MR phase imaging with bipolar acquisition.

We have previously proposed a novel magnetic resonance (MR) phase imaging framework (MAGPI) based on a three-echo sequence that demonstrated substantial gains in phase signal-to-noise ratio (SNR). We improve upon the performance of MAGPI by extending the formulation to handle (i) an alternating gradient polarity (bipolar) readout scheme and (ii) an arbitrary number of echoes. We formulate the phase-imaging problem using maximum-likelihood (ML) estimation. The acquisition uses an optimized multi-echo gradient echo (MEGE) sequence. The tissue-phase estimation algorithm is a voxel-per-voxel approach, which requires no reference scans, no phase unwrapping and no spatial denoising. Unlike other methods, our bipolar readout model is general and does not make simplifying assumptions about the even-odd echo phase errors. The results show that (a) our proposed bipolar MAGPI approach improves on the phase SNR gains achieved with monopolar MAGPI and (b) the phase SNR converges with the number of echoes more rapidly with bipolar MAGPI. Importantly, bipolar MAGPI enables phase imaging in severely SNR-constrained scenarios, where monopolar MAGPI is unable to find solutions. The substantial phase SNR gains achieved with our framework are used here to (a) accelerate acquisitions (full brain 0.89 mm in-plane resolution in 2 min 30 sec) and (b) enable high-contrast high-resolution phase imaging (310 µm in-plane resolution) at clinical field strengths. Copyright © 2016 John Wiley & Sons, Ltd.

MAGPI: A framework for maximum likelihood MR phase imaging using multiple receive coils.

PURPOSE: Combining MR phase images from multiple receive coils is a challenging problem, complicated by ambiguities introduced by phase wrapping, noise, and the unknown phase-offset between the coils. Various techniques have been proposed to mitigate the effect of these ambiguities but most of the existing methods require additional reference scans and/or use ad hoc post-processing techniques that do not guarantee any optimality. THEORY AND METHODS: Here, the phase estimation problem is formulated rigorously using a maximum-likelihood (ML) approach. The proposed framework jointly designs the acquisition-processing chain: the optimized pulse sequence is a single multiecho gradient echo scan and the corresponding postprocessing algorithm is a voxel-per-voxel ML estimator of the underlying tissue phase. RESULTS: Our proposed framework (Maximum AmbiGuity distance for Phase Imaging, MAGPI) achieves substantial improvements in the phase estimate, resulting in phase signal-to-noise ratio (SNR) gains by up to an order of magnitude compared to existing methods. CONCLUSION: The advantages of MAGPI are: (1) ML-optimal combination of phase data from multiple receive coils, without a reference scan; (2) voxel-per-voxel ML-optimal estimation of the underlying tissue phase, without the need for phase unwrapping or image smoothing; and (3) robust dynamic estimation of channel-dependent phase-offsets.

High-resolution, large dynamic range field map estimation.

PURPOSE: We present a theory and a corresponding method to compute high-resolution field maps over a large dynamic range. THEORY AND METHODS: We derive a closed-form expression for the error in the field map value when computed from two echoes. We formulate an optimization problem to choose three echo times which result in a pair of maximally distinct error distributions. We use standard field mapping sequences at the prescribed echo times. We then design a corresponding estimation algorithm which takes advantage of the optimized echo times to disambiguate the field offset value. RESULTS: We validate our method using high-resolution images of a phantom at 7T. The resulting field maps demonstrate robust mapping over both a large dynamic range, and in low SNR regions. We also present high-resolution offset maps in vivo using both, GRE and multiecho gradient echo sequences. Even though the proposed echo time spacings are larger than the well known phase aliasing cutoff, the resulting field maps exhibit a large dynamic range without the use of phase unwrapping or spatial regularization techniques. CONCLUSION: We demonstrate a novel three-echo field map estimation method which overcomes the traditional noise-dynamic range trade-off.

Towards a barrier-free anthropomorphic brain phantom for quantitative magnetic resonance imaging: Design, first construction attempt, and challenges.

Existing magnetic resonance imaging (MRI) reference objects, or phantoms, are typically constructed from simple liquid or gel solutions in containers with specific geometric configurations to enable multi-year stability. However, there is a need for phantoms that better mimic the human anatomy without barriers between the tissues. Barriers result in regions without MRI signal between the different tissue mimics, which is an artificial image artifact. We created an anatomically representative 3D structure of the brain that mimicked the T1 and T2 relaxation properties of white and gray matter at 3 T. While the goal was to avoid barriers between tissues, the 3D printed barrier between white and gray matter and other flaws in the construction were visible at 3 T. Stability measurements were made using a portable MRI system operating at 64 mT, and T2 relaxation time was stable from 0 to 22 weeks. The phantom T1 relaxation properties did change from 0 to 10 weeks; however, they did not substantially change between 10 weeks and 22 weeks. The anthropomorphic phantom used a dissolvable mold construction method to better mimic anatomy, which worked in small test objects. The construction process, though, had many challenges. We share this work with the hope that the community can build on our experience.

Comparison of Phase Estimation Methods for Quantitative Susceptibility Mapping Using a Rotating-Tube Phantom.

Quantitative Susceptibility Mapping (QSM) is an MRI tool with the potential to reveal pathological changes from magnetic susceptibility measurements. Before phase data can be used to recover susceptibility (Δχ), the QSM process begins with two steps: data acquisition and phase estimation. We assess the performance of these steps, when applied without user intervention, on several variations of a phantom imaging task. We used a rotating-tube phantom with five tubes ranging from Δχ=0.05 ppm to Δχ=0.336 ppm. MRI data was acquired at nine angles of rotation for four different pulse sequences. The images were processed by 10 phase estimation algorithms including Laplacian, region-growing, branch-cut, temporal unwrapping, and maximum-likelihood methods, resulting in approximately 90 different combinations of data acquisition and phase estimation methods. We analyzed errors between measured and expected phases using the probability mass function and Cumulative Distribution Function. Repeatable acquisition and estimation methods were identified based on the probability of relative phase errors. For single-echo GRE and segmented EPI sequences, a region-growing method was most reliable with Pr (relative error <0.1) = 0.95 and 0.90, respectively. For multiecho sequences, a maximum-likelihood method was most reliable with Pr (relative error <0.1) = 0.97. The most repeatable multiecho methods outperformed the most repeatable single-echo methods. We found a wide range of repeatability and reproducibility for off-the-shelf MRI acquisition and phase estimation approaches, and this variability may prevent the techniques from being widely integrated in clinical workflows. The error was dominated in many cases by spatially discontinuous phase unwrapping errors. Any postprocessing applied on erroneous phase estimates, such as QSM's background field removal and dipole inversion, would suffer from error propagation. Our paradigm identifies methods that yield consistent and accurate phase estimates that would ultimately yield consistent and accurate Δχ estimates.

A method for coordinating the distributed transmission of imagery.

Distributed imaging using sensor arrays is gaining popularity among various research and development communities. A common bottleneck within such an imaging sensor network is the large resulting data load. In applications for which transmission power and/or bandwidth are constrained, this can drastically decrease the sensor network lifetime. We present an algorithm that efficiently exploits inter- and intrasensor correlation for the purpose of power-constrained distributed transmission of sensor-network imagery. Gains in network lifetime up to 114% are obtained when using the suggested algorithm with lossless compression. Our results also demonstrate that when lossy compression is employed, much larger gains are achieved. For example, when a normalized root-mean-squared error of 0.78% can be tolerated in the received measurements, the network lifetime increases by a factor of 2.8, as compared to the (optimized) lossless case.

A Joint Acquisition-Estimation Framework for MR Phase Imaging.

Measuring the phase of the MR signal is faced with fundamental challenges such as phase aliasing, noise and unknown offsets of the coil array. There is a paucity of acquisition, reconstruction and estimation methods that rigorously address these challenges. This reduces the reliability of information processing in phase domain. We propose a joint acquisition-processing framework that addresses the challenges of MR phase imaging using a rigorous theoretical treatment. Our proposed solution acquires the multi-coil complex data without any increase in acquisition time. Our corresponding estimation algorithm is applied optimally voxel-per-voxel. Results show that our framework achieves performance gains up to an order of magnitude compared to existing methods.

Resource-constrained rate control for Motion JPEG2000.

With the increasing importance of heterogeneous networks and time-varying communication channels, fine scalability has become a highly desirable feature in both image and video coders. A single highly scalable bitstream can provide precise rate control for constant bitrate (CBR) traffic and accurate quality control for variable bitrate (VBR) traffic. We first propose two leaky-bucket rate allocation methods that provide constant quality video under buffer constraints. These methods can be used with all scalable coders. Moreover, we make use of one of these methods (DBRC) and extend it so that it can be used when multiple sequences are multiplexed over a single communications channel. The goal is to allocate the capacity of the channel between sequences to achieve constant quality across all sequences. Experimental results using Motion JPEG2000 demonstrate substantial benefits.

Efficient and robust estimation of blood oxygenation levels in single cerebral veins.

Blood oxygenation level is an important measure that can be used alongside functional magnetic resonance imaging data in order to obtain closer correlates of neuronal activation. A robust estimate of this measure has thus far not been demonstrated. This is mainly due to the lack of knowledge of the underlying parameters which influence the numerical estimates of blood oxygenation. In this paper, we present a systematic analysis of the estimation performance of venous hemoglobin oxygen saturation [Formula: see text] as a function of noise, physiologic and geometric parameters. Furthermore, we present a novel algorithm for estimating [Formula: see text] from the temporal decay of an MR signal. The proposed algorithm incorporates prior information about the functional dependence between [Formula: see text] and relaxation rates. We compare our algorithm to an existing method in the literature and analyze the estimation performance. We show that our proposed algorithm is more efficient, achieving gains in performance as high as 92 %. We also show how our estimation algorithm takes advantage of signal features that are specific to the underlying physiology and geometry. We argue that optimal acquisition sequences, and corresponding estimation methods, should take into account such features in order to obtain robust estimates of blood oxygenation.

A joint acquisition-reconstruction paradigm for correcting inhomogeneity artifacts in MR echo planar imaging.

One of the main sources of signal degradation in rapid MR acquisitions, such as Echo Planar Imaging (EPI), is magnetic field variations caused by field inhomogeneities and susceptibility gradients. If unaccounted for during the reconstruction process, this spatially-varying field can cause severe image artifacts. In this paper, we show that correcting for the resulting degradations can be formulated as a blind image deconvolution problem. We propose a novel joint acquisition-processing paradigm to solve this problem. We describe a practical implementation of this paradigm using a multi-image acquisition strategy and a corresponding joint estimation-reconstruction algorithm. The estimation step computes the spatial distribution of the field maps, while the reconstruction step yields a Minimum Mean Squared Error (MMSE) estimate of the imaged slice. Our simulations show that this proposed joint acquisition-reconstruction method is robust and efficient, offering factors of improvement in the quality of the reconstructed image as compared to other traditional methods.

Collaborative multihop transmission of distributed sensor imagery.

We consider a network of imaging sensors. We address the problem of energy-efficient communication of the measurements of the sensors. A novel algorithm is presented for the purpose of exploiting intersensor and intrasensor correlation, which is inherent in a network of imaging sensors. The collaborative algorithm is used in conjunction with a cooperative multihop routing strategy to maximize the lifetime of the network. The algorithm is demonstrated to achieve an average gain in the lifetime as high as 3.2 over previous methods.

Efficient storage and transmission of ladar imagery.

We develop novel methods for compressing volumetric imagery that has been generated by single-platform (mobile) range sensors. We exploit the correlation structure inherent in multiple views in order to improve compression efficiency. We show that, for lossless compression, three-dimensional volumes compress more efficiently than two-dimensional (2D) images by a factor of 60%. Furthermore, our error metric for lossy compression suggests that accumulating more than nine range images in one volume before compression yields as much as a 99% improvement in compression performance over 2D compression.