Image Reconstruction
 

Room 701 B

16:00-18:00

Chairs: Craig H. Meyer and Bradley P. Sutton


Time

Prog #

 
16:00 786. Image Reconstruction from Ambiguous PatLoc-Encoded MR Data

Gerrit Schultz1, Jürgen Hennig1, Maxim Zaitsev1

1University Hospital Freiburg, Freiburg, Germany

We present the first efficient and practical reconstruction algorithm for general non-bijective, curvilinear encoding fields. The algorithm makes use of parallel imaging techniques to resolve the occurring ambiguities. When overlapping coil sensitivities are considered, this can be done by a generalized SENSE-like reconstruction method, where acceleration is determined by the order of ambiguity. Nonlinearities are not treated as perturbations, but as an integral part of the reconstruction procedure. The algorithm is demonstrated for multipolar encoding fields, showing resolution enhancement at the periphery of the imaging region and loss of resolution towards the center, where the fields have a flat profile.

16:12 787.  Off-Resonance Effects in Non-Cartesian Parallel Imaging

Weitian Chen1, Peng Hu1, Chunlei Liu2, Craig H. Meyer1

1University of Virginia, Charlottesville, Virginia , USA; 2Stanford University, Palo Alto, California , USA

Off-resonance commonly exists in MR systems and can cause image artifacts in a number of MR imaging methods. MR scanning based on non-Cartesian sampling in k-space is sensitive to off-resonance. Off-resonance effects on non-Cartesian parallel imaging are complicated since they not only cause image blurring but can also interact with the unaliasing process. In this abstract, we use spiral scanning as an example to demonstrate off-resonance effects on non-Cartesian parallel imaging.

16:24 788. Fast, Mathematically Exact k-Space Sample Density Compensation for Rotationally Symmetric Interleaved Trajectories, and the SNR-Optimized Reconstruction from Non-Cartesian Samples

Dimitris Mitsouras1, Robert V. Mulkern1, 2, Frank J. Rybicki1

1Harvard Medical School & Brigham And Women's Hospital, Boston, Massachusetts, USA; 2Children's Hospital, Boston, Massachusetts, USA

A recently developed exact density compensation method solves a system of equations based on analytic cross-correlations of Fourier basis functions corresponding to the trajectory. Its application to image reconstruction is problematic since the system matrix size is equal to the number of samples (e.g., 65000-by-65000 for 256 image matrix). Here we show that rotationally symmetric interleaved trajectories, such as multi-shot spiral and PROPELLER, lead to a circulant block-Toeplitz system matrix, enabling fast solution within a few seconds.  Extensive simulation and experimental results show 10% RMS error reduction concurrently with 28% increase in SNR achieved for a 32-way interleaved spiral trajectory.

16:36  789. Improved Image Reconstruction for Partial Fourier Gradient-Echo EPI

Nan-kuei Chen1, Koichi Oshio2, 3, Lawrence P. Panych2

1Duke University, Durham, North Carolina, USA; 2Brigham and Women's Hospital, Boston, Massachusetts, USA; 3Keio University, Tokyo, Japan

The partial-Fourier EPI enables high-resolution fMRI scan at an optimal echo time. However, the partial-Fourier EPI may be degraded by artifacts that are not usually seen in other types of imaging. Those unique artifacts, to our knowledge, have not yet been systematically evaluated. Here we use k-space energy spectrum analysis to characterize two types of partial-Fourier EPI artifacts. We show that Type 1 artifact cannot be corrected with any post-processing method, and Type 2 artifact can be eliminated with an improved reconstruction method. We further propose a novel algorithm to reconstruct partial-Fourier EPI with minimal Type 2 artifact.

16:48  790. Magnetic Field Monitored Autofocus Deblurring for Improved Non-Cartesian Imaging

Florian Wiesinger1, Pekka Sipilae1, 2, Silke Maria Lechner1, 2, Rolf Feodor Schulte1

1GE Global Research, Munich, Germany; 2Munich University of Technology, Munich, Germany

Reliable non-Cartesian imaging requires experienced system tuning, as well as advanced reconstruction schemes incorporating B0 blurring correction. In this work an autofocus reconstruction method is described which achieves automatic B0 deblurring without prior information on the main magnetic field inhomogeneity. In order to also account for gradient field imperfections, magnetic field monitoring has been used to capture the exact encoding information simultaneous to data acquisition. Hence, high-quality spiral images were obtained in the presence of significant DB0 inhomogeneities and gradient field imperfections, without the need for extra system tuning, or calibration interactions.

17:00 791. Fast Conjugate Phase Image Reconstruction Based on a Chebyshev Approximation to Correct for B0 Field Inhomogeneity and Concomitant Gradients

Weitian Chen1, Christopher T. Sica1, Craig H. Meyer1

1University of Virginia, Charlottesville, Virginia , USA

Off-resonance can cause image blurring in spiral scanning and variousforms of image degradation in other methods.  Off-resonance can becaused by both B0 inhomogeneity and concomitant gradient fields.Previously developed off-resonance correction methods focus on thecorrection of a single source of off-resonance.  This work introducesa computationally efficient method of correcting for B0 inhomogeneityand concomitant gradients simultaneously.  The method is a fastalternative to conjugate phase reconstruction, with the off-resonancephase term approximated by Chebyshev polynomials. The proposedalgorithm is well suited for semi-automatic off-resonancecorrection, which works well even with an inaccurate or low-resolutionfield map.  The proposed algorithm is demonstrated using phantom andin vivo data sets acquired by spiral scanning.

17:12 792. Improved Time Series Reconstruction for Dynamic MRI

Uygar Sümbül1, Juan Manuel Santos1, John Mark Pauly1

1Stanford University, Stanford, California , USA

A fast statistical reconstruction algorithm for non-Cartesian dynamic MRI that yields a better temporal response is presented. The method is based on the famous Kalman filter and it provides increased frame rates through a fast algorithm. 4x accelerated reconstructions are presented that outperforms the sliding window reconstruction both in SNR and temporal resolution.

17:24 793. Optimizing K-T BLAST/SENSE Using FOCUSS and RIGR

Hong Jung1, Jong Chul Ye1

1Korea Advanced Institute of Science & Technology (KAIST), Guseong-dong Yuseong-gu, Republic of Korea

Recently, compressed sensing has been an active topic in dynamic MR imaging, and a new algorithm called k-t FOCUSS was proposed by employing FOCUSS algorithm to exploit the sparsity of the dynamic cine in x-f domain. Interestingly, k-t BLAST/SENSE algorithm turns out to be an approximation of k-t FOCUSS algorithm. Previously, temporal average was used to initialize these algorithms. However, it introduces the signal nulling artifact. The main contribution of this paper is to show that the signal nulling can be effectively suppressed by incorporating RIGR (Reduced-encoding Imaging by Generalized-series Reconstruction) initialization into k-t FOCUSS iteration. Interestingly, recently proposed SPEAR (Spatiotemporal domain based unaliasing employing sensitivity Encoding and Adaptive Regularization) algorithm turns out to be the first iteration of modified k-t FOCUSS that is optimal from compressed sensing perspective. Experimental result shows that highly accurate dynamic cine can be obtained even from severely down sampled data.

17:36 794. Generalized Reconstruction by Inversion of Coupled Systems (GRICS) Applied to Free-Breathing MRI

Freddy Odille1, 2, Pierre-André Vuissoz1, 2, Pierre-Yves Marie3, Jacques Felblinger1, 2

1Nancy University, Nancy, France; 2INSERM ERI 13, Nancy, France; 3University Hospital Nancy, Nancy, France

Correction for arbitrary motion (non rigid or affine) in reconstruction has been demonstrated recently. However, practical implementation is difficult, as a model is required to predict displacement fields at each sample time point. Small motion prediction errors are shown to propagate linearly in that reconstruction algorithm. This leads to reformulating reconstruction as two inverse problems which are coupled: motion compensated reconstruction (knowing motion), and motion model optimization (knowing the solution image). A fixed point multiresolution scheme is described for coupled systems inversion. This framework is shown to allow fully autocalibrated reconstructions, and is validated with free-breathing scans from healthy volunteers.

17:48  795. High-Resolution Pulmonary Perfusion Imaging in Rodents Using a Spatiotemporal Model

Sarah Schmitter1, Nilesh N. Mistry, 2, Cornelius Brinegar1, G. Allan Johnson, Zhi-Pei Liang1

1University of Illinois at Urbana-Champaign, Urbana, Illinois, USA; 2Duke University, Durham, North Carolina, USA

Pulmonary perfusion using Gd based contrast agents is an important tool for investigation of vascular pulmonary diseases. Imaging a rodent to study these disease models is challenging due to the high spatiotemporal demands. This work presents a new method for perfusion imaging in rodents using a novel spatiotemporal model. As compared to an existing method using interleaved radial imaging and sliding-window keyhole reconstruction, the technique presented in this work results in a ~6-fold improvement in temporal resolution.