MULTI-parametric MR imaging with fLEXible design (MULTIPLEX).

PURPOSE: To introduce a gradient echo (GRE) -based method, namely MULTIPLEX, for single-scan 3D multi-parametric MRI with high resolution, signal-to-noise ratio (SNR), accuracy, efficiency, and acquisition flexibility. THEORY: With a comprehensive design with dual-repetition time (TR), dual flip angle (FA), multi-echo, and optional flow modulation features, the MULTIPLEX signals contain information on radiofrequency (RF) B1t fields, proton density, T1 , susceptibility and blood flows, facilitating multiple qualitative images and parametric maps. METHODS: MULTIPLEX was evaluated on system phantom and human brains, via visual inspection for image contrasts and quality or quantitative evaluation via simulation, phantom scans and literature comparison. Region-of-interest (ROI) analysis was performed on parametric maps of the system phantom and brain scans, extracting the mean and SD of the T1 , T2∗ , proton density (PD), and/or quantitative susceptibility mapping (QSM) values for comparison with reference values or literature. RESULTS: One MULTIPLEX scan offers multiple sets of images, including but not limited to: composited PDW/T1 W/ T2∗ W, aT1 W, SWI, MRA (optional), B1t map, T1 map, T2∗ / R2∗ maps, PD map, and QSM. The quantitative error of phantom T1 , T2∗ and PD mapping were <5%, and those in brain scans were in good agreement with literature. MULTIPLEX scan times for high resolution (0.68 × 0.68 × 2 mm3 ) whole brain coverage were about 7.5 min, while processing times were <1 min. With flow modulation, additional MRA images can be obtained without affecting the quality or accuracy of other images. CONCLUSION: The proposed MUTLIPLEX method possesses great potential for multi-parametric MR imaging.

Augmented T1 -weighted steady state magnetic resonance imaging.

T1 contrasts obtained using short-TR incoherent steady state gradient echo (GRE) methods are generally suboptimal, to which non-T1 factors in the signals play a major part. In this work, we proposed an augmented T1 -weighted (aT1 W) method to extract the signal ratio between routine GRE T1 W and proton density-weighted signals that effectively removes the non-T1 effects from the original T1 W signals, including proton density, T2 * decay, and coil sensitivity. A recently proposed multidimensional integration (MDI) technique was incorporated in the aT1 W calculation for better signal-to-noise ratio (SNR) performance. For comparison between aT1 W and T1 W results, Monte Carlo noise analysis was performed via simulation and on scanned data, and region-of-interest (ROI) analysis and comparison was performed on the system phantom. For brain scans, the image contrast, noise behavior, and SNR of aT1 W images were compared with routine GRE and inversion-recovery-based T1 W images. The proposed aT1 W method yielded saliently improved T1 contrasts (potentially > 30% higher contrast-to-noise ratio [CNR]) than routine GRE T1 W images. Good spatial homogeneity and signal consistency as well as high SNR/CNR were achieved in aT1 W images using the MDI technique. For contrast-enhanced (CE) imaging, aT1 W offered stronger post-CE contrast and better boundary delineation than T1 MPRAGE images while using a shorter scan time.

A multi-dimensional integration (MDI) strategy for MR T2 * mapping.

MRI signals are intrinsically multi-dimensional, and signal behavior may be orthogonal among different dimensions. Such dimensional orthogonality can be utilized to eliminate unwanted effects and facilitate mathematical simplicity during image processing for improved outcomes. In this work, we will demonstrate and analyze the principles and performance of a newly developed multi-dimensional integration (MDI) strategy in MR T2 * mapping. By constructing a complex signal function to extract the inter-echo signal changes, MDI solves an optimization problem by processing all signal dimensions (eg echoes, flip angles and coil channels) in one integrative step. MDI was compared with routine curve fitting methods on noise behavior, quantification accuracy and computational efficiency. All methods were tested and compared on simulation, phantom and knee data. Monte Carlo simulations were performed on simulation and all MRI data to investigate noise propagation from k space to T2 * maps. For phantom tests, T2 * values in regions of interest were extracted on a voxel-wise basis and analyzed using a paired t-test between scanning parameters and mapping methods, with p < 0.05 being significantly different. MDI facilitated a straightforward processing procedure, yielding homogeneous, high-signal-to-noise-ratio (SNR) and artifact-free T2 * maps without explicit coil combination or additional measures. Compared with routine fitting methods, MDI offered significantly (p < 0.05) improved SNR and quantitative accuracy/robustness, with two to three orders higher computational efficiency. MDI also represented low-SNR signals with low T2 * values, avoiding misinterpretation with long-T2 * species.

Seed prioritized unwrapping (SPUN) for MR phase imaging.

BACKGROUND: Region-growing-based phase unwrapping methods have the potential for lossless phase aliasing removal, but generally suffer from unwrapping error propagation associated with discontinuous phase and/or long calculation times. The tradeoff point between robustness and efficiency of phase unwrapping methods in the region-growing category requires improvement. PURPOSE: To demonstrate an accurate, robust, and efficient region-growing phase unwrapping method for MR phase imaging applications. STUDY TYPE: Prospective. SUBJECTS, PHANTOM: normal human subjects (10) / brain surgery patients (2) / water phantoms / computer simulation. FIELD STRENGTH/SEQUENCE: 3 T/gradient echo sequences (2D and 3D). ASSESSMENT: A seed prioritized unwrapping (SPUN) method was developed based on single-region growing, prioritizing only a portion (eg, 100 seeds or 1% seeds) of available seed voxels based on continuity quality during each region-growing iteration. Computer simulation, phantom, and in vivo brain and pelvis scans were performed. The error rates, seed percentages, and calculation times were recorded and reported. SPUN unwrapped phase images were visually evaluated and compared with Laplacian unwrapped results. STATISTICAL TESTS: Monte Carlo simulation was performed on a 3D dipole phase model with a signal-to-noise ratio (SNR) of 1-9 dB, to obtain the mean and standard deviation of calculation error rates and calculation times. RESULTS: Simulation revealed a very robust unwrapping performance of SPUN, reaching an error rate of <0.4% even with SNR as low as 1 dB. For all in vivo data, SPUN was able to robustly unwrap the phase images of modest SNR and complex morphology with visually minimal errors and fast calculation speed (eg, <4 min for 368 × 312 × 128 data) when using a proper seed priority number, eg, Nsp = 1 or 10 voxels for 2D and Nsp = 1% for 3D data. DATA CONCLUSION: SPUN offers very robust and fast region-growing-based phase unwrapping, and does not require any tissue masking or segmentation, nor poses a limitation over imaging parameters. LEVEL OF EVIDENCE: 3 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2019;50:62-70.

Fast GRAPPA reconstruction with random projection.

PURPOSE: To address the issue of computational complexity in generalized autocalibrating partially parallel acquisition (GRAPPA) when several calibration data are used. METHOD: GRAPPA requires fully sampled data for accurate calibration with increasing data needed for higher reduction factors to maintain accuracy, which leads to longer computational time, especially in a three-dimensional (3D) setting and with higher channel count coils. Channel reduction methods have been developed to address this issue when massive array coils are used. In this study, the complexity problem was addressed from a different prospective. Instead of compressing to fewer channels, we propose the use of random projections to reduce the dimension of the linear equation in the calibration phase. The equivalence before and after the reduction is supported by the Johnson-Lindenstrauss lemma. The proposed random projection method can be integrated with channel reduction sequentially for even higher computational efficiency. RESULTS: Experimental results show that GRAPPA with random projection can achieve comparable image quality with much less computational time when compared with conventional GRAPPA without random projection. CONCLUSION: The proposed random projection method is able to reduce the computational time of GRAPPA, especially in a 3D setting, without compromising the image quality, or to improve the reconstruction quality by allowing more data for calibration when the computational time is a limiting factor. Magn Reson Med 74:71-80, 2015. © 2014 Wiley Periodicals, Inc.

Interpretable Dimension Reduction for MRI Channel Suppression.

Channel suppression can reduce the redundant information in multiple channel receiver coils and accelerate reconstruction speed to meet real-time imaging requirements. The principal component analysis has been used for channel suppression, but it is difficult to be interpreted because all channels contribute to principal components. Furthermore, the importance of interpretability in machine learning has recently attracted increasing attention in radiology. To improve the interpretability of PCA-based channel suppression, a sparse PCA method is proposed to reduce the most coils' loadings to be zero. Channel suppression is formulated as solving a nonlinear eigenvalue problem using the inverse power method instead of the direct matrix decomposition. Experimental results of in vivo data show that the sparse PCA-based channel suppression not only improves the interpretability with sparse channels, but also improves reconstruction quality compared to the standard PCA-based reconstruction with the similar reconstruction time.

Virtual Conjugate Coil for Improving KerNL Reconstruction.

Ker NL is a general kernel-based framework for auto calibrated reconstruction method, which does not need any explicit formulas of the kernel function for characterizing nonlinear relationships between acquired and unacquired k-space data. It is non-iterative without requiring a large amount of computational costs. Since the limited autocalibration signals (ACS) are acquired to perform KerNL calibration and the calibration suffers from the overfitting problem, more training data can improve the kernel model accuracy. In this work, virtual conjugate coil data are incorporated into the KerNL calibration and estimation process for enhancing reconstruction performance. Experimental results show that the proposed method can further suppress noise and aliasing artifacts with fewer ACS data and higher acceleration factors. Computation efficiency is still retained to keep fast reconstruction with the random projection.

KerNL: Kernel-Based Nonlinear Approach to Parallel MRI Reconstruction.

The conventional calibration-based parallel imaging method assumes a linear relationship between the acquired multi-channel k-space data and the unacquired missing data, where the linear coefficients are estimated using some auto-calibration data. In this paper, we first analyze the model errors in the conventional calibration-based methods and demonstrate the nonlinear relationship. Then, a much more general nonlinear framework is proposed for auto-calibrated parallel imaging. In this framework, kernel tricks are employed to represent the general nonlinear relationship between acquired and unacquired k-space data without increasing the computational complexity. Identification of the nonlinear relationship is still performed by solving linear equations. Experimental results demonstrate that the proposed method can achieve reconstruction quality superior to GRAPPA and NL-GRAPPA at high net reduction factors.

M-MRI: A Manifold-based Framework to Highly Accelerated Dynamic Magnetic Resonance Imaging.

High-dimensional signals, including dynamic magnetic resonance (dMR) images, often lie on low dimensional manifold. While many current dynamic magnetic resonance imaging (dMRI) reconstruction methods rely on priors which promote low-rank and sparsity, this paper proposes a novel manifold-based framework, we term M-MRI, for dMRI reconstruction from highly undersampled k-space data. Images in dMRI are modeled as points on or close to a smooth manifold, and the underlying manifold geometry is learned through training data, called "navigator" signals. Moreover, low-dimensional embeddings which preserve the learned manifold geometry and effect concise data representations are computed. Capitalizing on the learned manifold geometry, two regularization loss functions are proposed to reconstruct dMR images from highly undersampled k-space data. The advocated framework is validated using extensive numerical tests on phantom and in-vivo data sets.

A Kernel-Based Low-Rank (KLR) Model for Low-Dimensional Manifold Recovery in Highly Accelerated Dynamic MRI.

While many low rank and sparsity-based approaches have been developed for accelerated dynamic magnetic resonance imaging (dMRI), they all use low rankness or sparsity in input space, overlooking the intrinsic nonlinear correlation in most dMRI data. In this paper, we propose a kernel-based framework to allow nonlinear manifold models in reconstruction from sub-Nyquist data. Within this framework, many existing algorithms can be extended to kernel framework with nonlinear models. In particular, we have developed a novel algorithm with a kernel-based low-rank model generalizing the conventional low rank formulation. The algorithm consists of manifold learning using kernel, low rank enforcement in feature space, and preimaging with data consistency. Extensive simulation and experiment results show that the proposed method surpasses the conventional low-rank-modeled approaches for dMRI.

ACCELERATING DYNAMIC MAGNETIC RESONANCE IMAGING BY NONLINEAR SPARSE CODING.

Although being high-dimensional, dynamic magnetic resonance images usually lie on low-dimensional manifolds. Nonlinear models have been shown to capture well that latent low-dimensional nature of data, and can thus lead to improvements in the quality of constrained recovery algorithms. This paper advocates a novel reconstruction algorithm for dynamic magnetic resonance imaging (dMRI) based on nonlinear dictionary learned from low-spatial but high-temporal resolution images. The nonlinear dictionary is initially learned using kernel dictionary learning, and the proposed algorithm subsequently alternates between sparsity enforcement in the feature space and the data-consistency constraint in the original input space. Extensive numerical tests demonstrate that the proposed scheme is superior to popular methods that use linear dictionaries learned from the same set of training data.