Inherently self‐calibrating non‐cartesian parallel imaging

Abstract

The use of self-calibrating techniques in parallel magnetic resonance imaging eliminates the need for coil sensitivity calibration scans and avoids potential mismatches between calibration scans and subsequent accelerated acquisitions (e.g., as a result of patient motion). Most examples of self-calibrating Cartesian parallel imaging techniques have required the use of modified k-space trajectories that are densely sampled at the center and more sparsely sampled in the periphery. However, spiral and radial trajectories offer inherent self-calibrating characteristics because of their densely sampled center. At no additional cost in acquisition time and with no modification in scanning protocols, in vivo coil sensitivity maps may be extracted from the densely sampled central region of k-space. This work demonstrates the feasibility of self-calibrated spiral and radial parallel imaging using a previously described iterative non-Cartesian sensitivity encoding algorithm. The accuracy of coil sensitivity estimates is a major determinant of the quality of parallel magnetic resonance image reconstructions. While the level of tolerance for sensitivity miscalibration differs from technique to technique, any significant discrepancy in coil sensitivity references introduces systematic errors in reconstructed images while degrading overall image quality. Self-calibrating parallel imaging techniques that employ variable-density k-space acquisition schemes have been demonstrated (1-4). Selfcalibration eliminates the need for an external sensitivity reference, making parallel image reconstructions less susceptible to miscalibration and image degradation resulting from bulk patient motion.Self-calibrating parallel image reconstructions have been commonly implemented for rectilinear (or Cartesian) trajectories. Phase-encoded lines in the outer k-space region are omitted at a chosen outer reduction factor, whereas those in the central k-space region remain densely sampled. The densely sampled central lines are used for calibration, either through fitting directly in k-space (1-3) or through Fourier transformation to generate low-resolution in vivo sensitivity maps (4). The omitted phase-encoded lines are then reconstructed by a parallel imaging technique of choice.It has been noted that non-Cartesian trajectories such as spiral and radial trajectories are logical candidates for self-calibration due to their characteristic oversampled kspace center (4). Even for accelerated acquisitions, the center of k-space will generally be sampled with sufficient density to enable the creation of reliable low-resolution maps of component coil sensitivities without aliasing artifacts. These low-resolution maps can then provide the coil sensitivity references required for reconstruction of the outer k-space signal data.However, non-Cartesian k-space trajectories pose greater challenges for parallel image reconstruction than their Cartesian counterparts due to the memory requirements and computational demand of inverting large matrices. Recent work...

Nonlinear inverse reconstruction for real‐time MRI of the human heart using undersampled radial FLASH

“…For radial imaging, reconstructions by self-calibrated SENSE that estimate coil sensitivities from the k-space center are also possible (12). Unfortunately, the method is mathematically not exact and has previously been shown to cause errors in the Cartesian case (1,2).…”

3Parallel magnetic resonance imaging with adaptive radius in k‐space (PARS): Constrained image reconstruction using k‐space locality in radiofrequency coil encoded data

“…(15). SVD conditioning (truncation of singular values less than 0.01 of the maximum singular value) was used to further enhance the stable inversion of the localized encoding matrices.…”

“…⌳ PARS , the noise covariance matrix for the component coil image, is expressed as PARS ϭ F⍀ (F⍀) H , [15] where ⌿ is the noise covariance matrix of the acquired signal, and ͑ ⅐ ͒ H is the Hermitian conjugate operator. When a Cholesky factorization is performed, ϭ LL H [16] to decorrelate and normalize noise in the acquired signal (6), Eq.…”

Functional Imaging: CT and MRI

“…This will lead to relatively black lungs and excellent delineation of the chest wall, mediastinum and diaphragm. Parallel imaging sequences assist in obtaining images faster, within a single breath-hold, thus allowing for rapid image acquisition without the issue of motion artifact (136)(137)(138) A host of sequences are available, ranging from those focused on the diaphragm and mediastinum to those aimed at obtaining signal from the actual lungs themselves. Using intravenous Gadolinium-based contrast agents, it is possible to delineate the pulmonary vascular tree as well as the right heart.…”

“…The newer MRI systems all have the ability to perform parallel imaging techniques, which is somewhat similar to multidetector row CT in that multiple slices are excited and read out simultaneously, thus increasing temporal resolution by a factor of 2-8. (136)(137)(138) Although MRI has never really played a major role in routine chest imaging, several pathological processes can be evaluated using proton imaging, including pleural effusions, pneumonia, lung tumors (particularly useful in Pancoast tumors and for determination of tumor invasion in mesothelioma) and the assessment of the mediastinum (Figure 7). (139)(140)(141).The technique is complementary to CT, although tissue plane definition and characterization is better using MRI.…”