Martin Blaimer

SMASH, SENSE, PILS, GRAPPA How to Choose the Optimal Method

Fast imaging methods and the availability of required hardware for magnetic resonance tomography (MRT) have significantly reduced acquisition times from about an hour down to several minutes or seconds. With this development over the last 20 years, magnetic resonance imaging (MRI) has become one of the most important instruments in clinical diagnosis. In recent years, the greatest progress in further increasing imaging speed has been the development of parallel MRI (pMRI). Within the last 3 years, parallel imaging methods have become commercially available, and therefore are now available for a broad clinical use. The basic feature of pMRI is a scan time reduction, applicable to nearly any available MRI method, while maintaining the contrast behavior without requiring higher gradient system performance. Because of its faster image acquisition, pMRI can in some cases even significantly improve image quality. In the last 10 years of pMRI development, several different pMRI reconstruction methods have been set up which partially differ in their philosophy, in the mode of reconstruction as well in their advantages and drawbacks with regard to a successful image reconstruction. In this review, a brief overview is given on the advantages and disadvantages of present pMRI methods in clinical applications, and examples from different daily clinical applications are shown.

Controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA) for multi-slice imaging.

In all current parallel imaging techniques, aliasing artifacts resulting from an undersampled acquisition are removed by means of a specialized image reconstruction algorithm. In this study a new approach termed “controlled aliasing in parallel imaging results in higher acceleration” (CAIPIRINHA) is presented. This technique modifies the appearance of aliasing artifacts during the acquisition to improve the subsequent parallel image reconstruction procedure. This new parallel multi‐slice technique is more efficient compared to other multi‐slice parallel imaging concepts that use only a pure postprocessing approach. In this new approach, multiple slices of arbitrary thickness and distance are excited simultaneously with the use of multi‐band radiofrequency (RF) pulses similar to Hadamard pulses. These data are then undersampled, yielding superimposed slices that appear shifted with respect to each other. The shift of the aliased slices is controlled by modulating the phase of the individual slices in the multi‐band excitation pulse from echo to echo. We show that the reconstruction quality of the aliased slices is better using this shift. This may potentially allow one to use higher acceleration factors than are used in techniques without this excitation scheme. Additionally, slices that have essentially the same coil sensitivity profiles can be separated with this technique. Magn Reson Med 53:684–691, 2005.

Controlled Aliasing in Volumetric Parallel Imaging (2D CAIPIRINHA)

The CAIPIRINHA (Controlled Aliasing In Parallel Imaging Results IN Higher Acceleration) concept in parallel imaging has recently been introduced, which modifies the appearance of aliasing artifacts during data acquisition in order to improve the subsequent parallel imaging reconstruction procedure. This concept has been successfully applied to simultaneous multi‐slice imaging (MS CAIPIRINHA). In this work, we demonstrate that the concept of CAIPIRINHA can also be transferred to 3D imaging, where data reduction can be performed in two spatial dimensions simultaneously. In MS CAIPIRINHA, aliasing is controlled by providing individual slices with different phase cycles by means of alternating multi‐band radio frequency (RF) pulses. In contrast to MS CAIPIRINHA, 2D CAIPIRINHA does not require special RF pulses. Instead, aliasing in 2D parallel imaging can be controlled by modifying the phase encoding sampling strategy. This is done by shifting sampling positions from their normal positions in the undersampled 2D phase encoding scheme. Using this modified sampling strategy, coil sensitivity variations can be exploited more efficiently in multiple dimensions, resulting in a more robust parallel imaging reconstruction. Magn Reson Med, 2006.

General formulation for quantitative G-factor calculation in GRAPPA reconstructions.

In this work a theoretical description for practical quantitative estimation of the noise enhancement in generalized autocalibrating partially parallel acquisitions (GRAPPA) reconstructions, equivalent to the geometry (g)‐factor in sensitivity encoding for fast MRI (SENSE) reconstructions, is described. The GRAPPA g‐factor is derived directly from the GRAPPA reconstruction weights. The procedure presented here also allows the calculation of quantitative g‐factor maps for both the uncombined and combined accelerated GRAPPA images. This enables, for example, a fast comparison between the performances of various GRAPPA reconstruction kernels or SENSE reconstructions. The applicability of this approach is validated on phantom studies and demonstrated using in vivo images for 1D and 2D parallel imaging. Magn Reson Med, 2009.

Parallel magnetic resonance imaging using the GRAPPA operator formalism

In this article it is shown that GRAPPA reconstruction can be reformulated as a matrix operator, similar to ladder or propagator operators used in quantum mechanics, that shifts data in k‐space. Using this formalism, it is shown that there exists an infinitesimal GRAPPA operator that shifts data in k‐space by arbitrarily small amounts. Other desired k‐space shifts can then be accomplished through repeated applications of this infinitesimal GRAPPA operator. Implications of these ideas are described. Magn Reson Med, 2005.

Non-Cartesian Data Reconstruction Using GRAPPA Operator Gridding (GROG)

A novel approach that uses the concepts of parallel imaging to grid data sampled along a non‐Cartesian trajectory using GRAPPA operator gridding (GROG) is described. GROG shifts any acquired data point to its nearest Cartesian location, thereby converting non‐Cartesian to Cartesian data. Unlike other parallel imaging methods, GROG synthesizes the net weight for a shift in any direction from a single basis set of weights along the logical k‐space directions. Given the vastly reduced size of the basis set, GROG calibration and reconstruction requires fewer operations and less calibration data than other parallel imaging methods for gridding. Instead of calculating and applying a density compensation function (DCF), GROG requires only local averaging, as the reconstructed points fall upon the Cartesian grid. Simulations are performed to demonstrate that the root mean square error (RMSE) values of images gridded with GROG are similar to those for images gridded using the gold‐standard convolution gridding. Finally, GROG is compared to the convolution gridding technique using data sampled along radial, spiral, rosette, and BLADE (a.k.a. periodically rotated overlapping parallel lines with enhanced reconstruction [PROPELLER]) trajectories. Magn Reson Med, 2007.

2D‐GRAPPA‐operator for faster 3D parallel MRI

When using parallel MRI (pMRI) methods in combination with three‐dimensional (3D) imaging, it is beneficial to subsample the k‐space along both phase‐encoding directions because one can then take advantage of coil sensitivity variations along two spatial dimensions. This results in an improved reconstruction quality and therefore allows greater scan time reductions as compared to subsampling along one dimension. In this work we present a new approach based on the generalized autocalibrating partially parallel acquisitions (GRAPPA) technique that allows Fourier‐domain reconstructions of data sets that are subsampled along two dimensions. The method works by splitting the 2D reconstruction process into two separate 1D reconstructions. This approach is compared with an extension of the conventional GRAPPA method that directly regenerates missing data points of a 2D subsampled k‐space by performing a linear combination of acquired data points. In this paper we describe the theoretical background and present computer simulations and in vivo experiments. Magn Reson Med, 2006.

Accelerated volumetric MRI with a SENSE/GRAPPA combination

To combine the specific advantages of the generalized autocalibrating partially parallel acquisitions (GRAPPA) technique and sensitivity encoding (SENSE) with two‐dimensional (2D) undersampling.

Reconstruction of undersampled non-Cartesian data sets using pseudo-Cartesian GRAPPA in conjunction with GROG

Most k‐space‐based parallel imaging reconstruction techniques, such as Generalized Autocalibrating Partially Parallel Acquisitions (GRAPPA), necessitate the acquisition of regularly sampled Cartesian k‐space data to reconstruct a nonaliased image efficiently. However, non‐Cartesian sampling schemes offer some inherent advantages to the user due to their better coverage of the center of k‐space and faster acquisition times. On the other hand, these sampling schemes have the disadvantage that the points acquired generally do not lie on a grid and have complex k‐space sampling patterns. Thus, the extension of Cartesian GRAPPA to non‐Cartesian sequences is nontrivial. This study introduces a simple, novel method for performing Cartesian GRAPPA reconstructions on undersampled non‐Cartesian k‐space data gridded using GROG (GRAPPA Operator Gridding) to arrive at a nonaliased image. Because the undersampled non‐Cartesian data cannot be reconstructed using a single GRAPPA kernel, several Cartesian patterns are selected for the reconstruction. This flexibility in terms of both the appearance and number of patterns allows this pseudo‐Cartesian GRAPPA to be used with undersampled data sets acquired with any non‐Cartesian trajectory. The successful implementation of the reconstruction algorithm using several different trajectories, including radial, rosette, spiral, one‐dimensional non‐Cartesian, and zig–zag trajectories, is demonstrated. Magn Reson Med 59:1127–1137, 2008.

Fast MR parameter mapping using k-t principal component analysis.

Quantification of magnetic resonance parameters plays an increasingly important role in clinical applications, such as the detection and classification of neurodegenerative diseases. The major obstacle that remains for its widespread use in clinical routine is the long scanning times. Therefore, strategies that allow for significant decreases in scan time are highly desired. Recently, the k‐t principal component analysis method was introduced for dynamic cardiac imaging to accelerate data acquisition. This is done by undersampling k‐t space and constraining the reconstruction of the aliased data based on the k‐t Broad‐use Linear Acquisition Speed‐up Technique (BLAST) concept and predetermined temporal basis functions. The objective of this study was to investigate whether the k‐t principal component analysis concept can be adapted to parameter quantification, specifically allowing for significant acceleration of an inversion recovery fast imaging with steady state precession (TrueFISP) acquisition. We found that three basis functions and a single training data line in central k‐space were sufficient to achieve up to an 8‐fold acceleration of the quantification measurement. This allows for an estimation of relaxation times T1 and T2 and spin density in one slice with sub‐millimeter in‐plane resolution, in only 6 s. Our findings demonstrate that the k‐t principal component analysis method is a potential candidate to bring the acquisition time for magnetic resonance parameter mapping to a clinically acceptable level. Magn Reson Med, 2011.