Powerful Acquisition & Reconstruction Techniques

Monday 12 May 2014   14:15 - 15:15

Space 1/Power Poster Theatre & Traditional Poster Hall 
Moderators: Priti Balchandani, Ph.D. & Nicholas R. Zwart, Ph.D.

Click on this video icon to view the introductory session.

Florian Knoll¹, Thomas Koesters¹, Ricardo Otazo¹, Tobias Block¹, Li Feng¹, Kathleen Vunckx², David Faul³, Johan Nuyts², Fernando Boada¹, and Daniel K Sodickson^1
1Bernard & Irene Schwartz Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, New York, United States,²Department of Nuclear Medicine, K.U. Leuven, Leuven, Leuven, Belgium, ³Siemens Medical Solutions USA, New York, United States

While both measurements can be performed simultaneously with current state of the art PET-MR scanners, the data sets are processed in two separate reconstruction pipelines. The two different datasets are only combined at the visualization stage. We propose a new iterative reconstruction framework that treats MR and PET as one single data acquisition, and jointly reconstructs both image sets. In this way joint information of the underlying anatomy is shared during the iterations between both sets of images. In particular the lower resolution and lower SRN PET reconstruction can benefit from the superior soft tissue contrast of the MR.

Julianna Ianni¹ and William A. Grissom^1
1Biomedical Engineering, Vanderbilt University, Nashville, TN, United States

An algorithm for joint non-Cartesian image reconstruction and k-space trajectory error correction is presented. Results are shown from validations in simulated radial phantom data and in-vivo brain data collected with a center-out radial trajectory at 7T.

Daniel S. Weller¹ and Jeffrey A. Fessler^1
1EECS, University of Michigan, Ann Arbor, MI, United States

Fast, undersampled single-shot k-space trajectories have applications in functional and dynamic imaging, but their long readouts cause artifacts in the presence of field inhomogeneity. We propose an extension of a recently developed algorithm for fast L₁-SPIRiT reconstruction of undersampled non-Cartesian parallel imaging data, using a system matrix augmented with time-segmentation and a circulant preconditioner to yield high quality images quickly. We compare our method to the existing fast non-Cartesian L₁-SPIRiT using both simulated brain and real phantom data sets, where our proposed method effectively eliminates artifacts from field inhomogeneity.

Justin P. Haldar^1
1Signal and Image Processing Institute, University of Southern California, Los Angeles, CA, United States

This work presents a novel framework for constrained image reconstruction based on Low-Rank Modeling of Local k-Space Neighborhoods (LORAKS). We first demonstrate that k-space data for low-dimensional images can be mapped into high-dimensional matrices, such that the resulting matrices possess low-rank structure when the original images have limited support and/or slowly-varying phase. Subsequently, we propose a flexible approach to exploiting this low-rank structure that enables image reconstruction from undersampled data. The approach is analogous to a single-channel calibrationless generalization of GRAPPA, and is demonstrated to outperform sparsity-guided reconstructions of undersampled data in certain contexts.

Berkin Bilgic¹, Borjan Gagoski², Stephen Cauley¹, Audrey Fan³, Jonathan Polimeni¹, Ellen Grant², Lawrence Wald¹, and Kawin Setsompop^1
1Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ²Boston Children's Hospital, Boston, MA, United States, ³EECS, MIT, Cambridge, MA, United States

Wave-CAIPI acquisition enables highly accelerated parallel imaging with low g-factor penalty in a 3D gradient echo (GRE) scan by using i) 2D CAIPIRINHA controlled aliasing and ii) additional sinusoidal Gy and Gz encoding gradients during the readout of each phase encoding line. Herein, data acquisition and reconstruction time of QSM are dramatically reduced by the combination of Wave-CAIPI acquisition and fast phase processing and QSM algorithms. For Wave-CAIPI reconstruction, we extend the initial proposal by i) reducing the Wave reconstruction time 25× (from 360 min to 14 min), ii) estimating accurate point spread functions from a fast prior training acquisition, and iii) increasing the resolution 4-fold. This enables high quality whole-brain 7T QSM at 1×1×2 mm3 voxel size in 40 seconds.

Ziying Yin¹, Richard L. Magin¹, and Dieter Klatt^1
1Bioengineering, University of Illinois at Chicago, Chicago, Illinois, United States

Here we introduce a new pulse sequence, Diffusion-MRE (dMRE), for concurrent MRE and diffusion MRI. In dMRE shear motion and diffusion attenuation are encoded into the MR phase and magnitude by using a pair of bipolar gradients. The sequence timing is adjusted so that the bipolar gradients are sensitive to both coherent and incoherent intravoxel motion. The phantom results showed that simultaneous MRE and diffusion acquisition is feasible with no interference between MRE/diffusion acquisitions. The dMRE method may play a role in improving the use of MRE and diffusion for the detection of the diseases in liver and brain.

Steven Baete^1,2 and Fernando Emilio Boada^1,2
1Center for Biomedical Imaging, Dept. of Radiology, NYU Langone Medical Center, New York, New York, United States, ²CAI2R, Center for Advanced Imaging Innovation and Research, NYU Langone Medical Center, New York, New York, United States

Diffusion Spectrum Imaging (DSI) is able to non-invasively image the microstructure of the brain, including its complex distributions of intravoxel fiber orientations. A drawback of DSI is the requirement for a large number of q-space samples to adequately sample the Orientation Diffusion Function, leading to large measurement time. In order to accelerate DSI acquisitions we use a multi-echo stimulated echo diffusion sequence which samples multiple samples along a radial line in q-space in a single readout. This is combined with the recently proposed radial q-space sampling scheme, leading to, in the current configuration, a nearly fourfold speedup.

Zhiqiang Li¹, Dinghui Wang¹, Ryan K Robison¹, Nicholas R Zwart¹, Michael Schär^1,2, and James G Pipe^1
1Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ, United States, ²Philips Healthcare, Cleveland, OH, United States

Multi-slab 3D TSE imaging has better scan efficiency than its single-slab counterpart but suffers ringing and venetian blind artifacts. Meanwhile, spiral acquisition has high SNR efficiency and has been incorporated into 3D TSE imaging, but mostly using a spiral-out only trajectory. In this work we propose a 3D TSE technique, using a spiral-in/out trajectory to provide higher SNR efficiency, using sliding-slab to minimize the venetian blind artifacts, and using non-uniform slice phase encoding to reduce the ringing artifacts. The preliminary results demonstrate that the image quality is comparable to 2D Cartesian results.

Thomas Benkert¹, Martin Blaimer¹, Peter M. Jakob^1,2, and Felix A. Breuer^1
1Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Bavaria, Germany, ²Experimental Physics 5, University of Würzburg, Würzburg, Bavaria, Germany

Using balanced SSFP in combination with unequally spaced inversion pulses in between allows generating images with pure T2-contrast. Here, this concept is adapted using a radial trajectory to simultaneously generate several T2-weighted images and a standard bSSFP image (T2/T1-weighted) out of one single acquisition in a very short scan time (~1.3s). Additionally, a T1-weighted image can be obtained by a simple combination of these contrasts. Therefore, the proposed method is a promising candidate for clinical practice, especially for situations where long scan times limit the applicability of established protocols.

Daehyun Yoon¹, Valentina Taviani¹, Pauline Worters², and Brian Hargreaves^1
1Stanford University, Palo Alto, CA, United States, ²GE Healthcare, Menlo Park, CA, United States

We present a novel acquisition and reconstruction method to accelerate the Slice Encoding for Metal Artifact Correction (SEMAC) sequence for MR imaging near metallic implants. SEMAC adopted an additional encoding for slice-select dimension to resolve the off-resonance induced slice distortion, which severely increased the total scan time. In this abstract, we present a fast undersampling scheme and a simple reconstruction algorithm exploiting that the support of the extreme off-resonance spins is spatially limited. Our approach is to acquire on-resonance spins using less slice directional encoding and then to apply reduced FOV acquisition and reconstruction for off-resonance spins.

Anna Andreychenko¹, Sjoerd Crijns¹, Alexander Raaijmakers¹, Bjorn Stemkens¹, Peter Luijten¹, Jan Lagendijk¹, and Cornelis van den Berg^1
1Imaging Division, UMC Utrecht, Utrecht, Utrecht, Netherlands

Conventional methods to detect patient motion are based either on an external device, e.g. respiratory belt, or MR acquisition, i.e. navigator. An alternative technique has been proposed which monitors RF coil's impedance changes induced by the patient motion. However, this technique requires additional hardware. Here, we propose to detect the motion induced impedance variation by means of noise measurements. Using clinical MR systems we demonstrated the feasibility of the RF coil's thermal noise variance to detect respiratory motion. Moreover, noise covariance matrix of an array of coils contains spatial information which can potentially be used for motion prediction.

Kalina V Jordanova¹, Dwight G Nishimura¹, and Adam B Kerr^1
1Electrical Engineering, Stanford University, Stanford, California, United States

A new method to estimate the B₁ distribution in a volume by encoding in B₁ rather than along image space is presented. By acquiring multiple 1D projections using the BEAR B₁ mapping method with different phase sensitivities to B₁, an estimate of the B₁ distribution in each projected pixel is calculated using a convex optimization formulation. We validate this method through simulations and in vivo at 3T. With this method, the B₁ distribution in a volume can be estimated faster than in acquiring a 2D B₁ map. The method is potentially useful when B₁ varies rapidly in space.

Anagha Vishwas Deshmane¹, Dan Ma¹, Yun Jiang¹, Elizabeth Fisher², Nicole Seiberlich¹, Vikas Gulani^1,3, and Mark Griswold^1,3
1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, ²Biomedical Engineering, Cleveland Clinic Lerner Research Institute, Cleveland, Ohio, United States, ³Radiology, University Hospitals of Cleveland, Cleveland, OH, United States

In conventional weighted MRI, the presence of multiple species within a single voxel can alter signal intensity. However, it remains difficult to determine the species content which gives rise to this intensity change due to similarity in exponential-shaped signal evolutions. The uniqueness of signal evolutions generated through Magnetic Resonance Fingerprinting (MRF) allows for the identification of multiple species present within a single voxel. Here we demonstrate that MRF is able to resolve multiple material components from single, mixed voxels and validate the derived tissue fractions in a realistic simulation model.

Sandeep Kaushik¹, Dattesh Shanbhag¹, and Florian Wiesinger^2
1GE Global Research, Bangalore, Karnataka, India, ²GE Global Research, Munich, Germany

In this work, we propose a method for segmentation of bones in the head using a single echo zero TE (ZTE) pulse sequence using complex (magnitude and phase) information for efficient segmentation of bone and air regions in the head. We also introduce a histogram-based RF intensity correction method which enables threshold-based segmentation of bone, soft-tissue and air structures. We demonstrate excellent depiction of skull and vertebrae.

 

Ethan M Johnson¹ and John M Pauly^1
1Electrical Engineering, Stanford University, Stanford, California, United States

This work investigates UTE imaging with Dixon estimation of water, fat and short-T₂ anatomy. By searching over the space of feasible echo times, an acquisition scheme for multi-echo UTE acquisitions is identified that facilitates computation of component images using the echo images from a given acquisition strategy and field strength.