Improvement in High Acceleration Parallel Magnetic Resonance Imaging Using Efficient Graph-based Energy Minimization Methods PDF Download
Are you looking for read ebook online? Search for your book and save it on your Kindle device, PC, phones or tablets. Download Improvement in High Acceleration Parallel Magnetic Resonance Imaging Using Efficient Graph-based Energy Minimization Methods PDF full book. Access full book title Improvement in High Acceleration Parallel Magnetic Resonance Imaging Using Efficient Graph-based Energy Minimization Methods by Gurmeet Singh. Download full books in PDF and EPUB format.
Author: Gurmeet Singh
Publisher:
ISBN:
Category :
Languages : en
Pages : 210
Get Book Here
Book Description
Author: Gurmeet Singh
Publisher:
ISBN:
Category :
Languages : en
Pages : 210
Get Book Here
Book Description
Author: Mihir Rajendra Pendse
Publisher:
ISBN:
Category :
Languages : en
Pages :
Get Book Here
Book Description
Magnetic resonance imaging (MRI) is a powerful imaging modality that is widely used in medicine for both clinical and research purposes. Despite its success, there is still a demand for improved image quality in the form of higher SNR and resolution and a promising approach to achieve this is with higher static field strengths (7 Tesla and above) corresponding to the ultra high frequency (UHF) regime of the RF pulse. At these frequencies, wavelength effects and complex interactions with biological tissue become problematic leading to field inhomogeneity artifacts and tissue heating concerns quantified by the specific absorption rate (SAR). This dissertation will focus on the excitation portion of the imaging process with parallel transmission (pTx) that involves using a transmit RF coil with multiple independent transmit channels. pTx is an effective way to address the challenges of ultra high field MRI through optimization of the transmitted pulse in a patient-specific way. We introduce the Iterative Minimization Procedure with Uncompressed Local SAR Estimate (IMPULSE) which is a novel distributed optimization algorithm that has favorable scaling properties and eliminates the need for virtual observation points (VOPs) thus resulting in superior SAR performance and shorter computation time. The optimization problem is to minimize SAR over a pulse sequence consisting of multiple slice excitations while ensuring that the flip angle inhomogeneity (FAI) for each excited slice is within some user specified tolerance. IMPULSE uses the alternating direction method of mulitpliers (ADMM) to split the optimization into two subproblems, a SAR-update and a FAI-update, that are solved at each iteration until convergence. The SAR-update can be formulated as an unconstrained minimization of a piecewise quadratic function which can be solved efficiently by using a bundle method to build a piecewise linear surrogate that can easily be minimized. The computation time for the FAI-update can be reduced by exploiting parallelization and using an efficient algorithm for projection of a point onto an ellipsoid. IMPULSE achieves superior SAR performance and reduced computation time compared to a conventional approach using virtual observation points or compared to using a generic sequential quadratic programming (SQP) solver in MATLAB. Using the Duke head model consisting of over six million voxels, minimum SAR pTx pulses were designed for 120 slices within 45 seconds with an FAI tolerance of 5\% at each slice. IMPULSE combined with variable rate selective excitation (VERSE) can also be used to improve SAR performance and reduce computation time for simultaneous multislice (SMS) excitation with a pTx-SMS pulse. This method (IMPULSE-SMS) was used for the pTx-SMS task of the ISMRM RF Pulse Design competition in 2016 and resulted in a pulse that was about 20\% shorter than the second best submission and about 10 times shorter than a conventional approach (SAR-unaware pulse design without VERSE). Increasing the number of transmit channels in a coil can give more degrees of freedom to achieve flip angle uniformity and reduce SAR but also increases cost and complexity of the hardware. Studying the performance of massively parallel transmit arrays in simulation can help determine whether investment in these arrays is justified based on new applications that are enabled. An 84 channel loop array for 10.5T with 6 rows and 14 columns was simulated using the Ella body model and applied to two novel applications: power independent of number of slices (PINS) pulses combined with pTx for SMS excitation and SAR focusing for therapeutic hyperthermia. Using this coil in addition to an insertable head gradient (slew rate of 1500 T/m/s), a pulse duration of about 13ms for a 16 slice coronal excitation with 0.4mm slice thickness with 10\% FAI was achieved. SAR focusing is possible for a range of locations throughout the head (although focusing is better at the periphery than at the center). A solution to a simplified bioheat equation indicates that achievable temperature rise would be within acceptable range for some forms of hyperthermia (but not high enough to achieve for ablation). A significant concern in SAR-aware pTx is mismatch between the patient and the tissue model used for SAR estimation since running the optimization on a mismatched model can result in significantly higher SAR compared to a perfect match. One technique to introduce robustness to this mismatch is to use the SAR terms for voxels of multiple tissue models (rather than a single model) in the cost function of IMPULSE. Results indicate that using multiple poorly matched models can achieve similar SAR performance compared to using a single closely matched model indicating that the multiple model approach is a way to get by with a sparse model library that doesn't fully represent the entire human population. A more sophisticated approach is to use deep learning to predict the 3D SAR maps from measured magnetic field maps. An initial implementation of this concept shows promise but is still inconclusive.
Author: Ziwu Zhou
Publisher:
ISBN:
Category :
Languages : en
Pages : 180
Get Book Here
Book Description
Magnetic resonance imaging (MRI) is a powerful diagnostic medical imaging technique that provides very high spatial resolution. By manipulating the signal evolution through careful imaging sequence design, MRI can generate a wide range of soft-tissue contrast unique to individual application. However, imaging speed remains an issue for many applications. In order to increase scan output without compromising the image quality, the data acquisition and image reconstruction methods need to be designed to fit each application to achieve maximum efficiency. This dissertation concerns several application-tailored accelerated imaging methods through improved sequence design, efficient k-space traverse, as well as tailored image reconstruction algorithm, all together aiming to exploit the full potential of data acquisition and image reconstruction in each application. The first application is ferumoxtyol-enhanced 4D multi-phase cardiovascular MRI on pediatric patients with congenital heart disease. By taking advantage of the high signal-to-noise ratio (SNR) results from contrast enhancement, we introduced two methods to improve the scan efficiency with maintained clinical utility: one with reduced scan time and one with improved temporal resolution. The first method used prospective Poisson-disc under-sampling in combination with graphics processing unit accelerated parallel imaging and compressed sensing combined reconstruction algorithm to reduce scan time by approximately 50% while maintaining highly comparable image quality to un-accelerated acquisition in a clinically practical reconstruction time. The second method utilized a motion weighted reconstruction technique to increase temporal resolution of acquired data, and thus permits improved cardiac functional assessment. Compared with existing acceleration method, the proposed method has nearly three times lower computation burden and six times faster reconstruction speed, all with equal image quality. The second application is noncontrast-enhanced 4D intracranial MR angiography with arterial spin labeling (ASL). Considering the inherently low SNR of ASL signal, we proposed to sample k-space with the efficient golden-angle stack-of-stars trajectory and reconstruct images using compressed sensing with magnitude subtraction as regularization. The acquisition and reconstruction strategy in combination produces images with detailed vascular structures and clean background. At the same time, it allows a reduced temporal blurring delineation of the fine distal arteries when compared with the conventional k-space weighted image contrast (KWIC) reconstruction. Stands upon on this, we further developed an improved stack-of-stars radial sampling strategy for reducing streaking artifacts in general volumetric MRI. By rotating the radial spokes in a golden angle manner along the partition-encoding direction, the aliasing pattern due to under-sampling is modified, resulting in improved image quality for gridding and more advanced reconstruction methods. The third application is low-latency real-time imaging. To achieve sufficient frame rate, real-time MRI typically requires significant k-space under-sampling to accelerate the data acquisition. At the same time, many real-time application, such as interventional MRI, requires user interaction or decision making based on image feedback. Therefore, low-latency on-the-fly reconstruction is highly desirable. We proposed a parallel imaging and convolutional neural network combined image reconstruction framework for low-latency and high quality reconstruction. This is achieved by compacting gradient descent steps resolved from conventional parallel imaging reconstruction as network layers and interleaved with convolutional layers in a general convolutional neural network. Once all parameters of the network are determined during the off-line training process, it can be applied to unseen data with less than 100ms reconstruction time per frame, while more than 1s is usually needed for conventional parallel imaging and compressed sensing combined reconstruction.
Author: Mark Murphy
Publisher:
ISBN:
Category :
Languages : en
Pages : 250
Get Book Here
Book Description
Magnetic Resonance Imaging (MRI) is a non-invasive and highly flexible medical imaging modality that does not expose patients ionizing radiation. MR Image acquisitions can be designed by varying a large number of contrast-generation parameters, and many clinical diagnostic applications exist. However, imaging speed is a fundamental limitation to many potential applications. Traditionally, MRI data have been collected at Nyquist sampling rates to produce alias-free images. However, many recent scan acceleration techniques produce sub-Nyquist samplings. For example, Parallel Imaging is a well-established acceleration technique that receives the MR signal simultaneously from multiple receive channels. Compressed sensing leverages randomized undersampling and the compressibility (e.g. via Wavelet transforms or Total-Variation) of medical images to allow more aggressive undersampling. Reconstruction of clinically viable images from these highly accelerated acquisitions requires powerful, usually iterative algorithms. Non-Cartesian pulse sequences that perform non-equispaced sampling of k-space further increase computational intensity of reconstruction, as they preclude direct use of the Fast Fourier Transform (FFT). Most iterative algorithms can be understood by considering the MRI reconstruction as an inverse problem, where measurements of un-observable parameters are made via an observation function that models the acquisition process. Traditional direct reconstruction methods attempt to invert this observation function, whereas iterative methods require its repeated computation and computation of its adjoint. As a result, na\"ive sequential implementations of iterative reconstructions produce unfeasibly long runtimes. Their computational intensity is a substantial barrier to their adoption in clinical MRI practice. A powerful new family of massively parallel microprocessor architectures has emerged simultaneously with the development of these new reconstruction techniques. Due to fundamental limitations in silicon fabrication technology, sequential microprocessors reached the power-dissipation limits of commodity cooling systems in the early 2000's. The techniques used by processor architects to extract instruction-level parallelism from sequential programs face ever-diminishing returns, and further performance improvement of sequential processors via increasing clock-frequency has become impractical. However, circuit density and process feature sizes still improve at Moore's Law rates. With every generation of silicon fabrication technology, a larger number of transistors are available to system architects. Consequently, all microprocessor vendors now exclusively produce multi-core parallel processors. Additionally, the move towards on-chip parallelism has allowed processor architects a larger degree of freedom in the design of multi-threaded pipelines and memory hierarchies. Many of the inefficiencies inherent in superscalar out-of-order design are being replaced by the high efficiency afforded by throughput-oriented designs. The move towards on-chip parallelism has resulted in a vast increase in the amount of computational power available in commodity systems. However, this move has also shifted the burden of computational performance towards software developers. In particular, the highly efficient implementation of MRI reconstructions on these systems requires manual parallelization and optimization. Thus, while ubiquitous parallelism provides a solution to the computational intensity of iterative MRI reconstructions, it also poses a substantial software productivity challenge. In this thesis, we propose that a principled approach to the design and implementation of reconstruction algorithms can ameliorate this software productivity issue. We draw much inspiration from developments in the field of computational science, which has faced similar parallelization and software development challenges for several decades. We propose a Software Architecture for the implementation of reconstruction algorithms, which composes two Design Patterns that originated in the domain of massively parallel scientific computing. This architecture allows for the most computationally intense operations performed by MRI reconstructions to be implemented as re-usable libraries. Thus the software development effort required to produce highly efficient and heavily optimized implementations of these operations can be amortized over many different reconstruction systems. Additionally, the architecture prescribes several different strategies for mapping reconstruction algorithms onto parallel processors, easing the burden of parallelization. We describe the implementation of a complete reconstruction, $\ell_1$-SPIRiT, according to these strategies. $\ell_1$-SPIRiT is a general reconstruction framework that seamlessly integrates all three of the scan acceleration techniques mentioned above. Our implementation achieves substantial performance improvement over baseline, and has enabled substantial clinical evaluation of its approach to combining Parallel Imaging and Compressive Sensing. Additionally, we include an in-depth description of the performance optimization of the non-uniform Fast Fourier Transform (nuFFT), an operation used in all non-Cartesian reconstructions. This discussion complements well our description of $\ell_1$-SPIRiT, which we have only implemented for Cartesian samplings.
Author: Lei Hou Hamilton
Publisher:
ISBN:
Category : Diagnostic imaging
Languages : en
Pages :
Get Book Here
Book Description
Imaging speed is very important in magnetic resonance imaging (MRI), especially in dynamic cardiac applications, which involve respiratory motion and heart motion. With the introduction of reduced-data MR imaging methods, increasing acquisition speed has become possible without requiring a higher gradient system. But these reduced-data imaging methods carry a price for higher imaging speed. This may be a signal-to-noise ratio (SNR) penalty, reduced resolution, or a combination of both. Many methods sacrifice edge information in favor of SNR gain, which is not preferable for applications which require accurate detection of myocardial boundaries. The central goal of this thesis is to develop novel reduced-data imaging methods to improve reconstructed image performance. This thesis presents a novel reduced-data imaging method, PINOT (Parallel Imaging and NOquist in Tandem), to accelerate MR imaging. As illustrated by a variety of computer simulated and real cardiac MRI data experiments, PINOT preserves the edge details, with flexibility of improving SNR by regularization. Another contribution is to exploit the data redundancy from parallel imaging, rFOV and partial Fourier methods. A Gerchberg Reduced Iterative System (GRIS), implemented with the Gerchberg-Papoulis (GP) iterative algorithm is introduced. Under the GRIS, which utilizes a temporal band-limitation constraint in the image reconstruction, a variant of Noquist called iterative implementation iNoquist (iterative Noquist) is proposed. Utilizing a different source of prior information, first combining iNoquist and Partial Fourier technique (phase-constrained iNoquist) and further integrating with parallel imaging methods (PINOT-GRIS) are presented to achieve additional acceleration gains.
Author: Ashish Raj
Publisher:
ISBN:
Category :
Languages : en
Pages : 302
Get Book Here
Book Description
Author:
Publisher:
ISBN:
Category : Dissertations, Academic
Languages : en
Pages : 1006
Get Book Here
Book Description
Author: Curtis N. Wiens
Publisher:
ISBN:
Category :
Languages : en
Pages : 250
Get Book Here
Book Description
Chemical shift imaging is a magnetic resonance imaging technique that separates the signal from two or more chemical species. The cost of chemical shift encoding is increased acquisition time as multiple acquisitions are required at different echo times. Image accelera tion techniques, typically parallel imaging, are often used to improve coverage and resolution. This thesis describes a new technique for estimating the signal to noise ratio for parallel imaging reconstruction s and proposes new image reconstructions for a ccelerated chemical shift imaging using compressed sensing and/or parallel imaging for two applications: water- at separation and metabolic imaging of hyperpolarized [1-13C] pyruvate. Spatially varying noise in parallel imaging reconstructions makes measurements of the signal to noise ratio, a commonly used metric for image for image quality, difficult. Existing approaches have limitations: they are not applicable to all reconstructions, require significant computation time, or rely on repeated image acquisitions. A signal to noise ratio estimation technique is proposed that does not exhibit these limitations. Water-fat imaging of highly undersampled datasets from the liver, calf, knee, and abdominal cavity are demonstrated using a customized IDEAL-SPGR pulse sequence and an integrated compressed sensing, parallel imaging, water-fat reconstruction. This method offer s image quality comparable to fully sampled reference images for a range of acceleration factors. At high acceleration factors, this method offers improved image quality when compared to the current standard of parallel imaging. Accelerated metabolic imaging of hyperpolarized [1-13C] pyruvate and its metabolic by-products lactate, alanine, and bicarbonate is demonstrated using an integrated compressed sensing, metabolite separation reconstruction. Phantoms are used to validate this technique while retrospectively and prospectively accelerated 3D in vivo datasets are used to demonstrate feasibility. An alternative approach to accelerated metabolic imaging is demonstrated using high performance magnetic field gradient set. This thesis addresses the inherently slow acquisition times of chemical shift imaging by examining the role compressed sensing and parallel imaging can play in chemical shift imaging. An approach to SNR assessment for parallel imaging reconstruction is proposed and approaches to accelerated chemical shift imaging are described for applications in water-fat imaging and metabolic imaging of hyperpolarized [1-13C] pyruvate.
Author: Zhongnan Fang
Publisher:
ISBN:
Category :
Languages : en
Pages :
Get Book Here
Book Description
Functional magnetic resonance imaging (fMRI) is a technique that enables non-invasive monitoring of brain activity by detecting changes in blood oxygenation levels. With recent advancements in high performance computing and MRI hardware, real-time fMRI has become possible and the spatiotemporal resolution of fMRI has been significantly improved. However, there are still many challenges for fMRI to achieve its full potential. First, because many basic real-time fMRI modules still uses a large portion of the available processing time, there is insufficient time for the integration of advanced real-time fMRI techniques. Second, current high-resolution fMRI techniques do not provide the resolution needed for imaging activity of small but critical brain regions, such as cortical layers and hippocampal sub-regions. Third, it is still not trivial to achieve the high-resolution and real-time fMRI at once because significant higher computation power is needed. To address these challenges, three projects were conducted and illustrated in this dissertation. In the first project, a high-throughput real-time fMRI system is designed on the graphics processing unit (GPU) to overcome computation barriers associated with reconstruction of non-uniformly sampled image, motion correction and statistical analysis. This system achieves an overall processing speed of 15.01 ms per 3D image, which is more than 49-fold faster than widely used software packages. The high processing speed also enables sliding window reconstruction, which improves the temporal resolution. With this ultra high speed fMRI system, integration of CS reconstruction for real-time and high spatiotemporal resolution fMRI becomes possible. The second project explores the feasibility of CS fMRI and demonstrates a High SPAtial Resolution compressed SEnsing (HSPARSE) fMRI method. HSPARSE fMRI enables a 6-fold spatial resolution improvement with contrast to noise ratio (CNR) increase and no loss of temporal resolution. A novel randomly under-sampled, variable density spiral data acquisition trajectory is designed to achieve an imaging speed acceleration factor of 5.3, which is 32 \% higher than previously reported CS fMRI methods. HSPARSE fMRI also achieves high sensitivity and low false positive rate. Importantly, its high spatial resolution enables localization of brain regions that cannot be resolved using the highest spatial resolution fully-sampled reconstruction. The third project combines the methods in the previous two into a real-time high-resolution CS fMRI system. A random stack of variable density spiral trajectory is first designed to achieve highly incoherent CS sampling and 3.2 times imaging speed acceleration. An optimized CS reconstruction algorithm using wavelet regularization is then implemented on GPU, which achieves a reconstruction speed of 605 ms per 3D image. This method also achieves a 4-fold spatial resolution improvement, with increased CNR, high sensitivity, low false positive rate and no loss of temporal resolution. Notably, this is the first system that achieves the real-time 3D non-uniformly sampled image CS fMRI reconstruction.
Author: Daniel (Daniel Stuart) Weller
Publisher:
ISBN:
Category :
Languages : en
Pages : 148
Get Book Here
Book Description
Magnetic resonance imaging (MRI) is an increasingly versatile diagnostic tool for a variety of medical purposes. During a conventional MRI scan, samples are acquired along a trajectory in the spatial Fourier transform domain (called k-space) and the image is reconstructed using an inverse discrete Fourier transform. The affordability, availability, and applications of MRI remain limited by the time required to sample enough points of k-space for the desired field of view (FOV), resolution, and signal-to-noise ratio (SNR). GRAPPA, an accelerated parallel imaging method, and compressed sensing (CS) have been successfully employed to accelerate the acquisition process by reducing the number of k-space samples required. GRAPPA leverages the different spatial weightings of each receiver coil to undo the aliasing from the reduction in FOV induced by undersampling k-space. However, accelerated parallel imaging reconstruction methods like GRAPPA amplify the noise present in the data, reducing the SNR by a factor greater than that due to only the level of undersampling. Completely separate from accelerated parallel imaging, which capitalizes on observing data with multiple receivers, CS leverages the sparsity of the object along with incoherent sampling and nonlinear reconstruction algorithms to recover the image from fewer samples. In contrast to parallel imaging, CS actually denoises the result, because noise typically is not sparse. When reconstructing brain images, the discrete wavelet transform and finite differences are effective in producing an approximately sparse representation of the image. Because parallel imaging utilizes the multiple receiver coils and CS takes advantage of the sparsity of the image itself, these methods are complementary, and a combination of these methods would be expected to enable further acceleration beyond what is achievable using parallel imaging or CS alone. This thesis investigates three approaches to leveraging both multiple receiver coils and image sparsity. The first approach involves an optimization framework for jointly optimizing the fidelity to the GRAPPA result and the sparsity of the image. This technique operates in the nullspace of the data observation matrix, preserving the acquired data without resorting to techniques for constrained optimization. While this framework is presented generally, the effectiveness of the implementation depends on the choice of sparsifying transform, sparsity penalty function, and undersampling pattern. The second approach involves modifying the kernel estimation step of GRAPPA to promote sparsity in the reconstructed image and mitigate the noise amplification typically encountered with parallel imaging. The third approach involves imposing a sparsity prior on the coil images and estimating the full k-space from the observations using Bayesian techniques. This third method is extended to jointly estimate the GRAPPA kernel weights and the full k-space together. These approaches represent different frameworks for accelerating MRI imaging beyond current methods. The results presented suggest that these practical reconstruction and post-processing methods allow for greater acceleration with conventional Cartesian acquisitions.
