ISMRM 23rd Annual Meeting & Exhibition • 30 May - 05 June 2015 • Toronto, Ontario, Canada

Scientific Session • Novel Image Reconstruction Methods
 

Monday 1 June 2015

John Bassett Theatre 102

10:45 - 12:45

Moderators:

Justin P. Haldar, Ph.D., Daniel S. Weller, Ph.D.

10:45 0075.   Acquisition-free Nyquist ghost correction for parallel imaging accelerated EPI
Eric Peterson1, Murat Aksoy1, Julian Maclaren1, and Roland Bammer1
1Department of Radiology, Stanford University, Stanford, California, United States

When using parallel imaging accelerated echo planar imaging (EPI), Nyquist ghost correction typically necessitates an additional pre-scan or calibration lines. This work presents a method to perform Nyquist ghost correction in k-space on individual parallel imaging shots using an approached based on the singular value decomposition (SVD) that does not require a complete image or calibration lines. This obviates the standard Nyquist ghost correction pre-scan. A further advantage is that ghost correction can be performed on a shot-by-shot basis, which could be beneficial in the case of calibration drift or prospective motion correction.

10:57 0076.   
Externally Calibrated Parallel Imaging in the Presence of Metallic Implants
Curtis N Wiens1, Nathan S Artz1,2, Hyungseok Jang1, Alan B McMillan1, and Scott B Reeder1,3
1Department of Radiology, University of Wisconsin, Madison, Wisconsin, United States, 2Department of Radiological Sciences, St. Jude Children's Research Hospital, Memphis, Tennessee, United States, 3Department of Medical Physics, University of Wisconsin, Madison, Wisconsin, United States

3D multi-spectral imaging (3D-MSI) acquires multiple acquisitions at different frequency offsets, in order to excite signal over a wide range of off-resonance induced by metallic implants. Self-calibrated parallel imaging approaches can accelerate 3D-MSI techniques but a substantial amount of time is required to acquire calibration data for each offset. In this work we developed a method for exter-nally calibrated parallel imaging near metallic implants, and demonstrated feasibility using a hip prosthesis phantom and a volunteer with a cobalt/chromium/molybdenum alloy hip head placed posterior to the knee. Comparisons to self-calibrated parallel imaging showed significant reductions in acquisition time, particularly for SR-FPE with ~25% reductions in scan time.

11:09 0077.   Joint Compressed Sensing and Sparse Phase Retrieval: Reconstruction from a Combination of Complex and Magnitude-only k-space Measurements
Mehmet Akcakaya1, Vahid Tarokh2, and Reza Nezafat1
1Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, United States, 2Harvard University, Cambridge, MA, United States

In this study, we introduce a new sparsity-regularized reconstruction paradigm based on using conventional complex k-space measurements, along with magnitude-only k-space measurements available as side information.

11:21 0078.   
Simultaneous Multi-slice MRI Reconstruction using LORAKS
Tae Hyung Kim1 and Justin P. Haldar1
1Department of Electrical Engineering, University of Southern California, Los Angeles, CA, United States

This work proposes a novel approach to simultaneous multi-slice (SMS) parallel MRI reconstruction, based on the low-rank modeling of local k-space neighborhoods (LORAKS) framework. Compared to existing SMS reconstruction methods, the proposed SMS-LORAKS approach is flexible enough to reconstruct highly-undersampled SMS data in the absence of prior coil information or autocalibration data. SMS-LORAKS can also be applied to single-channel MRI data. Reconstruction results are shown with real retrospectively-undersampled MRI data to demonstrate the potential of the approach.

11:33 0079.   
Complex-Difference Constrained Reconstruction for Accelerated Phase Contrast Flow Imaging
Aiqi Sun1, Bo Zhao2, Rui Li1, and Chun Yuan1,3
1Center for Biomedical Imaging Research, School of Medicine, Tsinghua University, Beijing, China, 2Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States, 3Department of radiology, University of Washington, WA, United States

Phase-contrast (PC) cine MRI has been demonstrated as a promising technique for studying hemodynamics. However, it remains limited in clinical practice because of its long scan time. To date, a number of fast imaging methods have been applied to PC cine MRI, including kinds of k-t reconstruction algorithms, but most of them could cause temporal blurring and large deviation in flow measurements under higher acceleration rate. In this work, we propose a new complex-difference constrained reconstruction technique based on low-rank and sparsity model, and we further integrate it with ESPIRiT-based parallel imaging reconstruction to achieve even higher acceleration.

11:45 0080.   
Total Generalized Variation Based Joint Multi-Contrast, Parallel Imaging Reconstruction of Undersampled k-space Data
Adrian Martin1,2, Itthi Chatnuntawech1, Berkin Bilgic3, Kawin Setsompop3,4, Elfar Adalsteinsson1,5, and Emanuele Schiavi6
1Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 2Applied Mathematics, Universidad Rey Juan Carlos, Mostoles, Madrid, Spain, 3A. A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General hospital, Charlestown, MA, United States, 4Harvard Medical School, Boston, MA, United States, 5Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, 6Universidad Rey Juan Carlos, Mostoles, Madrid, Spain

Typical clinical MRI routines include multiple imaging of the same region of interest under different contrast settings. In this work we extend the Total Generalized Variation (TGV) operator to jointly reconstruct multiple MRI contrasts from undersampled k-space data using one or more receiver coils. The multi-contrast TGV operator exploits the structural similarities of the multi-contrast images to preserve these details in the reconstruction process. The proposed technique yields to improved reconstruction accuracy when compared to widely used parallel imaging reconstruction methods such as SENSE and Total Variation regularized SENSE.

11:57 0081.   Non-linear phase correction in model-based reconstruction of the diffusion tensor
Jose Raya1,2 and Florian Knoll1,2
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

In this work we investigate the value of a non-linear phase correction algorithm for model-based reconstruction of diffusion tensor images measured with non-Cartesian multi-shot diffusion-weighted sequences. The approach is used on data acquired with a diffusion-weighted radial spin echo sequence that acquires 2D navigators after each readout. We acquire data for the brain and the knee with resolutions of 1 and 0,7 mm2 respectively. Data reconstructed without phase correction showed systematic deviation of the MD and FA. In summary the use of non-linear phase correction is essential in iterative reconstruction of diffusion data acquired with multishot sequences.

12:09 0082.   Wave-CS: Combining wave encoding and compressed sensing
Andrew T Curtis1, Berkin Bilgic2, Kawin Setsompop2, Ravi S Menon3, and Christopher K Anand1
1Computing and Software, McMaster University, Hamilton, Ontario, Canada, 2Martinos Center for Biomedical Imaging, Charlestown, MA, United States,3Robarts Research Institute, London, Ontario, Canada

The recently introduced WAVE encoding modulates the phase/slice gradients such that the read trajectory corkscrews through k-space, sampling additional spatial frequencies. Here we investigate the combination of WAVE encoding with compressed-sensing (CS) via random phase/slice under-sampling patterns and sparsity enforcing reconstruction, which we term Wave-CS. The open source BART toolkit is leveraged for reconstruction. The additional phase encoding and the aliasing generated in the read direction from WAVE was found to provide significant performance benefits in the CS-framework as compared to regular Cartesian sampling, with improved reconstruction quality and faster iterative convergence for matched acceleration factors.

12:21 0083.   TrueCISS: Genuine bSSFP Signal Reconstruction from Undersampled Multiple-Acquisition SSFP Using Model-Based Iterative Non-Linear Inversion
Tom Hilbert1,2, Damien Nguyen3, Tobias Kober1,2, Jean-Philippe Thiran2, Gunnar Krueger1,2, and Oliver Bieri3
1Siemens ACIT – CHUV Radiology, Siemens Healthcare IM BM PI & Department of Radiology CHUV, Lausanne, Switzerland, 2LTS5, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 3Radiological Physics, Department of Radiology, University of Basel, Basel, Switzerland

Balanced steady-state free-precession (bSSFP) is prone to local field inhomogeneities, typically appearing as signal voids, i.e. banding-artifacts. A new method, termed true constructive interference in steady state (trueCISS), is proposed based on the acquisition of eight highly undersampled bSSFP k-spaces with different radio-frequency (RF) phase increments. A model-based non-linear inversion is used to fit the bSSFP signal model onto the undersampled data, effectively estimating parameter maps that allow synthesizing the genuine bSSFP signal over the whole image, thus without any noticeable banding artifacts.

12:33 0084.   Multiscale Image Reconstruction for MR Fingerprinting
Eric Y. Pierre1, Dan Ma1, Yong Chen2, Chaitra Badve2, and Mark A. Griswold1,2
1Department of Biomedical Engineering, Case Western Reserve University, Cleveland, Ohio, United States, 2Department of Radiology, Case Western Reserve University & University Hospitals, Cleveland, Ohio, United States

To perform parameter mapping, Magnetic Resonance Fingerprinting (MRF) relies on highly efficient, highly undersampled trajectories to acquire the image series, yielding images contaminated by high aliasing noise. We propose an iterative multiscale method to denoise these images so as to reduce the length of image series required for accurate parameter mapping. The proposed method is shown to allow the simultaneous T1, T2, field inhomogeneity and proton density estimation at 1.17 mm2 resolution in vivo from a single 5.1s acquisition, representing a potential 4-fold increase in acquisition speed for MRF methods.