Virtual coil concept for improved parallel MRI employing conjugate symmetric signals

Abstract

A new approach for utilizing conjugate k-space symmetry for improved parallel MRI performance is presented. By generating virtual coils containing conjugate symmetric k-space signals from actual coils, additional image-and coil-phase information can be incorporated into the reconstruction process for parallel acquisition techniques. In that way the reconstruction conditions are improved, resulting in less noise enhancement. In particular in combination with generalized autocalibrating partially parallel acquisitions (GRAPPA), the virtual coil concept represents a practical approach since no explicit spatial phase information is required. In addition, the influence of phase variations originating from the complex receiver coils as well as from the background is investigated. Many strategies for reducing the scan time of MRI experiments work by collecting only a fraction of the data required for an artifact-free image. In the phase-constrained or partial-Fourier approach for Cartesian MRI, the k-space is sampled asymmetrically along the phase-encoding (PE) direction with full sampling density. While one-half of k-space is fully Fourier-encoded, only a portion of the other half of k-space is covered. Dedicated reconstruction algorithms recover the missing data by exploiting conjugate symmetry properties of k-space (1-3).Alternative approaches for scan time reductions include the parallel MRI (pMRI) methods (for example (4 -6)). The basic idea of Cartesian pMRI is to cover the entire span of k-space, but skip a fraction of the PE steps. In that way the Nyquist criterion is violated along the PE dimension, resulting in aliasing artifacts. Parallel MRI algorithms use spatial information inherent in an array of multiple receiver coils to either recover missing data in k-space or remove aliasing artifacts in the image domain. Cartesian sensitivity encoding (SENSE) (5) is a widespread imagedomain pMRI technique that unfolds superimposed pixels by incorporating spatial coil sensitivity information. However, for nonideal coil configurations noise enhancement occurs as a result of ill-conditioning of the inverse solution. The noise enhancement varies within the field of view (FOV) and can be analytically described by the geometry factor (g-factor). In addition to SENSE, the k-space domain "generalized autocalibrating partially parallel acquisitions" (GRAPPA) (6) technique is widely used. In GRAPPA, missing k-space lines in a single coil are approximated by a linear combination of measured k-space lines from all coils. To calculate the reconstruction coefficients, additional Nyquist sampled k-space lines (autocalibration signal, ACS) have to be measured. Equivalent to SENSE, noise enhancement occurs depending on the coil sensitivity profiles. However, in contrast to SENSE an analytical prediction of the noise enhancement is not trivial for GRAPPA. Both SENSE and GRAPPA are successfully applied in daily clinical routine.Several approaches for combining partial-Fourier acquisitions with parallel MRI have been presented to achi...

Derivative encoding for parallel magnetic resonance imaging

“…[28][29][30][31] For example, it should be possible to combine k space partial derivatives constraints with Tikhonov regularization to reduce the magnitude of regularization parameters and hence the loss of spatial resolution. Combining derivative encoding with the virtual coil concept 32 to further expand the SENSE encoding matrix is also possible.…”

A kernel approach to parallel MRI reconstruction

“…The first and third parts of the Eq. (11) can be used to characterize other nonlinear effects in practice such that noise and approximation errors are suppressed.…”

“…(11). If the dimension is too low, prediction is inaccurate because the kernel space is not complex enough to accurately describe b .…”

“…When the net acceleration factor is high, GRAPPA reconstruction can suffer from aliasing artifacts and noise amplifications. Methods have been developed in recent years to improve GRAPPA using localized coil calibration and variable density sampling [2], multicolumn multiline interpolation [3], regularization [4,5], iteratively reweighted least-squares [6], high-pass filtering [7], cross-validation [8,9], iterative optimization [10], virtual coil using conjugate symmetric signals [11], multi-slice weighting [12], or infinite pulse response (IIR) filtering [13], etc.…”

Improving GRAPPA using cross‐sampled autocalibration data

“…When the number of ACS lines is insufficient, aliasing artifacts are present in reconstruction along the undersampling direction. A number of reconstruction methods have been proposed to reduce aliasing artifacts and improve image quality, such as multicolumn multiline interpolation (4), regularization (5), reweighted least squares (6), high-pass filtering (7), cross-validation (8), iterative optimization (9), virtual coil using conjugate symmetric signals (10), multislice weighting (11), and an infinite impulse response model (12).…”