Dynamically adaptive MRI with encoding by singular value decomposition

A new MRI spatial encoding method based upon the singular value decomposition (SVD) and using spatially selective RF excitation is described. This encoding technique is particularly; applicable to dynamic adaptive MRI, because it provides a near minimal set of spatial encoding profiles computed using an image estimate that is determined from a previously obtained image. Experimental results are presented for two cases, which exemplify its potential use in different dynamic imaging tasks. SVD-encoded MRI has demonstrated to be a highly efficient encoding scheme.     

Adaptive keyhole methods for dynamic magnetic resonance image reconstruction.

Dynamic magnetic resonance imaging (MRI) acquires a sequence of images for the visualization of the temporal variation of tissue or organs. Keyhole methods such as Fourier keyhole (FK) and keyhole SVD (KSVD) are the most popular methods for image reconstruction in dynamic MRI. This paper provides a class of adaptive keyhole methods, called adaptive FK (AFK) and adaptive KSVD (AKSVD), for dynamic MRI reconstruction. The proposed methods are based on the conventional Fourier encoding and SVD encoding schemes. Instead of the conventional keyhole methods' duplication of un-acquired data from the reference images, the proposed methods use a temporal model to depict the inter-frame dynamic changes and to estimate the un-acquired data in each successive frame. Because the model is online identified from the acquired data, the proposed methods do not require the pre-imaging process, the navigator signals, and any prior knowledge of the imaged objects. Furthermore, the new methods use the conventional keyhole encoding schemes without the bias to any particular object characters, and the temporal model for updating information is in the general form of AR process without the preference to any particular motion types. Hence, the proposed methods are designed as a generic approach to dynamic MRI, other than for any specific class of objects. Studies on dynamic MRI data set show that the new methods can produce images with much lower reconstruction error than the conventional FK and KSVD.     

Noquist: reduced field-of-view imaging by direct Fourier inversion.

A novel technique called "Noquist" is introduced for the acceleration of dynamic cardiac magnetic resonance imaging (CMRI). With the use of this technique, a more sparsely sampled dynamic image sequence is reconstructed correctly, without Nyquist foldover artifact. Unlike most other reduced field-of-view (rFOV) methods, Noquist does not rely on data substitution or temporal interpolation to reconstruct the dynamic image sequence. The proposed method reduces acquisition time in dynamic MRI scans by eliminating the data redundancy associated with static regions in the dynamic scene. A reduction of imaging time is achieved by a fraction asymptotically equal to the static fraction of the FOV, by omitting acquisition of an appropriate subset of phase-encoding views from a conventional equidistant Cartesian acquisition grid. The theory behind this method is presented along with sample reconstructions from real and simulated data. Noquist is compared with conventional cine imaging by retrospective selection of a reduced data set from a full-grid conventional image sequence. In addition, a comparison is presented, using real and simulated data, of our technique with an existing rFOV technique that uses temporal interpolation. The experimental results confirm the theory, and demonstrate that Noquist reduces scan time for cine MRI while fully preserving both spatial and temporal resolution, but at the cost of a reduced signal-to-noise ratio (SNR).     

Large field-of-view real-time MRI with a 32-channel system.

The emergence of parallel MRI techniques and new applications for real-time interactive MRI underscores the need to evaluate performance gained by increasing the capability of MRI phased-array systems beyond the standard four to eight high-bandwidth channels. Therefore, to explore the advantages of highly parallel MRI a 32-channel 1.5 T MRI system and 32-element torso phased arrays were designed and constructed for real-time interactive MRI. The system was assembled from multiple synchronized scanner-receiver subsystems. Software was developed to coordinate across subsystems the real-time acquisition, reconstruction, and display of 32-channel images. Real-time, large field-of-view (FOV) body-survey imaging was performed using interleaved echo-planar and single-shot fast-spin-echo pulse sequences. A new method is demonstrated for augmenting parallel image acquisition by independently offsetting the frequency of different array elements (FASSET) to variably shift their FOV. When combined with conventional parallel imaging techniques, image acceleration factors of up to 4 were investigated. The use of a large number of coils allowed the FOV to be doubled in two dimensions during rapid imaging, with no degradation of imaging time or spatial resolution. The system provides a platform for evaluating the applications of many-channel real-time MRI, and for understanding the factors that optimize the choice of array size.     

k‐t PCA: Temporally constrained k‐t BLAST reconstruction using principal component analysis

The k‐t broad‐use linear acquisition speed‐up technique (BLAST) has become widespread for reducing image acquisition time in dynamic MRI. In its basic form k‐t BLAST speeds up the data acquisition by undersampling k‐space over time (referred to as k‐t space). The resulting aliasing is resolved in the Fourier reciprocal x‐f space (x = spatial position, f = temporal frequency) using an adaptive filter derived from a low‐resolution estimate of the signal covariance. However, this filtering process tends to increase the reconstruction error or lower the achievable acceleration factor. This is problematic in applications exhibiting a broad range of temporal frequencies such as free‐breathing myocardial perfusion imaging. We show that temporal basis functions calculated by subjecting the training data to principal component analysis (PCA) can be used to constrain the reconstruction such that the temporal resolution is improved. The presented method is called k‐t PCA. Magn Reson Med, 2009. © 2009 Wiley‐Liss, Inc.     

Using UNFOLD to remove artifacts in parallel imaging and in partial-Fourier imaging.

In dynamic MRI, it is often difficult to achieve the acquisition speed required to resolve or freeze the temporal variations of the imaged object. Several MRI methods aim at speeding up the image acquisition process. Through assumptions and/or prior knowledge, these dynamic MRI methods allow part of the needed data to be calculated instead of acquired. For example, partial-Fourier imaging assumes that phase varies smoothly within the object, and parallel imaging (e.g., simultaneous acquisition of spatial harmonics (SMASH) and sensitivity encoding (SENSE)) uses prior knowledge about receiver-coil sensitivity. While these methods accelerate acquisition, they can introduce artifacts or amplify noise in doing so. The present work aims at accelerating image acquisition significantly, while introducing almost no artifacts or noise amplification. It is shown here that new, extra information is gained if dynamic MRI methods are modified so that the sampling function changes in specific ways from time-frame to time-frame. In other words, the set of k-space locations that are acquired (instead of calculated) changes with time. The present temporal strategy, based on the UNaliasing by Fourier-encoding the Overlaps in the temporaL Dimension (UNFOLD) method, can be incorporated into common dynamic MRI methods. Results with partial-Fourier, SMASH, and SENSE imaging are presented here, where UNFOLD's contribution is to very significantly reduce the artifact and/or amplified noise content. Used in this way, UNFOLD contributes indirectly, rather than directly to the improvement in image acquisition speed, as it allows companion methods to operate properly at greater acceleration settings than would otherwise be feasible.     

Adaptive sensitivity encoding incorporating temporal filtering (TSENSE).

A number of different methods have been demonstrated which increase the speed of MR acquisition by decreasing the number of sequential phase encodes. The UNFOLD technique is based on time interleaving of k-space lines in sequential images and exploits the property that the outer portion of the field-of-view is relatively static. The differences in spatial sensitivity of multiple receiver coils may be exploited using SENSE or SMASH techniques to eliminate the aliased component that results from undersampling k-space. In this article, an adaptive method of sensitivity encoding is presented which incorporates both spatial and temporal filtering. Temporal filtering and spatial encoding may be combined by acquiring phase encodes in an interleaved manner. In this way the aliased components are alternating phase. The SENSE formulation is not altered by the phase of the alias artifact; however, for imperfect estimates of coil sensitivities the residual artifact will have alternating phase using this approach. This is the essence of combining temporal filtering (UNFOLD) with spatial sensitivity encoding (SENSE). Any residual artifact will be temporally frequency-shifted to the band edge and thus may be further suppressed by temporal low-pass filtering. By combining both temporal and spatial filtering a high degree of alias artifact rejection may be achieved with less stringent requirements on accuracy of coil sensitivity estimates and temporal low-pass filter selectivity than would be required using each method individually. Experimental results that demonstrate the adaptive spatiotemporal filtering method (adaptive TSENSE) with acceleration factor R = 2, for real-time nonbreath-held cardiac MR imaging during exercise induced stress are presented.     

MRI estimation of contrast agent concentration in tissue using a neural network approach.

Using an MRI T(1) by multiple readout pulses (TOMROP) image set, an adaptive neural network (ANN) was trained to directly estimate the concentration of a contrast agent (CA), gadolinium-bovine serum albumin (Gd-BSA), in tissue. In nine rats implanted with a 9L cerebral tumor, MRI acquisition of TOMROP inversion-recovery data was followed by quantitative autoradiography (QAR) using radioiodinated serum albumin (RISA). QAR autoradiograms were used as a training set for the ANN. Precontrast and 25 min postcontrast TOMROP image sets were shown to the ANN in the form of a physical feature set related to 24 inversion-recovery images; QAR autoradiograms at 30 min after injection of RISA were taken as the training standard for the network. After training and optimization, the ANN produced a map of Gd-BSA concentration [g-moles/liter]. The prediction by the ANN of CA concentration at 25 min after injection was well correlated (r = 0.82, P < 0.0001) with the corresponding autoradiogram's measure of CA concentration.     

Self-navigated motion correction using moments of spatial projections in radial MRI.

Interest in radial MRI (also known as projection reconstruction (PR) MRI) has increased recently for uses such as fast scanning and undersampled acquisitions. Additionally, PR acquisitions offer intrinsic advantages over standard two-dimensional Fourier transform (2DFT) imaging with respect to motion of the imaged object. It is well known that aligning each spatial domain projection's center of mass (calculated using the 0th and 1st moments) to the center of the field of view (FOV) corrects shifts caused by in-plane translation. In this work, a previously unrealized ability to determine the in-plane rotational motion of an imaged object using the 2nd moments of the spatial domain projections in conjunction with a specific projection angle acquisition time order is reported. We performed the correction using only the PR data itself acquired with the newly proposed projection angle acquisition time order. With the proposed view angle acquisition order, the acquisition is "self-navigating" with respect to both in-plane translation and rotation. We reconstructed the images using the aligned projections and detected acquisition angles to significantly reduce image artifacts due to such motion. The theory of the correction technique is described, and its effectiveness is demonstrated in phantom and in vivo experiments.     

Motionless Movies of Myocardial Strain-Rates using Stimulated Echoes

We present methods to acquire and analyze NMR movies of myocardial strain rates in which cardiac motion is suppressed and the histories of strain rates are accurately defined for each voxel of myocardial tissue. By means of stimulated echoes, the myocardial strain-rate tensor is phase-encoded at progressive delays in the cardiac cycle while the slice-select and spatial encoding of the image acquisition are performed at a constant cardiac delay. In these data, every image shows the identical myocardial tissue, and the anatomic configuration of the heart appears motionless. The myocardial strainrate data, however, indicate the state of motion which existed in this slice at the time of the velocity phase-encoding, and these data evolve with the progressive delay as a movie. Using echo-planar MRI, motionless movies of myocardial strain rate of four to eight cardiac delays are obtained in a breath-hold. As an application, a quantitative characterization of cardiac mechanical synchrony is accomplished by principal component analysis (PCA) of the time series of strain rates.     

Effect of acquisition parameters on the accuracy of velocity encoded cine magnetic resonance imaging blood flow measurements.

PURPOSE: To investigate the effect of acquisition parameters on the accuracy of 2D velocity encoded cine magnetic resonance imaging (VEC MRI) flow measurements. MATERIALS AND METHODS: Using a pulsatile flow phantom, through-plane flow measurements were performed on a flexible vessel made of polyvinyl alcohol cryogel (PVA), a material that mimics the MR signal and biomechanical properties of aortic tissue. RESULTS: Repeated VEC MRI flow measurements (N = 20) under baseline conditions yielded an error of 0.8 +/- 1.5%. Slice thickness, angle between flow and velocity encoding directions, spatial resolution, velocity encoding range, and radio frequency (RF) flip angles were varied over a clinically relevant range. Spatial resolution had the greatest impact on accuracy, with a 9% overestimation of flow at 16 pixels per vessel cross-section. CONCLUSION: VEC MRI proved to be an accurate and reproducible technique for pulsatile flow measurements over the range of acquisition parameters examined as long as sufficient spatial resolution was prescribed.     

A dynamically adaptive imaging algorithm for wavelet-encoded MRI

A new adaptive algorithm based on wavelet-encoded MRI is presented for application in dynamic imaging. This algorithm is adaptive because the strategy for updating image data in the dynamic series of images is determined by the processing of the most recently acquired data. The spatially selective multi-resolution properties of the wavelet transform are exploited to selectively update only those regions of the field of view where change is actually occurring. A theoretical imaging model is presented to motivate use of the adaptive algorithm, and simulation results using both artificial and experimental wavelet-encoded data are presented.     

Parallel MRI reconstruction using variance partitioning regularization.

Multiple receivers can be utilized to enhance the spatiotemporal resolution of MRI by employing the parallel imaging technique. Previously, we have reported the L-curve Tikhonov regularization technique to mitigate noise amplification resulting from the geometrical correlations between channels in a coil array. Nevertheless, one major disadvantage of regularized image reconstruction is lengthy computational time in regularization parameter estimation. At a fixed noise level, L-curve regularization parameter estimation was also found not to be robust across repetitive measurements, particularly for low signal-to-noise ratio (SNR) acquisitions. Here we report a computationally efficient and robust method to estimate the regularization parameter by partitioning the variance of the noise-whitened encoding matrix based on the estimated SNR of the aliased pixel set in parallel MRI data. The proposed Variance Partitioning Regularization (VPR) method can improve computational efficiency by 2-5-fold, depending on image matrix sizes and acceleration rates. Our anatomical and functional MRI results show that the VPR method can be applied to both static and dynamic MRI experiments to suppress noise amplification in parallel MRI reconstructions for improved image quality.     

Method for rapid MRI needle tracking.

A new method for MRI needle tracking within a given two-dimensional (2D) image slice is presented. The method is based on k-space investigation of the difference image between the current dynamic frame and a reference frame. Using only a few central k-lines of the difference image and a nonlinear optimization procedure, one can resolve the parameters that define the 2D sinc function that best characterizes the needle in k-space. The spatial location and orientation of the needle are determined from these parameters. Rapid needle tracking is obtained by repeated acquisitions of the same set of several central k-lines (as in a "keyhole" protocol) and repeated computation of these parameters. The calculated needle tip is depicted on the reference image by means of a graphic overlay. The procedure was tested in computer simulations and in actual MRI scans (the computations were done offline). It was demonstrated that six k-lines out of 128 usually suffice to locate the needle. The refresh rate of the needle location depends on the time required to sample the subset of k-lines, calculate the current needle location, and refresh the reference image.     

On the Relationship Between Feature-Recognizing MRI and MRI Encoded by Singular Value Decomposition

This paper describes the similarity between two methods of non-Fourier MRI: feature-recognizing MRI (FR MRI) and MRI with encoding by singular value decomposition (SVD MRI). Both methods represented images as truncated expansions of non-Fourier basis functions; these basis images were derived from prior image data by using closely-related mathematical techniques: the Karhunen-Loeve decomposition (or principal components analysis) and singular value decomposition, respectively. We demonstrate that FR and SVD MRI are equivalent in the following sense: given the same prior image data, they lead to exactly the same basis functions. FR MRI utilized prior images of the same body part in many “training” subjects, thought to be similar to the “unknown” subject to be imaged. SVD MRI utilized a single prior image of one subject in order to perform dynamic imaging of that subject. We demonstrate that the basis function expansion derived from a single prior image may not be capable of representing new features (features not found in the prior image). Therefore, the SVD basis functions may be inappropriate for dynamic imaging.     

Dynamic MRI with projection reconstruction and KWIC processing for simultaneous high spatial and temporal resolution.

A method for dynamic imaging in MRI is presented that enables the acquisition of a series of images with both high temporal and high spatial resolution. The technique, which is based on the projection reconstruction (PR) imaging scheme, utilizes distinct data acquisition and reconstruction strategies to achieve this simultaneous capability. First, during acquisition, data are collected in multiple undersampled passes, with the view angles interleaved in such a way that those of subsequent passes bisect the views of earlier ones. During reconstruction, these views are weighted according to a previously described k-space weighted image contrast (KWIC) technique that enables the manipulation of image contrast by selective filtering. Unlike conventional undersampled PR methods, the proposed dynamic KWIC technique does not suffer from low image SNR or image degradation due to streaking artifacts. The effectiveness of dynamic KWIC is demonstrated in both simulations and in vivo, high-resolution, contrast-enhanced imaging of breast lesions.     

Dynamics of magnetization in hyperpolarized gas MRI of the lung

The magnetization in hyperpolarized gas (HP) MRI is generated by laser polarization that is independent of the magnet and imaging process. As a consequence, there is no equilibrium magnetization during the image acquisition. The competing processes of gas inflow and depolarization of the spins lead to large changes in signal as one samples k-space. A model is developed of dynamic changes in polarization of hyperpolarized ³He during infusion and in vivo imaging of the lung and verified experimentally in a live guinea pig. Projection encoding is used to measure the view-to-view variation with temporal resolution <4 ms. Large excitation angles effectively sample the magnetization in the early stages of inflow, highlighting larger airways, while smaller excitation angles produce images of the more distal spaces. The work provides a basis for pulse sequences designed to effectively exploit HP MRI in the lung.     

Single-shot,motion insensitive cardiac imaging on a standard clinical system

The overall goal of this study was the development and application of a less motion sensitive, single-shot MRI technique for use on a standard clinical system in a dynamic imaging setting, such as cardiac scanning. Time encoding, a single-shot line scanning technique, has been used to produce single-shot, small field-of-view cardiac images without the use of presaturation pulses. The major advantages of this method are: (1) as a line scanning technique, time encoding is minimally sensitive to motion when compared with 2D Fourier methods, and (2) aliasing will not occur if the object being imaged extends beyond the field of view.     

Adaptive bilateral breast MRI using projection reconstruction time-resolved imaging of contrast kinetics

PURPOSE: To introduce a bilateral implementation of an adaptive imaging technique in which both dynamic and high resolution breast MR images are acquired simultaneously. MATERIALS AND METHODS: Adaptive three-dimensional bilateral breast imaging in the sagittal plane was achieved by combining two elements: a projection reconstruction time-resolved imaging of contrast kinetics (PR-TRICKS) k-space trajectory and a slab interleaved sequence that imaged alternate breasts every TR. A pilot study was performed to evaluate image quality and contrast uptake behavior, using eight patients with previously identified benign lesions. RESULTS: Adaptive reconstruction demonstrated breast lesions in all eight women with similar image quality and signal-to-noise ratio (SNR) to Cartesian images with comparable imaging parameters. Contrast enhancement curves covering the entire postinjection time period were obtained from the dynamic images and in one case compared to previous enhancement profiles from a conventional Cartesian trajectory. CONCLUSION: Bilateral dynamic and high spatial resolution images with high SNR can be achieved in a clinically feasible manner, providing both kinetic and morphologic analysis with a single data set. This may obviate the need for multiple MRI examinations for a thorough breast MRI workup.     

Patient‐adaptive reconstruction and acquisition in dynamic imaging with sensitivity encoding (PARADISE)

MRI of the human heart without explicit cardiac synchronization promises to extend the applicability of cardiac MR to a larger patient population and potentially expand its diagnostic capabilities. However, conventional nongated imaging techniques typically suffer from low image quality or inadequate spatio‐temporal resolution and fidelity. Patient‐Adaptive Reconstruction and Acquisition in Dynamic Imaging with Sensitivity Encoding (PARADISE) is a highly accelerated nongated dynamic imaging method that enables artifact‐free imaging with high spatio‐temporal resolutions by utilizing novel computational techniques to optimize the imaging process. In addition to using parallel imaging, the method gains acceleration from a physiologically driven spatio‐temporal support model; hence, it is doubly accelerated. The support model is patient adaptive, i.e., its geometry depends on dynamics of the imaged slice, e.g., subject's heart rate and heart location within the slice. The proposed method is also doubly adaptive as it adapts both the acquisition and reconstruction schemes. Based on the theory of time‐sequential sampling, the proposed framework explicitly accounts for speed limitations of gradient encoding and provides performance guarantees on achievable image quality. The presented in‐vivo results demonstrate the effectiveness and feasibility of the PARADISE method for high‐resolution nongated cardiac MRI during short breath‐hold. Magn Reson Med, 2010. © 2010 Wiley‐Liss, Inc.