Parallel Imaging: Stretching the Limit
Thursday 6 May 2010
Room A6 10:30-12:30 Moderators: Ricardo Otazo and Jeffery Tsao

10:30 544

Fast MR Parameter Mapping Using K-T PCA
Frederike Hermi Petzschner1,2, Irene Paola Garcia Ponce3, Martin Blaimer4, Peter M. Jakob3, Felix A. Breuer4
1
Ludwig-Maximilians University, Institute of Clinical Neurosciences, Munich, Bavaria, Germany; 2Bernstein Center for Computational Neurosciences, Munich, Germany, Germany; 3University of Würzburg, Experimental Physics 5, Germany; 4Research Center Magnetic Resonance Bavaria, Germany

In this work, k-t PCA is demonstrated to be a promising acceleration technique for MR relaxation measurements, since the dynamics along the relaxation curve can be described by only a small number of principal components. In-vivo IR-TrueFISP experiments for quantitative T1, T2 & M0 parameter mapping acquired with up to 8-fold acceleration by using the k-t PCA concept are presented.

     
10:42 545.  

k-T Group Sparse Reconstruction Method for Dynamic Compressed MRI
Muhammad Usman1, Claudia Prieto1, Tobias Schaeffter1, Philip G. Batchelor1
1
King's College London, London, United Kingdom

Up to now, besides sparsity, the standard compressed sensing methods used in MR do not exploit any other prior information about the underlying signal. In general, the MR data in its sparse representation always exhibits some structure. As an example, for dynamic cardiac MR data, the signal support in its sparse representation (x-f space) is always in compact form.  In this work, exploiting the structural properties of sparse representation, we propose a new formulation titled ‘k-t group sparse compressed sensing’. This formulation introduces a constraint that forces a group structure in sparse representation of the reconstructed signal. The k-t group sparse reconstruction achieves much higher temporal and spatial resolution than the standard L1 method at high acceleration factors (9-fold acceleration).

     
10:54 546. 

Parallel Imaging Technique Using Localized Gradients (PatLoc) Reconstruction Using Compressed Sensing (CS)
Fa-Hsuan Lin1, Panu Vesanen2, Thomas Witzel, Risto Ilmoniemi, Juergen Hennig3
1
A. A. Martinos Center, Charlestown, MA, United States; 2Helsinki University of Technology, Helsinki, Finland; 3University Hospital Freiburg, Freiburg, Germany

The parallel imaging technique using localized gradients (PatLoc) system has the degree of freedom to encode spatial information using multiple surface gradient coils. Previous PatLoc reconstructions focused on acquisitions at moderate accelerations. Compressed sensing (CS) is the emerging theory to achieve imaging acceleration beyond the Nyquist limit if the image has a sparse representation and the data can be acquired randomly and reconstructed nonlinearly. Here we apply CS to PatLoc image reconstruction to achieve further accelerated image reconstruction. Specifically, we compare the reconstructions between PatLoc and traditional linear gradient systems at acceleration rates in an under-determined system.

     
11:06 547.  

Designing K-Space Trajectories for Simultaneous Encoding with Linear and PatLoc Gradients
Daniel Gallichan1, Gerrit Schultz1, Jürgen Hennig1, Maxim Zaitsev1

1University Hospital Freiburg, Freiburg, Germany

Recent work has shown that MR imaging can be performed using non-linear encoding gradients (PatLoc). Here we investigate the possibilities of combining non-linear encoding gradients with simultaneous use of the conventional linear gradients. We introduce the concept of a 'local k-space' to compare trajectories, as well as presenting a combination of a split-radial 4D trajectory which is able to exploit the advantages of varying spatial resolution across the FoV whilst retaining control over the resolution in the centre.

     
11:18 548.

A Time-Efficient Sub-Sampling Strategy to Homogenise Resolution in PatLoc Imaging
Hans Weber1, Daniel Gallichan1, Gerrit Schultz1, Jürgen Hennig1, Maxim Zaitsev1
1
University Hospital Freiburg, Dept. of Diagnostic Radiology, Medical Physics, Freiburg, Germany

Varying spatial resolution is one of the characteristic properties of MR imaging when using nonlinear gradient fields for spatial encoding, as realised by PatLoc. In the particular configuration of two orthogonal quadrupolar encoding fields, voxel size is inversely proportional to the distance to the FOV centre. In this work we present an iterative reconstruction method for sub-sampled PatLoc data that improves the local resolution at the centre and leads to shorter scan times for equivalent central resolution recovery. The method is demonstrated on simulated and experimentally acquired data.

     
11:30 549. 

An Assessment of O-Space Imaging Robustness to Local Field Inhomogeneities
Jason P. Stockmann1, R. Todd Constable2
1Biomedical Engineering, Yale University, New Haven, CT, United States; 2Diagnostic Radiology, Neurosurgery, and Biomedical Engineering, Yale University, New Haven, CT, United States

O-Space imaging permits highly-accelerated acquisitions using non-linear gradients to extract extra spatial encoding from surface coil profiles as compared with linear gradients.  For accurate reconstruction to occur, however, the curvilinear frequency contours created by the gradients must intersect one another at the appropriate locations, making the technique potentially vulnerable to local field inhomogeneity, such as the susceptibility gradients arising in the head near the sinuses.  This work shows that with appropriate regularization, O-Space imaging is robust to typical levels of field inhomogeneity.  Field inhomogeneity is shown to manifest itself as noise-like artifacts throughout the FOV rather than gross geometric distortion.

     
11:42 550. 

Highly Accelerated Multislice Parallel Imaging: Cartesian Vs Radial
Stephen R. Yutzy1, Nicole Seiberlich2, Jeffrey L. Duerk1,2, Mark A. Griswold2

1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States; 2Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, United States

Multiband imaging allows for multiple simultaneously acquired slices, thus giving an SNR benefit over conventional slice selection without potential artifacts from secondary phase encoding.  While methods have been shown that can separate the slides using parallel imaging for Cartesian trajectories, these methods are not compatible with non-Cartesian sampling.  Here we demonstrate the possibility of reconstructing two simultaneously acquired radial slices using an acquisition/reconstruction method known as radial CAIPIRINHA. We show that this method is capable of higher accelerations than possible with comparable Cartesian trajectories.

     
11:54 551. 

Blipped CAIPIRHINA for Simultaneous Multi-Slice EPI with Reduced G-Factor Penalty
Kawin Setsompop1,2, B. A. Gagoski3, J. Polimeni1,2, T. Witzel1, V. J. Wedeen1,2, L. L. Wald1,2
1Radiology, A. A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States; 2Harvard Medical School, Boston, MA, United States; 3EECS, Massachusetts Institute of Technology, Cambridge, MA, United States

The acquisition of simultaneous slices in EPI has the potential to increase the temporal sampling rate of fMRI or the number of diffusion directions obtained per unit time in diffusion imaging. In this work, we introduced a blipped CAIPIRINHA technique applicable to EPI acquisition and demonstrated its associated low g-factor penalty and 3x acceleration of the slices per second of acquisition. 3x slice-accelerated SE-EPI was acquired with retain SNR of close to unity. The 3x blipped CAIPIRINHA was also combined with 2x Simultaneous Image Refocusing (SIR) acquisition to create 6 simultaneous multi-slice GE-EPI acquisition with low g-factor penalty.

     
12:06 552.

SNR Quantification with Phased-Array Coils and Parallel Imaging for 3D-FSE
Charles Qingchuan Li1, Weitian Chen2, Philip J. Beatty2, Anja C. Brau2, Brian A. Hargreaves1, Reed F. Busse3, Garry E. Gold1
1Radiology, Stanford University, Stanford, CA, United States; 2Global Applied Science Laboratory, GE Healthcare, Menlo Park, CA, United States; 3Global Applied Science Laboratory, GE Healthcare, Madison, WI, United States

Current clinical MRI techniques often employ parallel imaging, partial Fourier and multicoil acquisition to decrease scan time while maintaining image quality. To aid in image quality assessment, image noise statistics can be measured by reconstructing noise-only acquisitions through an identical linear pipeline as signal data, which may involve signal data-dependent steps such as parallel imaging, partial Fourier homodyne and multichannel reconstructions. In this study it was shown that SNR and CNR measurements performed in 146 clinical knee MRIs using this quantification method significantly differ from the measurements obtained using the traditional foreground and background volume of interest approach.

     
12:18 553

A Mathematical Model Toward Quantitative Assessment of Parallel Imaging Reconstruction
Yu Li1, Feng Huang1, Wei Lin1, Arne Reykowski1

1Advanced Concept Development, Invivo Diagnostic Imaging, Gainesville, FL, United States

In this work, we propose a mathematical model that gives explicit representations for three different types of errors in parallel imaging reconstruction. These errors have different patterns in image space and affect the image quality in different fashions. This model offers a tool to extensively investigate how to quantitatively assess imaging quality beyond signal to noise ratio. Based on the proposed model, practical reconstruction techniques can be developed to suppress three types of errors to different degrees for improved overall imaging performance.

     

 

Back to Main Meeting

Back to Home