang², and Keigo Kawaji^2
1Biomedical Engineering, Northwestern University, Evanston, Illinois, United States, ²Medicine, Section of Cardiology, The University of Chicago, Chicago, Illinois, United States, ³Surgery, The University of Chicago, Chicago, Illinois, United States, ⁴Philips Medical Systems, Cleveland, Ohio, United States, ⁵College of Information and Electrical Engineering, and College of Economics & Management, China Agricultural University, Beijing, China, ⁶Imaging Technology group, Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, Illinois, United States

Myocardial T1 mapping, which is used to detect diffuse fibrosis and quantify extracellular volumes, often employs 8-10 mm 2D slice acquisitions. However, this approach may be unsuited for accurate quantification of small tissues samples. 3D T1 mapping is desirable, and a FAN-type view ordering has recently been shown to improve through-plane resolution over multiple breath-holds. In this study, we propose a new view order method called NINJA-STAR (or NJS), which allows efficiently acquisition of 3D MRI k-space in a reduced number of breath-holds, and examine this approach in both phantom experiments and in-vivo.

Stefan Wundrak^1,2, Jan Paul¹, Johannes Ulrici², Erich Hell², Margrit-Ann Geibel¹, and Volker Rasche^1
1Ulm University, Ulm, Baden-Württemberg, Germany, ²Sirona Dental Systems, Bensheim, Hessen, Germany

Assessment of the motion of the temporomandibular joint is of interest for a variety of pathologies. The dynamic visualization of the TMJ under realistic mastication is still limited by the poor spatiotemporal resolution. In this contribution we combine the advantages of self-gating and adaptive averaging by using contour fitting in the correlation matrix (CoFi-CoMa) to increase the achievable temporal resolution and SNR. Moving images of the TMJ at this fast opening / closing cycle rate in combination with the achieved high temporal resolution and high SNR have not been shown before.

Bjorn Stemkens¹, Rob HN Tijssen¹, Baudouin Denis de Senneville^2,3, Jan JW Lagendijk¹, and Cornelis A.T. van den Berg^1
1Department of Radiotherapy, UMC Utrecht, Utrecht, Netherlands, ²Image Science Institute, UMC Utrecht, Utrecht, Netherlands, ³IMB, UMR 5251 CNRS/University of Bordeaux, Bordeaux, France

The introduction of MR-guided radiotherapy allows real-time tracking of mobile tumors using fast 2D image navigators. In this work, a PCA-based motion model is introduced which characterizes the dynamic 3D motion of tissue outside the field-of-view of the 2D image navigators in order to estimate the deposited radiation dose. The PCA model was formed based on a 4D-MRI (training data) and updated using a coronal 2D navigator (imaging during treatment) and independently reviewed using a sagittal 2D navigator. The root-mean-squared error was below 2 mm in all directions. Both intra- and inter-cycle respiratory motion was visible in the motion trajectories.

Robin Simpson¹, Benjamin Knowles¹, Marius Menza¹, Michael Herbst^1,2, Cris Lovell-Smith¹, Maxim Zaitsev¹, and Bernd Jung^3
1Medical Physics, University Medical Centre, Freiburg, Germany, ²John A. Burns School of Medicine, Hawaii, United States, ³University Hospital of Bern, Switzerland

Removing respiratory motion is an important challenge for cardiac MRI. This abstract presents initial work to apply a novel method of prospective respiratory gating using ShapeTape, a bend-sensitive optical-fibre-based motion detection system. Gating can be performed without interrupting imaging, and motion is sampled at high rates. This allows gating separately for each cardiac frame, improving image quality towards the end of the cardiac cycle.

Yuji Iwadate¹, Kunihiro Miyoshi², Masanori Ozaki², and Hiroyuki Kabasawa^1
1Global MR Applications and Workflow, GE Healthcare Japan, Hino, Tokyo, Japan, ²MR Engineering, GE Healthcare Japan, Tokyo, Japan

Pencil beam navigator echo suffers from undesired excitation outside the beam�fs area when the tracker placement is inappropriate. We developed a signal combination method with gradient reversal to cancel out the effects of the side-lobe with the smallest radius from the beam center. The proposed method diminished the undesired signal contamination in both phantom and volunteer scan, and the resultant navigator-gated 3D-SPGR image had less motion artifacts than the conventional method.

Evan Levine^1,2, Shreyas Vasanawala², Brian Hargreaves², and Manojkumar Saranathan^2
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States

In many imaging scenarios, data is corrupted by motion and signal intensity changes, the exact severity and characteristics of which are not known a priori. To enable motion robust abdominal imaging, we employ a new ky-kz-t sampling trajectory based on complementary Poisson-disc sampling that allows viewsharing to be reduce retrospectively using compressed sensing parallel imaging reconstruction. Like golden angle approaches for radial imaging, the trajectory allows reduced viewsharing data from any of several time frames and temporal footprints chosen retrospectively to be effectively reconstructed. In poor breath holding scenarios, motion-free images can be recovered with schemes that reduce temporal footprint.

Zarko Celicanin¹ and Oliver Bieri^1
1Radiological Physics, University of Basel Hospital, Basel, Switzerland

Imaging of the organ dynamics is an important pretreatment step in interventional MR and radiotherapy. A novel 2D multislice 5DMRI acquisition method is presented, in which retrospective sorting based on principle component analysis was used to gate the acquired images. Simultaneous respiratory- and cardiac- resolved imaging was demonstrated in a feasibility study.

Anne Menini¹, Vladimir Golkov^1,2, and Florian Wiesinger^1
1DIBT, GE Global Research, Garching b. München, Germany, ²Department of Computer Science, Technical University Munich, Garching b. München, Germany

In lung imaging, high-resolution 3D imaging is highly desirable but requires motion management. Previous studies have proposed prospective gating solutions but suffer from low scan efficiency. Here, based on the 3D radial zero-TE RUFIS sequence, we propose a new sampling pattern associated with a motion compensated reconstruction to take advantage of the whole scan time. Compared to more classical motion management methods, the proposed method shows better sharpness, bronchial tube depiction and SNR. The proposed method presents potential benefits for PET/MR since it provides high-resolution anatomical imaging, renders the lung density, and extracts a custom motion model within 1min45s of free-breathing acquisition.

Gastao Cruz¹, David Atkinson², Tobias Schaeffter¹, and Claudia Prieto^1
1Division of Imaging Sciences & Biomedical Engineering, King's College London, London, London, United Kingdom, ²Centre for Medical Imaging, University College London, London, London, United Kingdom

Respiratory motion is a major challenge in 3D abdominal MRI, creating ghosting and blurring. Bin-to-bin motion correction techniques have been proposed to overcome these problems and reduce the scan time. However, intra-bin motion is unaccounted for resulting in, potentially significant, residual motion artifacts. Here we propose an intra-bin motion correction approach based on the assumption of a linear relation between the non-rigid motion of each acquired data within the bin and the inter-bin non-rigid motion model. The proposed method is compared with the conventional gated reconstruction, showing significantly improved sharpness and image quality.

Wei Huang¹, Yang Yang², Xiao Chen², and Michael Salerno^1,3
1Medicine, University of Virginia, Charlottesville, Virginia, United States, ²Biomedical Engineering, University of Virginia, Charlottesville, Virginia, United States,³Radiology, University of Virginia, Charlottesville, Virginia, United States

While k-t accelerated CMR perfusion techniques (k-t PCA) and CS techniques (k-t SLR) enable highly accelerated acquisition, their image quality is significantly degraded by respiratory motion which frequently occurs in clinical studies. Non-rigid motion correction techniques which are applied iteratively can reduce motion artifacts, but the repeated application of spatial interpolation causes blurring. We describe a technique which derives rigid motion estimates from the heart region and performs a single linear phase shift of the acquired k-space data. We demonstrate that this motion correction strategy greatly improves k-t PCA and k-t SLR in the setting of respiratory motion.

Masami Yoneyama¹, Masanobu Nakamura¹, Makoto Obara¹, Tomoyuki Okuaki¹, Tetsuo Ogino¹, Yuriko Suzuki¹, Yuriko Ozawa², Takashi Tabuchi², Satoshi Tatsuno², Ryuji Sashi², and Marc Van Cauteren^1
1Philips Electronics Japan, Tokyo, Japan, ²Yaesu Clinic, Tokyo, Japan

In breast or thoracic-spine MRI, motion artifacts due to cardiac- or aortic-pulsation inevitably appear, particularly in the images after contrast media injection. Such artifacts can impede the diagnosis. In this study, we proposed a new technique for simply and effectively reducing cardiac/pulsatile motion artifacts on contrast-enhanced images by using a cylindrical labeling inversion pulse which can be combined with various sequences.

KS Shriram¹, Dattesh D Shanbhag², Sheshadri Thiruvenkadam², Venkata Veerendranadh Chebrolu², Sandeep N Gupta³, and Rakesh Mullick^4
1Biomedical Signal Analysis Laboratory, GE Global Research, Bangalore, Karnataka, India, ²Medical Image Analysis Laboratory, GE Global Research, Bangalore, Karnataka, India, ³Clinical Systems & Signal Processing, GE Global Research, Niskayuna, NY, United States, ⁴Diagnostics & Biomedical Technologies, GE Global Research, Bangalore, Karnataka, India

We have introduced a fast group-wise registration scheme for correction of non-rigid motion in prostate DCE exams. We demonstrate the robustness of the algorithm in correcting motion even in presence of coil shine through artifacts, and retaining signal characteristics. Overall, motion correction results in lower dispersion of pK parameters in a given tissue ROI and improved confidence in interpretation of DCE data.

Christopher W. Roy¹, Mike Seed^2,3, and Christopher K. Macgowan^1,3
1Medical Biophysics and Medical Imaging, University of Toronto, Toronto, Ontario, Canada, ²Labatt Family Heart Centre, Division of Cardiology, Department of Paediatrics, The Hospital for Sick Children, Ontario, Canada, ³Diagnostic Imaging, The Hospital for Sick Children, Toronto, Ontario, Canada

A reconstruction method for undersampled time resolved fetal cardiac imaging is presented. First, Metric Optimized Gating is used to reduce artifact from non-gated cardiac motion. Second, Compressed Sensing is employed to reduce artifact from random undersampling. The accuracy of this approach is investigated through a numerical simulation of a prospectively undersampled CINE cardiac MRI acquisition comprised of realistic cardiac motion, heart rate variability, and image noise. Initial feasibility in the fetal population is presented through retrospectively undersampled in utero MRI data.

Quantitative & Model-based Image Reconstruction

Tuesday 2 June 2015

Umit Yoruk^1,2, Manojkumar Saranathan¹, Tao Zhang¹, Brian A Hargreaves¹, and Shreyas S Vasanawala^1
1Radiology, Stanford University, Stanford, CA, United States, ²Electrical Engineering, Stanford University, Stanford, CA, United States

Many pharmacokinetic models in DCE-MRI need subject-based aortic input function (AIF). Accurate AIF measurement requires high temporal resolution images, which compromises the image spatial resolution. A recently developed low-rank reconstruction technique allows reconstruction of two sets of images, a high temporal resolution image for quantitative analysis and a high spatial resolution image for clinical evaluation, at a cost of increased reconstruction time. We present a faster method for extracting high temporal resolution AIF from DCE-MRI data and compare it to the low-rank reconstruction. The feasibility of this method was demonstrated on a pediatric subject.

Vimal Singh¹, David Ress², and Ahmed Tewfik^1
1Electrical Engineering, University of Texas at Austin, Austin, Texas, United States, ²Baylor College of Medicine, Houston, Texas, United States

High spatial resolution in fMRI generally improves its sensitivity to brain activation signals by reducing partial volume effects. However, the long acquisition times required for high spatial resolution limit the temporal resolution in fMRI studies. Consequently, the low temporal sampling bandwidth leads to increase in physiological noise and poor temporal modeling of the functional activation dynamics. This paper presents an under-sampled fMRI recovery using low-rank plus sparse matrix decomposition signal model. The preliminary results on in-vivo fMRI data show recovery of BOLD activation in superior colliculus with contrast-to-noise ratio > 4.4 (85% of reference) up to acceleration factors of 3.

Matthias Schloegl¹, Martin Holler², Kristian Bredies², and Rudolf Stollberger^1
1Institute of Medical Engineering, Graz University of Technology, Graz, Styria, Austria, ²Department of Mathematics and Scientific Computing, University of Graz, Graz, Styria, Austria

Optimal coil sensitivity estimation for the phase-consistent array combination and SENSE like reconstruction for undersampled dynamic and static MR data remains a challenge. The best possible quality for advanced reconstruction methods is bounded by the quality of coil-sensitivity estimation. Furthermore a growing number of MR applications require acceleration together with measurement of phase and changes in phase. We propose an iterative variational approach that takes advantage of a-priory knowledge for the magnitude and phase of the coil sensitivities and also for the complex transverse magnetization.

Hassan Haji-Valizadeh¹, Elwin Bassett², Ganesh Adluru³, Edward DiBella⁴, and Daniel Kim^4
1Radiology, University of Utah, Salt lake city, Utah, United States, ²University of Utah, Utah, United States, ³Ucair,Radiology, Salt lake city, Utah, United States, ⁴Ucair,Radiology, Utah, United States

Widespread clinical use of phase contrast MRI is restricted due to this modalities low data acquisition efficiency. Low efficiency may lead to low temporal and spatial resolution within a clinically acceptable breath hold. Compressed sensing in combination with low Rank Block Wise constraint is a promising strategy in increasing acquisition efficiency. The goal of this study is to characterize the efficacy of Low Rank Block Wise reconstruction by reconstructing retrospectively undersampled data, and comparing maximum velocity measure from reconstruction to maximum velocity obtained from fully sampled data sets. Agreement between maximum velocities will be achieved by using Bland Altman plots.

Zachary B Rodgers¹ and Felix W Wehrli^1
1Radiology, University of Pennsylvania, Philadelphia, PA, United States

Simultaneous measurement of intravascular blood T₁ and T₂ is a promising approach for fast, accurate quantification of whole-brain venous oxygen saturation (Y_v), as the T₁ values can be used to determine hematocrit, which is needed to convert measured T₂ to Y_v. We describe a modification of the recently developed TRUST T₂-quantification method, with addition of multiple EPI readouts (mTRUST) to allow concurrent measurement of T₁. mTRUST was applied in four subjects, demonstrating good precision for T₁ and T₂ estimation. T₁ values were slightly lower than recent literature, potentially due to incomplete inversion of blood present in the later EPI readouts.

Darryl McClymont¹, Irvin Teh¹, Hannah Whittington¹, and Jurgen Schneider^1
1University of Oxford, Oxford, Oxfordshire, United Kingdom

Compressed sensing offers a means to decrease the long scan times of diffusion tensor MRI (DTI) by acquiring only a subset of k-space. In this work, we present and evaluate an algorithm for the reconstruction of diffusion signals using data-driven dictionaries. Data from one ex-vivo rat heart were prospectively under-sampled with accelerations of two to five using a novel k-space sampling scheme. Results indicate that this approach is able to reconstruct DTI with minimal compromise to image quality. To the authors’ knowledge, this is the first study using compressed sensing to reconstruct prospectively under-sampled cardiac DTI.

Matthew Goette^1,2, Shelton Caruthers¹, Gregory Lanza¹, and Samuel Wickline^1
1Cardiology, Washington University in St. Louis, St. Louis, MO, United States, ²Pediatric Radiology, Texas Children's Hospital, Houston, TX, United States

This study presents a strategy to more accurately quantify the sparse ¹⁹F signal from ₃integrin targeted perfluorocarbon nanoparticle emulsions with a ¹H image-based actual flip angle B₁-mapping correction to the ¹⁹F and ¹H images, acquired with a simultaneous dual-frequency ultra-short echo time balanced steady state free precession sequence, in a phantom and an in vivo setting using a VX2 tumor model implanted into rabbits.

Ina Vernikouskaya¹, Alexander Pochert², and Volker Rasche^1
1Internal Medicine II, University Hospital of Ulm, Ulm, Baden-Wuerttemberg, Germany, ²Inorganic Chemistry II, University of Ulm, Ulm, Baden-Wuerttemberg, Germany

In contrast to conventional MR contrast agents, 19F signal can be directly measured, instead of indirect measuring the impact on the surrounding environment. However inhomogeneous profile of the excitation field can lead to wrong quantitative results. Therefore B1 mapping has to be performed prior to quantification. Adaptation of one of the existing B1 technique to the problem of 19F signal quantification is proposed in this work. In vitro results are in a good correlation with the simulation prediction. Estimation error for the relative quantification after B1 correction is within 6-7%.

Mai Le¹ and Jeffrey A. Fessler^1
1University of Michigan, Ann Arbor, MI, United States

This work explores a fast convolution-based temporal basis for DCE-MRI image reconstruction.

Yihang Zhou¹, Chao Shi¹, Yanhua Wang¹, Jingyuan Lyu¹, 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

In this study, a novel reconstruction method using kernel-based manifold learning and regularized pre-imaging is proposed to accelerate the MR parameter mapping. The parametric-weighted image at a specific time point is assumed to lie in a low-dimensional manifold and is reconstructed individually. The low-dimensional manifold is learned from the training images generated by the parametric model. The underlying optimization problem is solved using kernel trick and split Bregman iteration algorithm. Our preliminary result demonstrated that the proposed method is able to accurately recover the T2 map at high reduction factors when the conventional compressed sensing methods with linear models fail.

Dongwook Lee¹, Sunghong Park¹, Chuan Huang², Eung Yeop Kim³, and Jong Chul Ye^1
1KAIST, Daejeon, Daejeon, Korea, ²Harvard Medical School, Boston, United States, ³Department of Radiology, Gachon University Gil Hospital, Incheon, Korea

Purpose of this study is in vivo acceleration of T2 parameter mapping. To achieve this purpose, random undersampled mask is applied on the pulse sequence of 2D multi-echo spin echo. The images are reconstructed using compressed sensing algorithm with patch based low-rank penalty. And common mono-exponential fitting is used to generate T2 map. The acceleration times are remarkably reduced.

Daeun Kim¹, Joong Hee Kim², and Justin P. Haldar^1
1Department of Electrical Engineering, University of Southern California, Los Angeles, CA, United States, ²Department of Neurology, Washington University, St. Louis, MO, United States

This work proposes a novel data-driven method for automatically decomposing a multi-contrast MRI dataset into a mixture of constituent spatially-overlapping tissue components. The approach is non-parametric (no physical models are necessary), instead relying on a combination of low-rank matrix modeling, sparsity, and nonnegativity constraints through the nonnegative matrix factorization (NMF) framework. We demonstrate that NMF, when combined with an appropriate non-central chi noise model, can be used to automatically decompose diffusion and relaxation MRI datasets, yielding partial volume maps of white matter, gray matter, cerebrospinal fluid, and abnormal/injured tissue components.

Jinseong Jang¹, Taejoon Eo¹, and Dosik Hwang^1
1Electrical and Electronic Engineering, Yonsei University, Seoul, Korea

Model-based compressed sensing for multi-contrast imaging using weighted data consistency coefficient.

Kourosh Jafari-Khouzani¹, Berkin Bilgic¹, Jayashree Kalpathy-Cramer¹, and Kawin Setsompop^1
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States

This abstract proposes a non-local means technique to reconstruct partially sampled images in MRI compressed sensing. Instead of imposing total variation constraint, we use a fully-sampled contrast as a prior estimate to reconstruct other undersampled contrasts. Partial volume information is extracted from the prior estimate by a feature-based non-local means approach and then applied as constraint to the undersampled images. Experiments show that the proposed method is comparable to M-FOCUSS with prior estimate in terms of normalized root-mean-square (NRMSE) error while being up to 30× faster. It also attains 50% NRMSE reduction and 20× speed-up relative to the sparseMRI algorithm.

Rohini Vidya Shankar¹ and Vikram D Kodibagkar^1
1Biomedical Engineering, Arizona State University, Tempe, AZ, United States

Tissue oximetry studies using MRI can play a key role in the diagnosis, treatment, and monitoring of cancer. PISTOL is a novel oximetry technique that maps the T1 of administered HMDSO (1H reporter molecule) to obtain the tissue oxygen tension (pO2) at different locations. The aim of this study was to accelerate PISTOL acquisitions by developing a HMDSO-selective Look-Locker sequence with EPI readout. The new oximetry sequence, PISTOL-LL speeds-up 1H MR oximetry by 4X, enabling rapid pO2 mapping in under one minute. Results from in vivo studies demonstrate the successful application of the new technique in fast MR oximetry.

R. Marc Lebel^1,2, Yi Guo³, Yinghua Zhu³, Sajan Goud Lingala³, Richard Frayne², Linda B Andersen², Jacob Easaw⁴, and Krishna S Nayak^3
1GE Healthcare, Calgary, Alberta, Canada, ²Radiology, University of Calgary, Calgary, Alberta, Canada, ³Electrical Engineering, University of Southern California, Los Angeles, California, United States, ⁴Oncology, University of Calgary, Calgary, Alberta, Canada

We describe the use of sparse data sampling and constrained iterative reconstruction to provide a comprehensive contrast-enhanced brain exam. Data acquisition is performed with a 3D Cartesian radial trajectory that is amenable to retrospective definition of the temporal resolution. Multiple image reconstructions are performed to obtain all salient pieces of information desired from a contrast enhanced brain exam. We demonstrate that pre- and post-contrast anatomical T1-weighted images, dynamic angiography, quantitative permeability mapping, and quantitative perfusion mapping can all be achieved simultaneously - with high spatial resolution and full brain coverage.

Nikolaos Dikaios¹, Shonit Punwani², and David Atkinson^2
1Centre of Medical Imaging, UCL, London, United Kingdom, ²Centre of Medical Imaging, UCL, Greater London, United Kingdom

Direct parametric reconstruction (DPR), offers a new perspective in MR, setting the model parameters as the aim of reconstruction by estimating them directly from k-space using a Bayesian inference algorithm. DPR was implemented to derive model parameters (i.e. plasma volume vp, extracellular extravascular volume (EES) ve, transfer rate between plasma and EES (min-1) Ktrans) from dynamic contrast enhanced (DCE) (k,t)-space data. Its performance was evaluated against the current “indirect” approach where (k,t)-space DCE data are reconstructed (either with a Fourier Transform or with kt-FOCUSS when undersampling was present) to images and then fitted using a pharmacokinetic (PK) model2. The purpose of this work is to address some of the limitations of the DPR algorithm, namely the suggested modifications are to jointly reconstruct proton density, and native T1 map, T10 from multi-flip angle and DCE data along with the PK parameters. Further, DPR was implemented for different PK models so as the enhancement at each pixel (“tissue”) is described by the appropriate PK model.

Kinam Kwon¹, Dongchan Kim¹, Hyunseok Seo¹, Jaejin Cho¹, and Hyunwook Park^1
1KAIST, Guseong-dong, Daejeon, Korea

Clinical diagnosis requires several examinations to present various characteristics of organs, which are very time-consuming. To reduce total imaging time, many techniques have been proposed. Among them, parallel imaging techniques utilize sensitivity difference between multichannel RF coils. However, it is difficult to apply these techniques to higher acceleration due to SNR degradation. In this study, it is a key concept that each image in clinical protocols has different contrast, but shares similar structure information, and they are helpful for reconstructing each other. We propose a reconstruction model based on artificial neural network to allow to use higher acceleration factors.

Mei-Lan Chu^1,2, Hing-Chiu Chang¹, Koichi Oshio³, and Nan-kuei Chen^1
1Brain Imaging and Analysis Center, Duke University Medical Center, Durham, North Carolina, United States, ²Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan, ³Department of Diagnostic Radiology, Keio University School of Medicine, Japan

A novel procedure is developed to produce high-quality single-shot and multi-shot GRASE images without aliasing artifacts, in which coil sensitivity profiles and inter-CPMG-echo signal variation models are used as the constraints in reconstruction. With the developed POCSMUSE method, multi-contrast images can be reconstructed from a single set of GRASE data, reliably enabling parametric T2 mapping.

Chandan Kumar Aladahalli¹, Dattesh D Shanbhag², Venkata Veerendranadh Chebrolu², Patrice Hervo³, Sandeep N Gupta⁴, and Rakesh Mullick^5
1Biomedical Signal Analysis Laboratory, GE Global Research, Bangalore, Karnataka, India, ²Medical Image Analysis Laboratory, GE Global Research, Bangalore, Karnataka, India, ³GEHC, Buc, France, ⁴Clinical Systems and Signal Processing, GE Global Research, Niskayuna, NY, India, ⁵Diagnostics & Biomedical Technologies, GE Global Research, Bangalore, Karnataka, India

A web-enabled tool to allow generic collaborative exploration of multi-parameter-multi-time-point data is conceptualized and completed components of this tools are presented. The tool also allows clinicians to quickly apply existing biomarkers definitions using prior data and offers pre-set visualization templates for broad based diagnostics and therapeutic evaluation of longitudinal studies.

Itthi Chatnuntawech¹, Berkin Bilgic², Adrian Martin^1,3, Kawin Setsompop^2,4, and Elfar Adalsteinsson^1,5
1MIT, Cambridge, MA, United States, ²A. A. Martinos Center for Biomedical Imaging, MA, United States, ³Universidad Rey Juan Carlos, Mostoles, Madrid, Spain,⁴Harvard Medical School, MA, United States, ⁵Harvard-MIT Heath Sciences and Technology, MA, United States

We present an efficient algorithm to jointly reconstruct a set of images with different contrasts that has faster reconstruction time and better quality as measured by the normalized root-mean-square error (RMSE). To efficiently solve the _2,1-regularized optimization problem, our proposed algorithm first adopts the Split-Bregman (SB) technique to break down the problem into sub-problems. We efficiently compute a closed-form solution to each of the sub-problems with the help of a finite difference operator in k-space. The proposed algorithm (SB-L21) offers up to 32x faster reconstruction with up to 30% reduction in an average RMSE of the reconstructed images across all contrasts and slices, compared to other methods, including M-FOCUSS and SparseMRI.

Alexey Samsonov^1
1University of Wisconsin, Madison, Wisconsin, United States

Model-based reconstruction of undersampled data is a popular strategy to accelerate MR parameter mapping. In practice, however, the model-based strategy may lead to sub-optimal performance because actual signal may deviate from the model in many voxels (e.g., due to modeling simplifications, partial voluming, motion, etc). In this work, we propose a new model-based reconstruction technique, whose intrinsic insensitivity to the model mismatch in such voxels results in improved reconstruction of MR parameter maps.

Alessandro Sbrizzi¹, Annette van der Toorn¹, Hans Hoogduin¹, Peter R Luijten¹, and Cornelis A van den Berg^1
1UMC Utrecht, Utrecht, Utrecht, Netherlands

We present a new approach of quickly measuring MR parameter maps by treating the quantitative MR problem as a dynamic system identification process. The system equations are inverted to match the response of the MR scanner to the data in time domain. Due to advances in numerical optimization and computing power, this approach has become possible and it is routinely applied, for instance, to seismology. For the first time, we apply it to MR and we recover all the desired parameters, for example: T1, T2, B1, B0, M0. Data acquisition takes just a few seconds.

Volkert Roeloffs¹, Xiaoqing Wang¹, Tilman Sumpf¹, and Jens Frahm^1
1Biomedizinische NMR Forschungs GmbH, Max Planck Institute for Biophysical Chemistry, Göttingen, Niedersachsen, Germany

In this work we present a model-based reconstruction algorithm for the reconstruction of high resolution T1 maps from a fast spoiler-gradient free radial IR-FLASH sequence. By using minimal TR and formulating the problem in the parameter domain, no further regularization techniques are necessary. The resulting non-linear optimization problem is solved by the Gauss-Newton method. A phantom study revealed good agreement of the T1 values with the gold standard and a first in vivo experiment using the same parameter set as in the phantom study proofed the proposed technique to be robust.

Artifacts & Correction I

Tuesday 2 June 2015

Sungmin Cho¹, Joonsung Lee², Jaewook Shin¹, Min-Oh Kim¹, and Dong-Hyun Kim^1
1Yonsei University, SeodaemunGu, Seoul, Korea, ²Severance Hospital, Seoul, Korea

Magnetic resonance electrical properties tomography (MREPT) is a technique which estimates conductivity and permittivity by measuring B1 information. Previous MREPT assumed that electrical properties are locally homogeneous so this assumption incurred the "Boundary Artifact". Lee¡¯s proposed method reduces boundary artifact by iterative process. However, this method was verified just for simulation data with high SNR condition. To verify the practical applicability of this method, in this study, the iterative boundary artifact reduction algorithm is implemented for a phantom experiment.

Moran Wei¹, Weiwei Zhang¹, Yongchuan Lai¹, and Bing Wu^1
1GE Healthcare, Beijing, Beijing, China

In hybrid 3D FSE, the close proximity of pre-phaser and the first refocus pulse makes the first refocus RF very vulnerable to the short-term eddy current induced by the pre-phaser. This causes image shadings in regions distant from the magnet center. We propose to shift the position of pre-phaser to post of the first refocus RF to eliminate the image shading of hybrid 3D FSE sequence.

Candice A. Bookwalter¹, Samir D. Sharma¹, and Scott B. Reeder^1,2
1Department of Radiology, University of Wisconsin-Madison, Madison, WI, United States, ²Department of Medical Physics, University of Wisconsin-Madison, Madison, Wisconsin, United States

A commonly seen “bite mark” susceptibility artifact in the inferolateral and anteroseptal myocardial wall in cardiac cine SSFP images can degrade morphologic and functional evaluation, particularly for R2* mapping. Two explanations have been previously proposed, that these artifacts originate from 1) the heart-lung interface or 2) deoxygenated blood in adjacent cardiac veins. The purpose of this study was to determine whether the geometry of the heart-lung interface alone accounts for "bite mark" artifacts in vivo.

Ming Yang¹, Haikun Qi², Shuo Zhang³, Guang Qiang Geng¹, Chen Guang Zhao¹, Huijun Chen², and Feng Huang^1
1Philips Healthcare, Suzhou, Jiangsu, China, ²Center for Biomedical Imaging Research, Tsinghua University, Beijing, China, ³Philips Healthcare, Singapore, Singapore

This abstracts introduces a method to correct the off-center issue caused by gradient delays in radial arbitrary-angle acquisition. Instead of using 0º-180º and 90º-270º projection pairs, this method only requires a number of pairs of pseudo anti-parallel projections and then applies Linear regression method such as least square method to estimate the gradient delays. Therefore, the off-center error can be compensated during the reconstruction for arbitrary-angle, such as golden-angle radial acquisition. This method is tested on phantom and human Golden-angle radial MR images and result shows the image quality is improved with off-center error corrected using this method.

Stephen James Wastling¹, Mark Symms², Mauro Costagli^3,4, Laura Biagi^3,4, Mirco Cosottini^3,5, Gareth John Barker¹, and Michela Tosetti^3,4
1Department of Neuroimaging, King's College London, London, United Kingdom, ²GE Healthcare, Pisa, Italy, ³Imago7, Pisa, Italy, ⁴IRCCS Stella Maris, Pisa, Italy, ⁵Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Pisa, Italy

A method of reducing signal drop-out in Gradient Echo imaging using a Hyperbolic Secant pulse was developed and tested at 7T. The coronal images showed reduced signal drop-out in the brain-stem, inferior temporal lobe, and the cerebellum, producing images of superior visual quality in human subjects.

Shengzhen Tao¹, Joshua D Trzasko¹, Yunhong Shu¹, John Huston III¹, Paul T Weavers¹, and Matt A Bernstein^1
1Radiology, Mayo Clinic, Rochester, MN, United States

Due to engineering limitations, achieving perfect gradient linearity across the imaging field-of-view is infeasible. Gradient nonlinearity(GNL), if not accounted for, causes image geometrical distortion, which is conventionally corrected by image-domain interpolation. Direct interpolation techniques, however, exert smoothing effects on corrected images which results in resolution loss. In non-Cartesian MRI, B0 inhomogeneity can also cause image blurring. In this work, a non-iterative gridding reconstruction framework with integrated GNL and B0 off-resonance correction is developed for non-Cartesian MRI. The proposed strategy can mitigate the image blurring that occurs in standard interpolation-based GNL-correction and from B0 inhomogeneity while still effectively correcting geometrical distortion.

Shengzhen Tao¹, Joshua D Trzasko¹, Paul T Weavers¹, Yunhong Shu¹, John Huston III¹, and Matt A Bernstein^1
1Radiology, Mayo Clinic, Rochester, MN, United States

Partial Fourier homodyne reconstruction is a widely used method to reduce the amount of data required to form an image in MRI by up to 50%. Recently, a non-iterative MR image reconstruction method with integrated gradient nonlinearity (GNL) correction was proposed. The method was shown to be able to mitigate the image blurring and resolution loss introduced by conventional, image-domain interpolation based GNL correction, while still correcting the GNL-induced image geometrical distortion. In this work, we discuss the addition of partial Fourier acquisition to this integrated reconstruction paradigm to allow for similar maintenance of spatial resolution while reducing acquisition time.

Henrik Marschner¹, André Pampel¹, and Harald E. Möller^1
1Nuclear Magnetic Resonance, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Sachsen, Germany

We investigate a novel averaging approach for MRI series that does not rely on repetitions of an image. The principle is based on complex averaging but translated into the wavelet space where conservation of signals is possible that are not meant to be averaged out. The novel averaging is applied for a qMTI experiment which consists of a series of images with different contrasts. The observed effect of the averaging for the investigated post-mortem brain is similar to twofold averaging but without the need for repetitions of any image.

Kevin Perkins^1,2, Reeve Ingle², Juan Santos², Galen Reed², Ken Johnson², and William Overall^2
1BYU, Provo, Utah, United States, ²HeartVista, Menlo Park, Ca, United States

One source of artifacts in spiral scans is concomitant gradient fields (CGFs), which are higher order field terms that accompany linear gradients. Demodulation is a zero-order CGF correction method that has been implemented in real-time. However, this simple approach does not effectively correct CGF effects in non-axial slices. Multi-frequency binning is a more advanced technique that requires more computational time. We propose an intermediate, first-order solution that retains the non-axial correction capabilities of the multifrequency approach along with the speed of simple demodulation correction. This first-order approach is shown to eliminate blurring artifact inside a user-defined FOV.

Hing-Chiu Chang¹ and Nan-Kuei Chen^1
1Brain Imaging and Analysis Center, Duke University Medical Center, Durham, North Carolina, United States

In addition to minuscule motion induced aliasing artifacts, interleaved EPI based DWI are also susceptible to Nyquist artifact. 1D Nyquist correction is routinely used for single-shot and interleaved DWI. However, in many cases (e.g., oblique-plane EPI), the Nyquist artifacts due to phase errors along the phase-encoding direction are significant, which can only be suppressed with The 2D phase correction Nyquist correction methods for Nyquist ghost reduction reveal significant improvement of image quality for both single-shot and segmented EPI data acquired in oblique plane or in the presence of cross-term eddy current. Unfortunately, existing 2D Nyquist correction procedures are not compatible with interleaved DWI data. To address this limitation, here we report a novel integrated approach to simultaneously remove 2D Nyquist artifact and shot-to-shot phase variations in interleaved DWI data.

Francesco Santini¹, Carl Ganter², Philipp Ehses³, Klaus Scheffler³, and Oliver Bieri^1
1Radiological Physics, University of Basel Hospital, Basel, Switzerland, ²Department of Diagnostic Radiology, Klinikum rechts der Isar, Munich, Germany, ³Max Planck Institute for Biological Cybernetics, Tübingen, Germany

Phase imaging is especially important at high to ultra high fields to both map local field inhomogeneities and highlight important anatomical structures. Obtaining a proper coherent phase image from the combination of multiple receiver coils, however, is challenging. Here, we introduce a generic referenceless phase combination approach able to eliminate most artifacts in phase imaging without the need of a special acquisition protocol or geometric assumptions for the coils.

Shayan Guhaniyogi¹, Mei-Lan Chu¹, and Nan-Kuei Chen^1
1Brain Imaging and Analysis Center, Duke University, Durham, NC, United States

A major concern of multishot MRI acquisitions is the effect of subject motion, which can result in undesirable image artifacts. In order to discard or correct these images, the first step is to identify the images which have been corrupted. We describe an automated machine-learning method to identify motion-corrupted multishot images using unsupervised feature learning and a convolutional neural network. We demonstrate that the method can accurately classify motion-corrupted images of different contrasts and different multishot acquisition types. The result is an effective technique which eliminates the need for manual identification of motion artifacts in multishot images.

Bo Zhang¹, Hans Peeters², Ad Moerland², Helene Langet¹, and Niccolo Stefani^3
1Philips Research, Suresnes, France, ²Philips Healthcare, Netherlands, ³Philips Healthcare, OH, United States

MR images are usually degenerated by artifact of intensity inhomogeneity, or bias field, undesirable for perception and diagnosis. In this work, we present an optimized 3-dimensional retrospective nonparametric inhomogeneity correction method by minimizing a regularized-entropy criterion. The inhomogeneity estimator is numerically particularly efficient, scalable and parallelizable compared to exisiting entropy-based approaches. Its effectiveness and robustness have also been validated by vast clinical evaluations on 1.5T and 3T scans of brain and breast applications.

Maysam Mahmood Jafar¹, Christopher Dean², Malcolm J Birch¹, and Marc E Miquel^1
1Medical Physics, Barts Health NHS Trust, London, London, United Kingdom, ²Radiotherapy, Barts Health NHS Trust, London, London, United Kingdom

Radiotherapy treatment planning (RTTP) necessitates accurate delineation of tumor volume and adjacent structures at risk. Magnetic resonance imaging (MRI) offers superior soft-tissue contrast compared with CT but suffers from inherent geometric distortions. The problem is exacerbated at higher field strengths where there is an increase in inhomogeneties in the main B0 magnetic field as well as the B1 RF field. Furthermore, non-linearities in the applied gradients add a further element to these distortions. In this study, we propose a cost-effective regularly structured three-dimensional 3D printed grid phantom, which enables one to quantify machine-related MR distortion by comparing the locations of corresponding features in both MR and CT data sets.

Ameneh Boroomand¹, Mohammad Javad Shafiee,¹, Alexander Wong¹, Farzad Khalvati², Paul Fieguth¹, and Masoom Haider^3
1System Design Engineering, University of Waterloo, Waterloo, Ontario, Canada, ²Medical Imaging, University of Toronto, Toronto, Ontario, Canada,³Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada

The bias field inhomogeneity in Magnetic Resonance Imaging (MRI) often makes difficulties for the physicians who interpret and analyze the MR images. One important challenging aspect of the most bias field correction methods is the presence of MRI noise which should be handled. Here, we propose a Bayesian based image reconstruction framework which concurrently corrects for the MRI bias field as well as compensates for MRI noise in the final reconstructed MR image.

Alto Stemmer¹ and Berthold Kiefer^1
1Healthcare, Siemens AG, Erlangen, Germany

The acquisition of a field map is integrated into a diffusion-weighted single shot EPI prototype sequence. The field information retrieved from the unwrapped and calibrated field map is used at run time for slice-specific center frequency update and gradient offset shimming and during the reconstruction for pixel-wise unwarping of remaining distortions.

Se-Hong Oh¹ and Mark J Lowe^1
1Imaging Institute, Cleveland Clinic Foundation, Cleveland, OH, United States

DWI sequences based on single-shot EPI have an additional source of N/2 ghosting artifacts that are associated with B0 field perturbations resulting from diffusion gradient-induced eddy currents. Hence, diffusion gradient-induced phase error must be considered in DWI. In this study, we investigate the impact of navigator echo acquisition locations. In addition, the effect of a dummy diffusion gradient is investigated as an alternative method reduces eddy current.

Brian Burns¹, Neil Wilson², and M. Albert Thomas^2,3
1Department of Bioengineering, UCLA, Los Angeles, CA, United States, ²Medical Physics, IDP, UCLA, Los Angeles, CA, United States, ³Department of Radiology, UCLA, Los Angeles, CA, United States

Current phase correction techniques in multi-dimensional spectroscopic imaging (MRSI) do not take into account the spatio-temporal nature of phase errors because they were designed for single voxel methods. This work categorizes phase errors in 4D MRSI (2D spatial+2D spectral) as time, space, or space and time varying, and proposes a post-processing pipeline that decouples these errors so they are removed in the appropriate domain. The Interleaved Navigator Scan corrected Echo-Planar J-Resolved Spectroscopic Imaging (INSEP-JRESI) sequence is proposed. Results from gray matter phantom scans using this new sequence and pipeline demonstrate the viability of this technique compared to MRSI-adapted Klose's methods.

Daiki Tamada¹ and Katsumi Kose^1
1Institute of Applied Physics, University of Tsukuba, Tsukuba, Ibaraki, Japan

To design incoherent trajectories is important to achieve an efficient compressed sensing (CS) reconstruction. However, in practice, CS reconstruction using non-Cartesian sampling trajectories suffer from the off-resonance effect, which makes image distortion and artifacts. To overcome this problem, we developed a new CS reconstruction strategy using a cross-sampling, and a time-segmented off-resonance correction method. A B₀ distribution map used for the off-resonance correction was estimated using an image registration based method. And a wavelet regularized split Bregman method was used for the reconstruction. Imaging experiments of a chemically fixed mouse demonstrated usefulness of our method.

Cheng Li¹, Jeremy F. Magland¹, Alan C. Seifert¹, and Felix W. Wehrli^1
1Laboratory for Structural NMR Imaging, Department of Radiology, University of Pennsylvania, Philadelphia, PA, United States

Zero echo-time (ZTE) sequence is a promising technique for short-T2 imaging. However, the presence of an imaging gradient during excitation causes blurring and shadow artifacts due to limited RF pulse bandwidth. Our previous work proposed an approach by applying phase-modulated RF excitation and iterative reconstruction to correct the artifacts. However, this iterative method is computationally intensive. In this work, we developed a fast non-iterative correction algorithm to dramatically reduce the computation time while maintaining image quality. Results from phantom and in vivo scans demonstrate the effectiveness of the method. The proposed method allows online ZTE reconstruction with artifact correction on clinical scanners.

Xinwei Shi¹, Daehyun Yoon², Kevin Koch³, and Brian Hargreaves^2
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, CA, United States, ³Radiology, Medical College of Wisconsin, WI, United States

3D Multi-Spectral Imaging techniques have made significant advancements toward imaging near metal, with their ability to correct for most of the distortion and signal loss. However, in close vicinity of implants, artificial signal voids and dark tissues are indistinguishable with lack of signal in the location of the actual implant, and this makes it challenging to visualize the implant geometry or to examine the tissue/implant interface. In this work, we demonstrate a regularized inversion approach to estimate the susceptibility map from the B0 field maps and thereby differentiate the metallic voxels from artificial signal void or dark tissues.

Joshua Trzasko¹, Shengzhen Tao¹, Jeffrey Gunter¹, Yunhong Shu¹, John Huston III¹, and Matt Bernstein^1
1Mayo Clinic, Rochester, MN, United States

Gradient nonlinearity (GNL) correction is a standard process performed on MRI scanners to eliminate geometric spatial distortion that arises from imperfect hardware performance. Typically, the gradient field is estimated via electromagnetic (EM) simulation for a scanner type, but does not account for scanner-specific variations due to hardware construction (e.g., winding) or siting. Recently, a phantom-based calibration procedure was developed that enables accurate individual field estimation without needing proprietary information. In this work, we develop a new iterative estimation strategy — based on post-GNL correction distortion mean square error (MSE) minimization — that further improves scanner-specific gradient field estimation accuracy.

Paul Polak¹, Robert Zivadinov^1,2, and Ferdinand Schweser^1,2
1Department of Neurology, Buffalo Neuroimaging Analysis Center, State University of New York at Buffalo, Buffalo, NY, United States, ²Molecular and Translational Imaging Center, MRI Center, Clinical and Translational Research Center, Buffalo, NY, United States

Images reconstructed by direct Fourier transform from k-space data are hindered by gradient non-linearities which result in imaging voxel distortions. Correction of these effects, or gradient unwarping, is provided by MR manufacturers near the end of their image reconstruction pipeline; however, this is typically applied only to multi-channel combined magnitude images. Advanced reconstruction techniques utilizing compressed sensing, non-Cartesian sampling or multi-channel phase images typically use data from a more primary step (i.e. k-space or single channel data), and are thus subject to gradient warping effects in the final reconstruction. We present here a technique to unwarp complex-valued MRI data which is then suitable for advanced phase imaging reconstruction.

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

To perform correction of axis dependent system/gradient delays for radial imaging we propose an improved algorithm which works by iteratively optimizing the recon-structed image until the local minima of a calibration function is reached. The calibration function is calculated from the reconstructed image magnitude and thus uses the actually measured radial raw data for correction, requiring no additional calibration data to be acquired. We demonstrate the performance of the algorithm for both phantom and in vivo measurements.

Image Processing & Segmentation

Tuesday 2 June 2015

Jyh-Miin Lin¹, Andrew Patterson², Hing-Chiu Chang³, Tzu-Chao Chuang⁴, Hsiao-Wen Chung⁵, Jonathan H. Gillard¹, and Martin J. Graves^2
1Department of Radiolgoy, University of Cambridge, Cambridge, Cambridgeshire, United Kingdom, ²Cambridge University Hospitals NHS Foundation Trust, Cambridge, United Kingdom, ³Brain Imaging and Analysis Center, Duke University Medical Center, NC, United States, ⁴Department of Electrical Engineering, National Sun Yat-sen University, Kaohsiung, Taiwan, Taiwan, ⁵Department of Electrical Engineering, National Taiwan University, Taiwan, Taiwan

The colored noise pattern in periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) images is described theoretically by Cramér-Rao lower bound (CRLB), followed by confirmation using simulation and phantom studies. An iterative regularized method named Pseudo-Inverse as COnstraint (PICO) for reconstructing PROPELLER images is proposed and tested on phantom images to examine the whitening of noise power spectra at various angular under-sampling factors. Comparison against conventionally reconstructed PROPELLER images using density compensation demonstrates the advantages of PICO by reducing streaks artifacts and high-spatial-frequency noise on human images in vivo.

Vanessa Wiggermann^1,2, Enedino Hernández Torres^2,3, Anthony Traboulsee^3,4, David K.B. Li^2,4, and Alexander Rauscher^2,3
1Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, ²Radiology, University of British Columbia, Vancouver, BC, Canada, ³UBC MRI Research Centre, Vancouver, BC, Canada, ⁴Medicine (Neurology), University of British Columbia, Vancouver, BC, Canada

Visualization of cortical lesions on conventional MR images is often challenging due to restricted contrast-to-noise levels. Further, cerebrospinal fluid (CSF) in the vicinity of lesions hampers lesion detection. The here proposed combination of conventional T2-weighted and FLAIR images doubles the contrast-to-noise ratio, while providing suppression of CSF signal. The enhanced contrast may aid automated lesion segmentation and detection of cortical lesions.

Jisu Hu #¹, Wenbo Wu #², Bin Zhu #², Huiting Wang², Renyuan Liu², Xin Zhang², Ming Li², Yongbo Yang³, Jing Yan⁴, Fengnan Niu⁵, Chuanshuai Tian², Kun Wang², Haiping Yu², Weibo Chen⁶, Suiren Wan*¹, Yu Sun*¹, and Bing Zhang*^2
1The Laboratory for Medical Electronics, School of Biological Sciences and Medical Engineering, Southeast University, Nanjing, China, ²Department of Radiology, The Affiliated Drum Tower Hospital of Nanjing University Medical School, Nanjing, China, ³Department of Neurosurgery, The Affiliated Drum Tower Hospital of Nanjing University Medical School, Nanjing, China, ⁴Department of Oncology, The Affiliated Drum Tower Hospital of Nanjing University Medical School, Nanjing, China, ⁵Department of Pathology, The Affiliated Drum Tower Hospital of Nanjing University Medical School, Nanjing, China, ⁶Philips Healthcare, Shanghai, China

In order to combine multiple modalities of MRI in preoperative cerebral glioma grading, a diagnosing tool based on Bayesian Network was developed to integrate features extracted from conventional MR imaging, perfusion weighted imaging and MR spectroscopic imaging. The structure of the network was determined in cooperation with experienced neuroradiologists and the parameters learned using EM (Expectation-Maximization) algorithm with the incomplete dataset of 52 clinical cases. The grading performance was evaluated in a leave-one-out analysis, achieving the highest grading accuracy of 88.24% with all the features observed.

Eric Van Reeth¹, Michael Sdika¹, Sophie Gaillard¹, Pierre-Hervé Luppi², Paul-Antoine Libourel², and Olivier Beuf^1
1Université de Lyon, CREATIS; CNRS UMR5220; Inserm U1044; INSA-Lyon; Université Lyon 1, Villeurbanne, Rhone, France, ²Centre de Recherche en Neurosciences de Lyon; Inserm U1028 - CNRS UMR5292, Lyon, Rhone, France

An image processing method called super-resolution is applied on rat brain MR data in order to improve the spatial resolution and signal-to-noise ratio of 2D multi-slices MR acquisitions. The high-resolution output image has an isotropic spatial resolution and is obtained from several low-resolution anisotropic images. It clearly improves the trade-off between acquisition time, SNR and spatial resolution when compared to a high-resolution image that was acquired in approximately the same time. The few changes required on the acquisition protocol and the flexibility of the imaging model to handle different applications make it an interesting post-processing method for practical applications.

Di Zhao^1
1The Dorothy M. Davis Heart & Lung Research Institute, The Ohio State University, Columbus, Ohio, United States

A generic problem of MRI images is the low SNR, and filtering is the widely used technique to suppress MRI image noise. Images filters based on machine learning algorithms, such as Support Vector Machine (SVR), have been shown to have superior performance because the signal can be preserved better. In this abstract, we apply Support Vector Regression based denoiser (SVR denoiser) for MRI image processing.

Dirk K. Müller¹, René Küttner¹, Ralf Hannig¹, Thomas Frank¹, Juliane Müller¹, and Michael Marxen^1
1Department of Psychiatry and Neuroimaging Center, Technische Universität Dresden, Dresden, 01187, Germany

We developed a pipeline manager to offer standardized and parallelized imaging processing pipelines to a community of applied scientists with an efficient and intuitive web-based interface. The software is based on nipype allowing to interface with algorithms from different software packages (e.g., FSL, FreeSurfer, SPM ...). Pipelines can be executed without the need of programming. Quality assurance for motion correction, coregistration and normalization is included.

Nicole Wake^1,2, William Huang³, Todd Pietila⁴, and Hersh Chandarana^1
1The Center for Advanced Imaging Innovation and Research (CAI2R), Department of Radiology, New York University School of Medicine, New York, New York, United States, ²The Sackler Institute of Graduate Biomedical Sciences, New York University School of Medicine, New York, New York, United States,³Department of Urology, New York University School of Medicine, New York, New York, United States, ⁴Materialise USA, Plymouth, Michigan, United States

Soft tissue three-dimensional (3D) printing from magnetic resonance (MR) imaging data is feasible, but implementation is challenging and time consuming. We performed 3D segmentation of an abdominal MR dataset and created a high-fidelity, life-sized 3D model of a kidney and in-situ renal tumor for pre-operative surgical planning. The 3D model provided additional information to the attending surgeon that influenced surgical planning and allowed real-time changes to be applied during the operative procedure.

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

Spatiotemporally encoded (SPEN) single-shot imaging is a recently proposed ultrafast approach, which has great advantages in resisting field inhomogeneity and chemical shift effects compared to echo planar imaging, but limited by its low inherent spatial resolution. Super-resolved (SR) reconstruction is indispensable to SPEN approach, which can improve the spatial resolution without additional acquisition. The existing SR algorithms always compromise in spatial resolution to suppress aliasing artifacts. In this abstract, we proposed a novel SR algorithm termed super-resolved enhancing and edge deghosting, which can provide better spatial resolution compared to state-of-the-art SR reconstruction algorithms in SPEN MRI.

Angel Torrado-Carvajal^1,2, Eduardo Alcain³, Joaquin L. Herraiz^2,4, Antonio S. Montemayor³, Juan A. Hernandez-Tamames^1,2, Elfar Adalsteinsson^5,6, Larry L. Wald^6,7, and Norberto Malpica^1,2
1Medical Image Analysis and Biometry Lab, Universidad Rey Juan Carlos, Mostoles, Madrid, Spain, ²Madrid-MIT M+Vision Consortium, Madrid, Spain, ³Dept. of Computer Science, Universidad Rey Juan Carlos, Mostoles, Madrid, Spain, ⁴Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA, United States, ⁵Dept. of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States,⁶Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, ⁷Martinos Center for Biomedical Imaging, Dept. of Radiology, MGH, Charlestown, MA, United States

In this work, we propose a fast pseudo-CT generation from a patient-specific MRI T1-weighted image using a group-wise patch-based approach and a limited MRI and CT atlas dictionary. The use of patch-based techniques to estimate a pseudo-CT from MR T1-weighted images allows determining accurate AC maps for use in hybrid PET/MR systems. The proposed method provides an accurate estimation of the pseudo-CT with a similar accuracy as patient-specific CT does. This avoids the over-simplification of most previous proposed methods based on segmented MR images that assume that all voxels in the same tissue type should have the same attenuation coefficients.

Jason Su^1,2, Thomas Tourdias³, Manojkumar Saranathan², and Brian K. Rutt^2
1Electrical Engineering, Stanford University, Stanford, California, United States, ²Radiology, Stanford University, Stanford, California, United States,³Neuroradiology, Bordeaux University Hospital, Bordeaux, France

A method for automatic segmentation of thalamic nuclei was developed and optimized using 7T white-matter-nulled MP-RAGE images, which provide excellent contrast and detail for segmentation and for ANTS nonlinear registration. The PICSL multi-atlas label fusion algorithm by Wang and Yushkevich was optimized for 12 thalamic nuclei and validated in 9 subjects using an atlas of prior manual delineations from 20 subjects, including multiple sclerosis patients and healthy controls. Performance in accuracy, resolution, and acquisition time surpasses other published methods that require DTI. The Dice coefficients for whole thalamus (0.92), pulvinar nucleus (0.86), and mediodorsal nucleus (0.87) were notably high.

Daniel S. Cho¹, Farzad Khalvati², Alexander Wong¹, David A Clausi¹, and Masoom Haider^2
1Systems Design Engineering, University of Waterloo, Waterloo, Ontario, Canada, ²University of Toronto, Ontario, Canada

Diffusion weighted imaging (DWI) has gained significant attention for prostate cancer imaging as its derived modalities such as apparent diffusion coefficient and computed high b-value images are commonly employed for prostate cancer analysis. In this work, a novel technique to register a set of DWI acquisitions across multiple b-values was proposed. The proposed registration adapted b-spline registration with a new hybrid similarity metric, which utilized statistical likelihood and cross-correlation. The DWI co-registration showed the improved contrast-to-noise ratio on DWI acquisitions across multiple b-values as well as ADC map.

Yi He^1,2, Kun Zhang³, and Xin Yu^1,2
1Research Group of Translational Neuroimaging and Neural Control, High-Field Magnetic Resonance, Max Planck Institute for Biological Cybernetics, Tuebingen, Baden-Wuerttemberg, Germany, ²Graduate School of Neural Information Processing, University of Tuebingen, Tuebingen, Baden-Wuerttemberg, Germany,³Department of Empirical Inference, Max Planck Institute of Intelligent System, Tuebingen, Germany

The hemodynamic response function (HRF) of fMRI signal varies according to different conditions and tasks. The estimate of HRF highly relies on the spatiotemporal resolution of fMRI raw images. Here, we developed a new algorithm to estimate the single-vessel specific HRF from the fMRI signal acquired by multi-echo line-scanning method (MELS-fMRI). This method simultaneously performed T2* decay deconvolution and HRF optimization. Given the millisecond scale sampling rate for multi-echo acquisition, the estimated HRF bears high temporal feature of fMRI signal propagation through different venules in the deep layer cortex. This is the first step to model millisecond scale fMRI signal.

Elizabeth Wicks¹, Jason P.C. Chiu¹, Lisa Y.W. Tang^1,2, Kevin Lam¹, Andrew Riddehough¹, David K.B. Li^1,2, Anthony Traboulsee¹, and Roger Tam^1,2
1MS/MRI Research Group, Division of Neurology, University of British Columbia, Vancouver, BC, Canada, ²Dept. of Radiology, University of British Columbia, BC, Canada

Established methods of brain volume normalization on T1-weighted images typically require affine registration to a standard template, which can introduce a significant amount of measurement noise. We have developed a fully-automated method to compute a normalized brain volume from T1w images by directly estimating intradural volume. This method correlated highly (r = 0.845) with an established method on T2w/PDw images, when applied to a completed multiple sclerosis clinical trial dataset of 131 patients, which included both T1w and T2w/PDw sequences of each patient.

Alberto De Luca^1,2, Denis Peruzzo², Fabio Triulzi³, Filippo Arrigoni², and Alessandra Bertoldo^1
1Department of Information Engineering, University of Padova, Padova, PD, Italy, ²Department of Neuroimaging, Scientific Institute, IRCCS "Eugenio Medea", Bosisio Parini, LC, Italy, ³Neuroradiology department, Scientific Institute, IRCCS "Cà Granda" - Ospedale Maggiore Policlinico, Milan, MI, Italy

Malformations of cortical development (MCDs) encompass a wide spectrum of brain abnormalities which extension and localization are extremely variable from subject to subject and their analysis with existing methods is difficult. First we extended a fractal geometry algorithm to compute voxelwise maps, then defined two distance maps used to quantify the distance of a single subject from a population. Results suggest that fractal values are sensible to the structural properties of the tissues being statistically different values between healthy and malformed cortex. The classification based on these indices is able to reveal malformed areas with high specificity.

Bilal A Tahir^1,2, Helen Marshall², Matthew Q Hatton¹, Jim M Wild², and Rob H Ireland^1,2
1Academic Unit of Clinical Oncology, University of Sheffield, Sheffield, South Yorkshire, United Kingdom, ²Academic Unit of Academic Radiology, University of Sheffield, Sheffield, South Yorkshire, United Kingdom

Image registration of inspiratory and expiratory lung CT has been proposed to generate surrogates of ventilation. However, validation against established ventilation modalities is required prior to clinical implementation. Here, we present a method using deformable image registration, via same-breath 1H MRI, to enable ROI correlation analysis of ventilation calculated by dual breath-hold CT with hyperpolarized 3He MRI, which is demonstrated in 3 patients with moderate-to-severe asthma.

Till Huelnhagen¹, Andreas Pohlmann¹, and Thoralf Niendorf^1,2
1Berlin Ultrahigh Field Facility (B.U.F.F.), Max-Delbrueck Center for Molecular Medicine (MDC), Berlin, Germany, ²Experimental and Clinical Research Center, a joint cooperation between the Charite Medical Faculty and the Max-Delbrueck Center, Berlin, Germany

This works investigates the impact of spatially adaptive non local means filtering (SANLM) on T₂^* mapping accuracy using numerical simulations and in vivo data derived from animal and human imaging at ultrahigh magnetic fields. The presented results suggest, that SANLM filtering prior to T₂^* mapping can substantially improve T₂^* fitting accuracy, but should be used with due caution for very small structures and very low SNR. The in vivo results suggest that SANLM filtering provides means for improving parametric mapping for a broad range of applications including neurovasculaar and cardiac parametric mapping.

Parmida Moradi Birgani^1,2, Anahita Fathi Kazerooni^1,2, Hamidreza Haghighatkhah³, Pedram Fadavi⁴, Mohsen Shojaei Moghaddam⁵, Meghdad Ashtiyani⁶, and Hamidreza Saligheh Rad^1,2
1Quantitative MR Imaging and Spectroscopy Group, Research Center for Molecular and Cellular Imaging, Tehran University of Medical Sciences, Tehran, Iran,²Department of Medical Physics and Biomedical Engineering, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran, ³Department of Radiology, School of Medicine, Shahid Beheshti University of Medical Sciences, Tehran, Iran, ⁴Radiation Oncology Department, Iran University of Medical Sciences, Tehran, Iran, ⁵Imaging Center, Payambaran Hospital, Tehran, Iran, ⁶Department of Medical Physics and Biomedical Engineering, School of Medicine, International Campus, Tehran University of Medical Sciences, Tehran, Iran

Accurate assessment of bone marrow, as a common site of metastasis among patients with breast cancer, in DW-MR images is of high clinical importance. In this regard, bone marrow extraction could play an important role in providing early markers of tumor progression to be followed by proper therapy planning. However, the existing bone marrow segmentation methods are prone to errors due to heterogeneous nature of tumors, mis-registration and lack of reproducibility. This issue becomes even more challenging in extracting bone marrows from ADC-maps where there is a lack of morphological information. In this work, we proposed a novel texture-based segmentation approach, which could be reliably used for bone marrow extraction from ADC-maps.

Peter Adany¹, In-Young Choi^1,2, Erica Sherry¹, and Phil Lee^1,3
1Hoglund Brain Imaging Center, University of Kansas Medical Center, Kansas City, KS, United States, ²Neurology, University of Kansas Medical Center, Kansas City, KS, United States, ³Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, KS, United States

Despite the functional importance of the thalamus including the regulation of consciousness, sleep and alertness, its structural analysis is challenging, requiring significant manual segmentation with cumbersome editing of 3D shapes as 2D contours. The currently available interpolation tools (linear, sinc, trilinear, etc.) present an obstacle because they do not create smooth contours, i.e., the interpolation is based on the intensity rather than a shape of the target structure. Our proposed universal Shape Interpolation using the Radon Transform (uSHIRT) presents an advancement from currently available analysis methods, as it greatly improves and facilitates coregistration, reslicing and editing masks in different planes.

Max W.K. Law¹, Jing Yuan¹, Gladys G. Lo², Oi Lei Wong¹, Abby Y. Ding¹, and Siu Ki Yu^1
1Medical Physics and Research Department, Hong Kong Sanatorium & Hospital, Hong Kong, Hong Kong, ²Department of Diagnostic and Interventional Radiology, Hong Kong Sanatorium & Hospital, Hong Kong, Hong Kong

This work describes a new algorithm that automatically segments the cranium from two MRI sequences - gradient echo based "Blackbone" MRI and Ultra-short-TE "Silenz" MRI. This algorithm deforms an ellipsoid template according to the Hessian based image statistics, to find the boundaries where abrupt intensity changes are observed. We also studied the bone thickness consistency and bone signal contrast compared to the neighboring tissues for these two sequences. This method is potentially helpful for clinical applications such as MR-based cephalometry and radiotherapy planning to reduce or eliminate radiation deposition in patients.

Joel Kullberg¹, Lars Johansson¹, Lars Lind², Håkan Ahlström¹, and Robin Strand^1
1Radiology, Uppsala University, Uppsala, Sweden, ²Medical Sciences, Uppsala University, Uppsala, Sweden

We have developed an image processing concept, Imiomics (imaging –omics), a set of methods, including image registration, that allow statistical and holistic analysis of whole-body image data and non-imaging data. The image registration gives point-to-point correspondences between images allowing whole-body comparisons of image intensity values and morphology. The purpose of this work was to present Imiomics and initial examples of cross sectional interaction between MRI (fat content and local volume) and non-imaging data (anthropometrical and body fat measurements) where there is information on expected associations. We conclude that Imiomics can be used for cross-sectional anomaly detection, associations and group comparisons.

Danielle F. Pace¹, Polina Golland¹, David Annese², Tal Geva^2,3, Andrew J. Powell^2,3, and Mehdi H. Moghari^2,3
1Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, United States, ²Department of Cardiology, Boston Children's Hospital, Boston, MA, United States, ³Department of Pediatrics, Harvard Medical School, Boston, MA, United States

We present a semi-automatic segmentation algorithm to create 3D heart models of children with complex congenital heart disease from 3D magnetic resonance images, which have promise for planning interventions. After 10-15 short-axis slices are segmented manually (in less than one hour of interaction time), a patch-based algorithm segments the remaining slices automatically. 3D surface models are then generated from the segmented blood pool and epicardium. The semi-automatic algorithm was evaluated using images acquired from 4 patients. Compared to manual segmentation, the proposed algorithm had surface-to-surface distance errors of 0.51 +/- 0.90 mm (blood pool) and 0.60 +/- 0.99 mm (epicardium).

Phillip G. D. Ward^1,2, Nicholas J. Ferris^2,3, Amanda C. L. Ng^2,4, David G. Barnes^1,5, David L. Dowe¹, Gary F. Egan^2,6, and Parnesh Raniga^2
1Clayton School of Information Technology, Monash University, Clayton, Victoria, Australia, ²Monash Biomedical Imaging, Monash University, Clayton, Victoria, Australia, ³Monash Imaging, Monash Health, Clayton, Victoria, Australia, ⁴Department of Anatomy and Neuroscience, The University of Melbourne, Parkville, Victoria, Australia, ⁵Monash eResearch Centre, Monash University, Victoria, Australia, ⁶School of Psychology and Psychiatry, Monash University, Victoria, Australia

This study introduces a new method for segmenting the cerebral venous vasculature, using quantitative susceptibility mapping (QSM) and susceptibility-weighted imaging (SWI). The method employs a Gaussian mixture-model to incorporate the QSM and SWI contrast, which then feds into a Markov random field model, augmented with a Gabor filter bank, to enhance hyper-intense, vessel-like structures and provide patient-specific venous cerebrovascular models.

Phillip G. D. Ward^1,2, Parnesh Raniga², Nicholas J. Ferris^2,3, Amanda C. L. Ng^2,4, David G. Barnes^1,5, David L. Dowe¹, Elsdon Storey⁶, Robyn L. Woods⁷, and Gary F. Egan^2,8
1Clayton School of Information Technology, Monash University, Clayton, Victoria, Australia, ²Monash Biomedical Imaging, Monash University, Clayton, Victoria, Australia, ³Monash Imaging, Monash Health, Clayton, Victoria, Australia, ⁴Department of Anatomy and Neuroscience, The University of Melbourne, Parkville, Victoria, Australia, ⁵Monash eResearch Centre, Monash University, Victoria, Australia, ⁶Department of Medicine, Monash University, Victoria, Australia,⁷Department of Epidemiology & Preventive Medicine, Monash University, Melbourne, Australia, ⁸School of Psychology and Psychiatry, Monash University, Victoria, Australia

The calculation of venous vascular metrics has been made possible without a contrast agent using susceptibility based magnetic resonance imaging (MRI) techniques, such as susceptibility weighted imaging (SWI) and quantitative susceptibility mapping (QSM), and a suitable vessel-enhancing filter. Whilst multiple filters have been proposed, the sensitivity of the final measurement to the choice of filter is an unexplored relationship. This study examines the correlation between venous density and the choice of image type and filtering technique in a large cohort of healthy elderly subjects.

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

Radiation therapy planning (RTP) based on magnetic resonance imaging (MRI) is an emerging application that benefits from the superior display of soft tissue contrasts and the delineation of tumor and critical organs. A new approach based on a Cartesian T1-Dixon acquisition has been introduced which makes it possible to classify soft tissue and cortical bone structures in the pelvic region. In this study, the consistency of the density value assignment of bone voxels is investigated on patient datasets who received both MR and CT imaging. The proposed assignment scheme gives correct overall mass densities on a population level.

Artifacts & Correction II

Tuesday 2 June 2015

Jon-Fredrik Nielsen¹ and Douglas C Noll^1
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States

RF-spoiled steady-state sequences (SPGR/T1-FFE/FLASH) offer rapid data acquisition and T1-weighting. The unbalanced gradient lobes (“spoiler gradients”) in these sequences are generally chosen empirically to be sufficiently large to achieve good spoiling and suppress ghosting artifacts, however the required spoiler gradient size for “good” ghost suppression is subjective and application-dependent. We present a simple acquisition and data processing strategy for improved ghost suppression in dynamic (or averaged) SPGR imaging, based on dynamically modulating the phase of the ghosts at the Nyquist frequency and removing that frequency component in pre-processing.

Kangrong Zhu¹, Robert F Dougherty², Matthew J Middione³, Hua Wu², Greig Scott¹, John M Pauly¹, and Adam B Kerr^1
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Center for Cognitive and Neurobiological Imaging, Stanford University, Stanford, CA, United States, ³Applied Sciences Laboratory West, GE Healthcare, Menlo Park, CA, United States

Multiband RF pulses are more likely to incur RF system nonlinearities than standard RF pulses because they typically push B1 to the RF amplifier limit and their envelope requires rapid RF supply current changes. Nonlinear distortion on a multiband RF pulse results in undesired higher harmonics being excited in the frequency domain. In this work, an RF pre-distortion approach, which compensates for the error between desired and measured RF waveforms, is adapted to mitigate the nonlinear distortion on multiband RF pulses.

Florent Eggenschwiler¹, Kieran R. O'Brien², Daniel Gallichan¹, Rolf Gruetter^1,2, and Jose P. Marques^3
1Laboratory for Functional and Metabolic Imaging, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Vaud, Switzerland, ²Department of Radiology, University of Geneva, Geneva, Geneva, Switzerland, ³Department of Radiology, University of Lausanne, Lausanne, Vaud, Switzerland

At ultra-high magnetic field, the inhomogeneous distribution of the B₁⁺ field can significantly impair the quality of turbo spin echo sequences. Static k_T-points where a unique pulse is designed in the STA regime and then dynamic k_T-points where a specific k_T-point pulse is optimized for each pulse of the turbo spin echo sequence were proposed to generate T₂-weighted images with increasingly improved signal and contrast homogeneities. In this work, the dynamic k_T-point design is further improved by allowing an increased flexibility for the optimization of the excitation pulse and by choosing carefully the cost function used in the optimization algorithm.

Rita Schmidt¹, Jean-Noel Hyacinthe², Andrea Capozzi³, Nikolas Kunz⁴, Rolf Gruetter^4,5, Arnaud Comment³, Lucio Frydman¹, and Mor Mishkovsky^5
1Chemical Physics, Weizmann Institute of Science, Rehovot, Israel, ²School of health, University of Applied Sciences and Arts Western Switzerland, Geneva, Switzerland, ³Institute of the Physics of Biological Systems, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁴Center of biomedical imaging (CIBM), Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁵Laboratory of Functional and Metabolic Imaging, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland

A general challenge of doing MRI with surface coils arises from the strong inhomogeneity of the radio-frequency (RF) field. This challenge also arises in ultrahigh field human MRI applications. Hybrid spatiotemporal encoding (SPEN) sequence is a new alternative for ultrafast acquisition, which recently was demonstrated providing higher immunity to B0 inhomogeneity compared to Echo Planar Imaging (EPI). The aim of the present work was to use SPEN scheme to correct for the B1 inhomogeneity by compensating in the “chirped” RF pulse amplitude for the spatial B1 distribution. Phantom as well as in-vivo animals’ experiments were conducted in 9.4 T MRI.

Chemseddine Fatnassi^1,2, Rachid Boucenna¹, Michael Betz¹, and Habib Zaidi^3
1Radio-oncology, Hirslanden Lausanne, Lausanne, vaud, Switzerland, ²Faculty of biology and Medicine, UNIL, Lausanne, vaud, Switzerland, ³Division of Nuclear Medicine and Molecular Imaging, Geneva University Hospital, Geneva, Switzerland

In 3D gradient echo (GRE) imaging, strong macroscopic B0 field gradients (Gr) are observed at air/tissue interfaces and in the presence of metallic objects. In particular, at low spatial resolution, the respective field gradients lead to an apparent increase in the intra-voxel dephasing and subsequently to signal loss or inaccurate R2* estimates. If the macroscopic gradient through a voxel (Gr) can be estimated, its influence can be removed through post-processing. The proposed correction strategies usually assume a linear phase evolution with time. However, near the edge of the brain, the paranasal sinus and temporal lobes, this assumption is often broken. In this work, we explore a model that considers a non-linear dependence of the phase evolution with echo time. The correction model is then weighted by the SNR map computed from the magnitude image in order to remove singularities caused by inaccurate field map estimation. We tested the performance of the proposed model for correcting of artifacts in a physical phantom with different MnCl2 concentrations and in vivo clinical studies.

Kun Zhou¹, Wei Liu¹, and Nan Xiao^1
1Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, Guangdong, China

B0 map is needed by many MR applications, for instance EPI distortion correction. Field mapping is normally accomplished via evaluating the phase evolution of multi-echo gradient echo measurements, which is acquired separately from routine scans and consume additional time. In this abstract, a novel B0 mapping method is presented, which reconstructs the field map via exploiting the information provided by active shimming. Its application on EPI distortion correction is demonstrated.

Hong Shang^1,2, Peder E.Z. Larson^1,2, Galen Reed³, Eugene Milshteyn^1,2, Cornelius von Morze¹, Frank Ong⁴, Jeremy W. Gordon¹, Jonathan I. Tamir⁴, 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, ³HeartVista, Menlo Park, California, United States, ⁴Electrical Engineering and Computer Science, UC Berkeley, Berkeley, California, United States

A variable flip angle approach was designed for bSSFP transient state imaging to improve signal profile uniformity and off-resonance insensitivity by solving a non-convex optimization problem. HP C-13 MRI investigations with this variable flip angle scheme resulted in less blurring and higher SNR.

Jong Bum Son¹, John Hazle¹, and Jingfei Ma^1
1Imaging Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, United States

We developed an optimized region growing algorithm for phase correction that includes jointly considering the two candidate vectors in selecting the final output vector for each pixel during the region growing and an automated segmentation to handle spatially isolated objects. The algorithm was implemented and evaluated for water and fat separation for in vivo two-point Dixon imaging with flexible echo times.

Barbara Dymerska¹, Benedikt Poser², Markus Barth³, Siegfried Trattnig¹, and Simon Daniel Robinson^1
1High Field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Vienna, Austria, ²Department of Psychology and Neuroscience, Cognitive Neuroscience, Maastricht University, Maastricht, Netherlands, ³Centre for Advanced Imaging, The University of Queensland, Brisbane, Australia

Dynamic maps of B0 can be extracted from a time series of standard, single-echo EPI by calculating the echo-time-independent phase (or “offset”) from a reference measurement and subtracting this from the phase of each EPI. This yields a field map for each volume. In this study we i) improve the SNR of the field maps by using a GE rather than EPI reference scan and ii) modify the method for multi-channel coils. In experiments at 7 T we find that while B0 varies dramatically with head motion, phase offsets remain stable, allowing distortions to be accurately corrected.

Wieland A. Worthoff¹, Stefan Schwan¹, Johannes Lindemeyer¹, and N. Jon Shah^1,2
1Institute of Neuroscience and Medicine, Forschungszentrum Jülich GmbH, Jülich, Germany, ²Faculty of Medicine, Department of Neurology, RWTH Aachen University, JARA, Aachen, Germany

We present results from simulations of the magnetic fields generated by the susceptibility distribution of a field probe and its magnetic environment. In particular, the effect of an unmatched object residing at various distances in the vicinity of the probe is evaluated numerically and verified experimentally on a 9.4 T MRI scanner. Furthermore, we explore the impact of a susceptibility mismatch between the sample droplet and the liquid buffers on the field homogeneity within the droplet as well as the relationship to the length of buffers in order to optimize the field probe signal.

Barbara Dymerska¹, Benedikt Poser², Wolfgang Bogner¹, Eelke Visser³, Korbinian Eckstein¹, Pedro Cardoso¹, Roland Beisteiner^1,4, Markus Barth⁵, Siegfried Trattnig¹, and Simon Daniel Robinson^1
1High Field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Vienna, Austria, ²Department of Psychology and Neuroscience, Cognitive Neuroscience, Maastricht University, Maastricht, Netherlands, ³FMRIB Centre, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, ⁴Department of Neurology, Medical University of Vienna, Vienna, Austria, ⁵Centre for Advanced Imaging, The University of Queensland, Brisbane, Australia

The geometric distortions encountered in EPI can be corrected using static or dynamic distortion correction methods. The static approach does not account for temporal B0 changes caused by breathing and motion over the course of an fMRI experiment. Here, we propose a single-echo EPI sequence in which the echo time is jittered between two values. Field maps are calculated between adjacent volumes, enabling a reference-free dynamic distortion correction. We show that this method generates accurate maps of local deviations from the static magnetic field even in the presence of large motion (up to 12°) and fast breathing (0.36 Hz) without compromising BOLD sensitivity or spatiotemporal resolution.

Jacco A de Zwart¹, Peter van Gelderen¹, and Jeff H Duyn^1
1Advanced MRI, LFMI, NINDS, National Institutes of Health, Bethesda, MD, United States

MR signal stability in brain, and thus image quality, is significantly affected by physiological noise, predominantly from the respiratory cycle. Navigator echoes or real-time shimming (dynamic updating of shims and center frequency) can be used to compensate or correct for this but have practical limitations. We propose the incorporation of a covariance map of temporal signal fluctuations, acquired separately, in a navigator-based correction strategy. This covariance map describes the spatial distribution of the temporal field changes that are derived from navigator phase evolution during the multi-shot experiment.

Shinji Kurokawa¹, Yoshitaka Bito², and Hisaaki Ochi^1
1Central Research Laboratory, Hitachi, Ltd., Kokubunji-shi, Tokyo, Japan, ²Hitachi Medical Corporation, Kashiwa-shi, Chiba, Japan

Coherent ghost artifacts are likely to occur in 2D cine compressed sensing MRI. They are caused by a lack of randomness in sampling patterns and insufficient parallel reconstruction. We propose a novel method that suppresses these artifacts. A new constraint on mean in the time direction is applied to a conventional compressed sensing reconstruction. The method is applicable to general compressed sensing MRI as well as cine MRI. In numerical simulations, artifacts were reduced to 33% in RMSE without temporal blurring. In volunteer scanning, artifacts were suppressed without image degradation.

Yasha Khatamian¹ and J. Jean Chen^1
1Rotman Research Institute, Toronto, Ontario, Canada

This study investigated a fat-related artifact that occurs when spectral-fat saturation is used with spin-echo EPI. This artifact was detected in humans and phantoms for a variety of imaging parameter settings, was significant compared to task associated signal changes, diminished with longer TR as well as fewer slices, and exhibited a power spectrum with a single broad peak at 0.133/TR Hz. This finding is relevant for all spin-echo EPI experiments, including fMRI and DTI. While further investigation is required to understand the mechanisms causing this artifact, water excitation is recommended for acquiring fat-free images with spin-echo EPI.

Shengzhen Tao¹, Joshua D Trzasko¹, Yunhong Shu¹, Paul T Weavers¹, Seung-Kyun Lee², and Matt A Bernstein^1
1Radiology, Mayo Clinic, Rochester, MN, United States, ²GE Global Research, Niskayuna, NY, United States

The spatial encoding gradient field used in MRI always includes spatially-varying higher order fields known as concomitant fields. Different from conventional gradient systems whose concomitant fields only contain 2nd-order spatial dependency, some emerging MRI platforms employ asymmetric gradient system, whose concomitant fields also include zero- and first-order spatial dependencies. The first-order terms cause further image distortion and echo shift in echo planar imaging (EPI). In this work, we develop a generalized waveform pre-emphasis framework to correct first-order concomitant fields for arbitrary axial-coronal oblique EPI acquisitions on a head-only asymmetric gradient system, and provide closed-form mathematical expressions for determining pre-emphasis factors.

Elias Kellner¹, Bibek Dhital¹, Valerij G. Kiselev¹, and Marco Reisert^1
1Department of Radiology, Medical Physics, University Medical Center Freiburg, Freiburg, Germany

Gibbs-ringing originates from the convolution of sharp edges in an object with the point-spread function, which is typically a sinc-function. They are strongest when the sinc is sampled at the extrema, and virtually disappear when it is sampled at its zero crossings. In this work, we propose a correction method based on local subvoxel pixel shifts, such that the oscillations are locally sampled at the zero crossings, and hence disappear. Compared to the popular global filtering approach, the proposed significantly better removes the artifact, while it introduces less smoothing and preserves the edges.

Aaron T Hess¹ and Matthew D Robson^1
1Oxford Centre for Clinical Magnetic Resonance Research (OCMR), Oxford, Ox, United Kingdom

When acquired spoiled gradient echo images at flip angles significantly larger than the Ernst angel, the transition to the steady state may not be smooth as predicted by the theoretical signal equations but can oscillate causing artefacts in images. A novel six point spoiling scheme is proposed that minimizes the number of unwanted echoes refocused and ensure that refocused echoes are return to the center of k-space which intern enables the use of RF-spoiling to further remove their effect. This technique is found to improve the approach to steady state compared to using a constant gradient with RF spoiling.

Yongchuan Lai¹, Weiwei Zhang¹, Baogui Zhang¹, and Bing Wu^1
1GE Healthcare, Beijing, China

FSE cusp artifacts are caused by collective effects of the gradient non-linearity and B0 inhomogeneity in regions distant from the magnet center. This can present as the so-called feather-like artifacts associated with sagittal plane FSE images when the phase encoding is along the S/I direction. We propose a simple yet effective method to eliminate the root cause of the artifacts by spatial saturation and further improve its practical implementation to reduce total RF bandwidth and overall SAR with two-stage saturation: frequency saturation and special saturation.

Genevieve M LaBelle¹ and Brad P Sutton^2,3
1Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ²Bioengineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ³Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Illinois, United States

Geometric distortion is a highly prevalent issue for Echo-Planar Imaging (EPI), due to long readout times and field inhomogeneities. Previously, the measured point spread function (PSF) has been shown to be effective in correcting this distortion. In this work, we reconstruct an image quickly and address the distortion with a point spread function map that was generated entirely through simulation using the trajectory and field map. The distortion correction with this approach is shown to be better than k-space based iterative reconstructions and is robust to high magnetic field maps when we use an optimal trajectory.

Marco Reisert¹ and Michael Herbst^1,2
1Medical Physics, University Medical Center Freiburg, Freiburg, Germany, ²Department of Radiology, John A. Burns School of Medicine, Honolulu, Hawai, United States

Off-resonance effects cause distortion artefacts in single shot EPI data, even when parallel imaging is incorporated. Multi-shot acquisition methods can partially solve these issues, but are susceptible to phase differences between different shots, in particular in DWI. In this work we combine multiplexed sensitivity encoding (MUSE) with an interleaved bottom-up/top-down traversal of k-space of the individual k-space segments to completely remove susceptibility induced distortions without any reference scan.

Guoxiang LIU¹ and Takashi UEGUCHI^1
1CiNet, National Institute of Information and Communications Technology, Suita, Osaka, Japan

We present a technique for planar imaging (EPI) ghost reduction at an ultrahigh field human whole body MRI scanner with a high power gradient insert system. Our approach can be implemented as post processing without extra scan. The key of the proposed method is the idea to derive the parent only region automatically without defined region of interest (ROI) placed on the parent image. Phantom experiments were performed using gradient-echo EPI and spin-echo EPI sequences.

Amy Walker^1,2, Gary Liney^1,2, Lois Holloway^1,2, Jason Dowling³, David Rivest-Henault³, and Peter Metcalfe^1,2
1Center for Medical Radiation Physics, University of Wollongong, Wollongong, NSW, Australia, ²Medical Physics, Liverpool and Macarthur Cancer Therapy Centres and Ingham Institute for Applied Medical Research, Liverpool, NSW, Australia, ³Commonwealth Scientific and Industrial Research Organisation, Australian E-Health Research Centre, Brisbane, Queensland, Australia

This study compares image acquisitions with static and continuously moving table. The goal was to investigate differences in geometric distortion and the impact this may have on radiotherapy treatment planning.

Hyun-Soo Lee¹, Seung Hong Choi², and Sung-Hong Park^1
1Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea, ²Department of Radiology, Seoul National University College of Medicine, Seoul, Korea

The centric PE order is often preferable for bSSFP in physiological MR imaging to maximize signal contrast. However, the centric PE ordered bSSFP can cause artifacts due to eddy current and transient oscillation. Existing compensation schemes such as pairing or double averaging still have limitations in application for physiological MRI. In this study, we tested three new compensation schemes based on complex averaging of two datasets with different dummy scans or two datasets with different pairing schemes. The proposed schemes suppressed the eddy current and transient oscillation artifacts, while maintaining the temporal resolution the same as the original centric scheme.

Michael N Hoff¹, Jalal B Andre¹, and Qing-San Xiang^2
1Radiology, University of Washington, Seattle, Washington, United States, ²Physics, University of British Columbia, Vancouver, British Columbia, Canada

Two unique analytical techniques have shown to demodulate balanced steady state free precession (bSSFP) images of their dependence on magnetic field inhomogeneity and mitigate subsequent artifacts. Here both the Algebraic and Geometric solutions are compared through simulation and in vivo application, with a focus on evaluating artifact correction, solution error, and solution variability as a function of tissue and noise level. A hybrid Geometric-Algebraic solution is formulated to exploit the strengths of both solutions and minimize reconstructed image variance, yielding a robust solution when compared to standard complex averaging of phase-cycled bSSFP images.

Reconstruction of Dynamic Data

Tuesday 2 June 2015

Loris Cannelli¹, Paolo Scarponi¹, Gesualdo Scutari¹, and Leslie Ying^1
1Electrical Engineering, University at Buffalo, Buffalo, NY, United States

In this work we present a novel and very general optimization algorithm customized for dynamic MRI reconstruction under the compressed sensing framework. The size of this kind of problems is usually huge: for this reason is compulsory to design algorithms capable to manage a large amount of data in an efficient way. Our approach exploits the benefits of a parallel nature, it relies on a smart decomposition of the original problem and it also possesses the ability of recognizing the elements that will be zero at the solution, taking thus advantage of the sparse structure of the problem itself.

Albert Jang^1,2, Alexander Gutierrez³, Di Xiao², Curtis A. Corum¹, Vuk Mandic⁴, Jarvis Haupt², and Michael Garwood^1
1Center for Magnetic Resonance Research and Department of Radiology, University of Minnesota, Minneapolis, MN, United States, ²Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN, United States, ³Department of Mathematics, University of Minnesota, Minneapolis, MN, United States, ⁴School of Physics and Astronomy, Department of Physics, University of Minnesota, Minneapolis, Minneapolis, MN, United States

Spatiotemporal-based encoding offers certain advantages over traditional Fourier-based encoding, enabling an alternative way of doing MRI. Two new reconstruction approaches, maximum-likelihood estimation (MLE) and total variation regularization (TVR), are evaluated for spatiotemporal encoding and compared with conventional methods (Cartesian gridded Fourier Transform and pseudo-inverse). It is demonstrated that MLE and TVR generate better images in terms of resolution and can compensate for non-uniform excitation profiles as well.

Muhammad Usman¹ and Claudia Prieto^1
1Division of Imaging Sciences and Biomedical Engineering, King's College London, London, United Kingdom

Standard Compressed Sensing (CS) techniques require signal/image to be a linear combination of very few coefficients in a transform representation. For dynamic cardiac MRI, examples of commonly used linear transforms are Wavelets, finite differences, temporal Fourier Transform and Principal Component Analysis (PCA). Nonlinear data reduction techniques such as Kernel PCA (KPCA) have the advantage over linear methods that these can detect nonlinearity or higher order moments within the given data set. By using appropriate nonlinear basis, complex features in the signal are expected to become separable that can be exploited for better signal classification or more compact representation of the signal. For MRI, this could be useful for a) better signal sparsity for CS and/or b) separation of signal content from artifacts in the undersampled reconstruction. Recently for retrospectively undersampled Cartesian cardiac CINE, compared to standard CS techniques, KPCA has been shown to more efficiently represent intra-frame spatial correlations for frame by frame reconstruction. In this work, we propose to accelerate real time dynamic cardiac CINE by exploiting both spatial and temporal denoising using kernel PCA. Prospective golden angle radial MR acquisitions, performed in 3 volunteers, demonstrate the feasibility of proposed framework for up to 8 fold accelerated real time CINE.

Mei-Lan Chu^1,2, Hing-Chiu Chang¹, Hsiao-Wen Chung², Trong-Kha Truong¹, Mustafa R Bashir³, and Nan-kuei Chen^1,3
1Brain Imaging and Analysis Center, Duke University, Durham, North Carolina, United States, ²Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan, ³Department of Radiology, Duke University Medical Center, Durham, North Carolina, United States

POCSMUSE is a general post-processing algorithm capable of reducing motion-related artifacts in MRI data, using the RF coil sensitivity profile as a constraint. It can be applied to reduce artifacts of various patterns, ranging from breathing-induced artifacts in abdominal FSE imaging to aliasing artifacts due to shot-to-shot phase variations in interleaved DWI. POCSMUSE is compatible with existing motion artifact correction schemes, and can be used to further improve the image quality in data produced by existing artifact correction procedures.

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

A novel application-specific reconstruction approach is proposed for accelerated intracranial time-resolved contrast-enhanced MR angiography. The proposed model-based compressed sensing (CS) technique utilizes a gamma-variate based model of contrast bolus propagation to constrain the reconstruction. The use of robust L1 norm allows to reconstruct abnormal dynamics not accounted for by the model. The proposed technique was initially validated in phantom simulations and in-vivo data and shown to improve spatial resolution compared to parallel imaging and general CS, while maintaining temporal fidelity.

Qiu Wang¹, Boris Mailhe¹, Robert Grimm², Marcel Dominik Nickel², Kai Tobias Block³, Hersh Chandarana³, and Mariappan S. Nadar^1
1Imaging and Computer Vision, Siemens 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

Dynamic contrast-enhanced MRI is widely used in clinical practice, due to its ability to reveal clinically significant pathology. Faster acquisition is critical since the acquisition has to be completed within a short time after contrast injection. Sparse-model based reconstruction is one of the techniques to recover high quality image for accelerated acquisitions. Sparse constraints correlated with the temporal dimension allow high spatio-temporal resolution. This work proposes a new sparse model and a reconstruction acceleration algorithm designed for DCE MRI. Experimental results demonstrate the effectiveness of the proposed method with superior image quality and time curves.

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

To address the tradeoff of spatial and temporal resolution in dynamic contrast-enhanced MRI, schemes using pseudorandom trajectories and view-sharing (VS) have been proposed. Compressed sensing (CS) has shown promise to reduce VS and temporal footprint in these schemes. However, validating these techniques and determining a reduced temporal footprint is challenging without a ground truth. We present a novel approach that uses variable flip angle (VFA) acquisition and retrospectively applies sampling and CS/VS schemes to fully-sampled VFA data, allowing comparison with a ground truth in vivo. Results suggest that CS reconstruction with reduced VS data is a suitable alternative to VS.

Courtney K Morrison¹, Roberta M Strigel^1,2, Kang Wang³, James H Holmes³, Alexey Samsonov², Frank R Korosec^1,2, and Julia Velikina^1
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ²Radiology, University of Wisconsin-Madison, Madison, WI, United States,³Global MR Applications and Workflow, GE Healthcare, Madison, WI, United States

Breast DCE-MRI with undersampled k-space segmentation schemes and reconstructed using view-sharing allows for improved temporal resolution while maintaining high spatial resolution. However, view sharing introduces a temporal footprint longer than the frame rate. This work demonstrates the feasibility of reconstructing undersampled breast DCE-MRI data with a novel compressed sensing technique. This technique preserves image quality while reducing the temporal footprint to a single phase.

Yijing Wu¹, Kevin M Johnson¹, Patrick A Turski², Kai Niu¹, YinSheng Li¹, GuangHong Chen¹, and Chuck A Mistretta^1
1Medical Physics, University of Wisconsin, Madison, WI, United States, ²Radiology, University of Wisconsin, Madison, WI, United States

Time-resolved 3D contrast-enhanced MR angiography (TR CE-MRA) often requires highly accelerated image acquisition to achieve clinically desired temporal and spatial resolutions. Combination of VIPR acquisition with improved multi-channel coils and advanced reconstruction techniques such as Parallel Imaging (PI) and Compressed Sensing (CS), offers substantially greater acceleration than past methods. However, image quality is restricted by poor SNR due to the limited amount of data used for reconstruction. In this work, we explore the temporal similarity of TR CE-MRA to further improve the SNR and image quality.

Ziyi Wang¹, Sheng Fang¹, and Hua Guo^1
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, Beijing, China

High spatiotemporal dynamic imaging reconstruction is a challenging topic that has important clinical applications [1]. k-t low-rank structure (SLR) exploits sparsity and low-rank structure of MRI image series, and has been successfully applied [2]. In k-t SLR, the low rank penalty is imposed to the entire dynamic image series, which may result in loss of fine image structures. In this study, a k-t nonlocal SLR algorithm (k-t nSLR) that introduces nonlocal patch similarities is proposed [3], [4]. It imposes low rank penalty only to patches selected out throughout the image series and adaptively determines threshold. The in vivo dynamic cardiac image reconstruction demonstrate that our k-t nSLR has an enhanced performance in anatomic details preservation.

Elwin Bassett^1,2, Ganesh Adluru², Promporn Suksaranjit³, Brent D. Wilson³, Edward VR DiBella², and Daniel Kim^2
1Physics, University of Utah, Salt Lake City, Utah, United States, ²UCAIR, Radiology, University of Utah, Salt Lake City, Utah, United States, ³Cardiology, Internal Medicine, University of Utah, Salt Lake City, Utah, United States

We sought to improve our previously described rapid real-time cine MRI with Cartesian undersampling and temporal total variation (TTV) constraint reconstruction by using a combination of radial k-space sampling and block-wise low-rank (BWLR) constraint. We imaged a resolution phantom, and our results show that BWLR produces 76% higher effective spatial resolution than TTV. We imaged 14 patients and 1 volunteer. Our experiments show that, compared with TTV, BWLR yields 17% higher effective spatial resolution without sacrificing diagnostic confidence determined by 2 cardiologists.

Samuel T Ting¹, Rizwan Ahmad¹, Ning Jin², Juliana Serafim da Silveira¹, and Orlando P Simonetti^1
1The Ohio State University, Columbus, Ohio, United States, ²Siemens Healthcare, Chicago, Illinois, United States

We combine the Variable density Incoherent Spatio-Temporal Acquisition (VISTA) sampling pattern with a Fast Iterative Shrinkage Thresholding Algorithm (FISTA) implementation of SENSE to achieve online real-time, free-breathing cardiac cine imaging with < 40 ms temporal resolution and < 2x2 mm2 in-plane spatial resolution with low latency (< 10 second) reconstruction time. We test our method in five healthy volunteers and demonstrate diagnostically sufficient image quality compared to conventional segmented techniques with no significant difference in measurement of volumetric parameters.

Merlin J Fair^1,2, Peter D Gatehouse^1,2, Peter Drivas², and David N Firmin^1,2
1NHLI, Imperial College London, London, United Kingdom, ²NIHR Cardiovascular BRU, Royal Brompton Hospital, London, United Kingdom

A coil sensitivity calibration technique for parallel imaging that makes use of multiple free-breathing prescans (MFP) to give accurate calibration data during free-breathing, without reducing acceleration, is compared with a temporal calibration technique in 20 prospectively subsampled first-pass myocardial perfusion datasets.

Yuan Le¹, Marcel Dominik Nickel², Randall Kroeker³, Christian Geppert², Bruce Spottiswoode³, and Chen Lin^1
1Radiology and Imaging Science, Indiana University School of Medicine, Indianapolis, IN, United States, ²Siemens Healthcare, Erlangen, Bavaria, Germany,³Siemens Medical Solutions, NC, United States

In order to investigate the representation of complex tumors when using modern, view sharing MRI sequences such as TWIST, dynamic contrast-enhanced TWIST raw data were generated from static background and a fractal tumor phantom raw data scaled with an enhancement model. Our results show that, for typical spatial resolution in clinical breast DCE-MRI, TWIST parameters of pA=20%-25% and pB=33% provided an enhancement measurement with lowest RMS error. The measured shape irregularity of the tumor changed in a less predictable way, however, which should be taken into consideration in the texture analysis of the images.

Yu Y. Li^1
1Radiology, Imaging Research Center, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, United States

The presented work developed a new approach to k-t space image reconstruction from highly undersampled data using low-rank matrix completion. It was found that the low-rank nature of k-t space data matrix is dependent on k-space locations due to motion sensitivity differences and k-t imaging can benefit from region-by-region image reconstruction. This work demonstrated that low-rank matrix completion can generate multi-cycle multi-phase cardiac images from non-segmented data collected at a speed of <50 ms per time frame with free-breathing, providing an approach to examining non-periodic cardiac behaviors in clinical practice.

Suhyung Park¹, Eung Yeop Kim², and Jaeseok Park^3
1Center for Neuroscience Imaging Research, Institute for Basic Science (IBS), Sungkyunkwan University, Suwon, Gyeong Gi-Do, Korea, ²Department of Radiology, Gachon University Gil Medical Center, Incheon, Korea, ³Biomedical Imaging and Engineering Lab., Department of Global Biomedical Engineering, Sungkyunkwan University, Suwon, Gyeong Gi-Do, Korea

Dynamic contrast-enhanced magnetic resonance angiography (DCE-MRA) requires high spatiotemporal resolution, and typically employs subtraction between static reference and dynamic images followed by maximum intensity projection (MIP) to visualize time-varying angiograms. Nevertheless, the subtraction-based DCE-MRA suffers from incomplete suppression of background signals in the presence of motion-induced voxel misregistration, potentially impairing the detectability of small distal vessels. In this work, we propose a novel reconstruction framework, dual projected background nulling compressed sensing (BANC), for robust separation of dynamic contrast-enhanced angiograms, in which we decompose x-t images into background static tissue signals (low rank component), background motion-induced signals (sparse component I), and DCE angiograms of interest (sparse component II) and then jointly estimate them while selectively nulling multiple background signals. Simulations and experiments validate that the proposed method is, if compared with conventional methods, highly effective in generating dynamic angiograms with robust background suppression even at very high reduction factors (R~30).

Jaekyun Ryu¹ and Jang-Yeon Park^1
1Biomedical Engineering, IBS Center for Neuroscience Imaging Research, Sungkyunkwan University, Suwon, Gyungki-do, Korea

In this study, we show that there is an effective way to circumvent this problem in the original SPEN imaging scheme with no special reconstruction technique like SR reconstruction if we employ a 3D imaging scheme for SPEN imaging. The guideline for parameter setup not to meet the overlapping artifacts was also discussed. The proposed method was demonstrated by theory and phantom imaging

Zhifeng Chen¹, Ming Yang², Liyi Kang³, Ling Xia³, and Feng Liu^4
1Zhejiang University, Hangzhou, Zhejiang, China, ²Philips Healthcare, Jiangsu, China, ³Zhejiang University, Zhejiang, China, ⁴The University of Queensland, Queensland, Australia

DCE-MRI has been widely used for diagnosis of liver diseases like hepatic cirrhosis, tumor, etc. Now the existing DCE-MRI reconstruction algorithms such as iGRASP and L+S mainly focus on iteratively minimize the energy equation combine parallel imaging and sparsity penalties. The iterative reconstruction schemes require a lot of computational cost. The expensive computation has impeded the clinical application. We investigate a non-iterative scheme with separating parallel imaging and denoising operator in this abstract. Our non-iterative parallel imaging and image denoising reconstruction can result in comparable image quality to iterative schemes with greatly reduced time cost. The scheme improves the clinical applicability of high spatiotemporal resolution DCE.

Jingyuan Lyu¹, Yihang Zhou¹, Ukash Nakarmi¹, 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

This abstract presents a new approach to highly accelerated dynamic MRI using partial separability (PS) model. In data acquisition, k-space data is moderately randomly undersampled at the center k-space navigator locations, but highly undersampled at the outer k-space for each temporal frame. In reconstruction, the navigator data is reconstructed from undersampled data using structured low-rank matrix completion. After all the unacquired navigator data is estimated, the partial separable model is used to obtain the entire dynamic image series from highly undersampled data. The proposed method has shown to achieve high quality reconstructions with reduction factors up to 44, when the conventional PS method fails.

Feiyu Chen^1,2, Feng Huang³, Dan Zhu¹, Jia Ning¹, and Huijun Chen^1
1Center for Biomedical Imaging Research, School of Medicine, Tsinghua University, Beijing, China, ²Electrical Engineering, Stanford University, Stanford, California, United States, ³Philips Healthcare (Suzhou). Co. Ltd, Jiangsu, China

Dynamic contrast-enhanced MR imaging is a promising technique for treating various hepatic diseases. Breath-hold 3D Gradient echo sequence is currently used for achieving multi-phase volumetric images of the liver. However, conventional 3D whole-liver imaging requires a breath-hold duration of more than 30 seconds. 2D-CAIPIRINHA, which accelerates the acquisition through controlled-aliased under-sampling, has been applied to reducing breath-hold duration to eight seconds at a reduction factor of four. However, down-sampling of k-spaces leads to aliasing artifacts in reconstructed images. In this research, we propose a new method combining temporal-shifted 2D-CAIPIRINHA sequence with PEAK-GRAPPA reconstruction. This approach further utilizes additional information acquired from the free-breathing periods before and after the breath-holding, which were wasted in traditional acquisition, to reduce the aliasing artifacts.

Bo Li^1,2, Cihat Eldeniz¹, Jue Zhang^2,3, Jing Fang^2,3, and Hongyu An^1
1Biomedical Research Imaging Center, Department of Radiology, School of Medicine, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States, ²College of Engineering, Peking University, Beijing, China, ³Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China

We proposed a compressed sensing reconstruction method that utilizes neighboring respiration phases as constrain to minimize image reconstruction artifacts using highly under-sampled stack-of-stars acquisition. In vivo results show the promise of the approach.

Sunrita Poddar¹, John D Newell², and Mathews Jacob^1
1Electrical and Computer Engineering, University of Iowa, Iowa City, IA, United States, ²Radiology, University of Iowa, IA, United States

We propose an algorithm for high time resolution multi-slice free breathing ungated dynamic imaging. This is of utmost importance to paediatric patients, obese subjects and subjects with compromised pulmonary function who cannot hold their breath sufficiently long for a breath-hold exam. We acquire cardiac k-space data using a novel acquisition scheme and reconstruct the image series assuming that the images lie on a low-dimensional manifold. We find that good spatial and temporal resolution images are obtained in a reasonable acquisition time without the need for any physiological monitors. Our results are compared to the traditional ECG gated breath-held method.

Morteza Mardani¹, Leslie Ying², and Georgios B Giannakis^3
1University of Minnesota, Falcon Heights, MN, United States, ²Buffalo University, New York, United States, ³University of Minnesota, Minneapolis, MN, United States

Our advocated approach builds on three-way tensors and leverages spatiotemporal correlations of the ground truth images through tensor low rank. CP/PARAFAC decomposition of tensors is adapted [7], and a tomographic approach is put forth that leverages the tensor low rank to recursively learn the low-dimensional subspace from undersampled k-space data. In the nutshell, the novel approach allows real-time data acquisition without gating or breath-holding, yet being able to recover high-quality dynamic cardiac images from high-dimensional even under-sampled tensors `on-the-fly'. It means the images can be reconstructed while the data is still being acquired.

Esben Plenge¹, Tal Shnitzer¹, and Michael Elad^1
1Technion - Israel Institute of Technology, Haifa, Israel

In this study we present a new dictionary-based model and its application as sparsifying operator in a low-rank plus sparse matrix decomposition. According to the model, short temporal signals of a dynamic MRI sequence are sparse under a non-linear transformation using a trained dictionary. We validate the model, quantitatively and qualitatively in the context of reconstruction of under-sampled abdominal MRI using a numerical phantom and in vivo MRI data.