Lipid-suppressed and tissue-fraction corrected metabolic distributions in human central brain structures using 2D 1 H magnetic resonance spectroscopic imaging at 7 T.

INTRODUCTION: Magnetic resonance spectroscopic imaging (MRSI) has the potential to add a layer of understanding of the neurobiological mechanisms underlying brain diseases, disease progression, and treatment efficacy. Limitations related to metabolite fitting of low signal-to-noise ratios data, signal variations due to partial-volume effects, acquisition and extracranial lipid artifacts, along with clinically relevant aspects such as scan time constraints, are among the challenges associated with in vivo MRSI. METHODS: The aim of this work was to address some of these factors and to develop an acquisition, reconstruction, and postprocessing pipeline to derive lipid-suppressed metabolite values of central brain structures based on free-induction decay measurements made using a 7 T MR scanner. Anatomical images were used to perform high-resolution (1 mm3 ) partial-volume correction to account for gray matter, white matter (WM), and cerebral-spinal fluid signal contributions. Implementation of automatic quality control thresholds and normalization of metabolic maps from 23 subjects to the Montreal Neurological Institute (MNI) standard atlas facilitated the creation of high-resolution average metabolite maps of several clinically relevant metabolites in central brain regions, while accounting for macromolecular distributions. Partial-volume correction improved the delineation of deep brain nuclei. We report average metabolite values including glutamate + glutamine (Glx), glycerophosphocholine, choline and phosphocholine (tCho), (phospo)creatine, myo-inositol and glycine (mI-Gly), glutathione, N-acetyl-aspartyl glutamate(and glutamine), and N-acetyl-aspartate in the basal ganglia, central WM (thalamic radiation, corpus callosum) as well as insular cortex and intracalcarine sulcus. CONCLUSION: MNI-registered average metabolite maps facilitate group-based analysis, thus offering the possibility to mitigate uncertainty in variable MRSI data.

Over-discretized SENSE reconstruction and B0 correction for accelerated non-lipid-suppressed 1 H FID MRSI of the human brain at 9.4 T.

The aim of this work was to use post-processing methods to improve the data quality of metabolite maps acquired on the human brain at 9.4 T with accelerated acquisition schemes. This was accomplished by combining an improved sensitivity encoding (SENSE) reconstruction with a B0 correction of spatially over-discretized magnetic resonance spectroscopic imaging (MRSI) data. Since MRSI scans suffer from long scan duration, investigating different acceleration techniques has recently been the focus of several studies. Due to strong B0 inhomogeneity and strict specific absorption rate (SAR) limitations at ultra-high fields, the use of a low-SAR sequence combined with an acceleration technique that is compatible with dynamic B0 shim updating is preferable. Hence, in this study, a non-lipid-suppressed ultra-short TE and TR1 H free induction decay MRSI sequence is combined with an in-plane SENSE acceleration technique to obtain high-resolution metabolite maps in a clinically feasible scan time. One of the major issues in applying parallel imaging techniques to non-lipid-suppressed MRSI is the presence of strong lipid aliasing artifacts, which if not thoroughly resolved will hinder the accurate quantification of the metabolites of interest. To achieve a more robust reconstruction, an over-discretized SENSE reconstruction (with direct control over the shape of the spatial response function) was combined with an over-discretized B0 correction. This method is compared with conventional SENSE reconstruction for seven acceleration schemes on four healthy volunteers. The over-discretized method consistently outperformed conventional SENSE, resulting in an average of 23 ± 1.2% higher signal-to-noise ratio and 8 ± 2.9% less error in the fitting of the N-acetylaspartate signal over a whole brain slice. The highest achievable acceleration factor with the proposed technique was determined to be 4. Finally, using the over-discretized method, high-resolution (97 µL nominal voxel size) metabolite maps can be acquired in 3.75 min at 9.4 T. This enables the acquisition of high-resolution metabolite maps with more spatial coverage at ultra-high fields.

MultiNet PyGRAPPA: Multiple neural networks for reconstructing variable density GRAPPA (a 1H FID MRSI study).

Magnetic resonance spectroscopic imaging (MRSI) is a powerful tool for mapping metabolite levels across the brain, however, it generally suffers from long scan times. This severely hinders its application in clinical settings. Additionally, the presence of nuisance signals (e.g. the subcutaneous lipid signals close to the skull region in brain metabolite mapping) makes it challenging to apply conventional acceleration techniques to shorten the scan times. The goal of this work is, therefore, to increase the overall applicability of high resolution metabolite mapping using 1H MRSI by introducing a novel GRAPPA acceleration acquisition/reconstruction technique. An improved reconstruction method (MultiNet) is introduced that uses machine learning, specifically neural networks, to reconstruct accelerated data. The method is further modified to use more neural networks with nonlinear hidden layers and is then combined with a variable density undersampling scheme (MultiNet PyGRAPPA) to enable higher in-plane acceleration factors of R = 5.6 and R = 7 for a non-lipid suppressed ultra-short TR and TE 1H FID MRSI sequence. The proposed method is evaluated for high resolution metabolite mapping of the human brain at 9.4T. The results show that the proposed method is superior to conventional GRAPPA: there is no significant residual lipid aliasing artifact in the images when the proposed MultiNet method is used. Furthermore, the MultiNet PyGRAPPA acquisition/reconstruction method with R = 5.6 results in reproducible high resolution metabolite maps (with an in-plane matrix size of 64 × 64) that can be acquired in 2.8 min on 9.4T. In conclusion, using multiple neural networks to predict the missing points in GRAPPA reconstruction results in a more reliable data recovery while keeping the noise levels under control. Combining this high fidelity reconstruction with variable density undersampling (MultiNet PyGRAPPA) enables higher in-plane acceleration factors even for non-lipid suppressed 1H FID MRSI, without introducing any structured aliasing artifact in the image.

Compressed sensing for high-resolution nonlipid suppressed 1 H FID MRSI of the human brain at 9.4T.

PURPOSE: The aim of this study was to apply compressed sensing to accelerate the acquisition of high resolution metabolite maps of the human brain using a nonlipid suppressed ultra-short TR and TE 1 H FID MRSI sequence at 9.4T. METHODS: X-t sparse compressed sensing reconstruction was optimized for nonlipid suppressed 1 H FID MRSI data. Coil-by-coil x-t sparse reconstruction was compared with SENSE x-t sparse and low rank reconstruction. The effect of matrix size and spatial resolution on the achievable acceleration factor was studied. Finally, in vivo metabolite maps with different acceleration factors of 2, 4, 5, and 10 were acquired and compared. RESULTS: Coil-by-coil x-t sparse compressed sensing reconstruction was not able to reliably recover the nonlipid suppressed data, rather a combination of parallel and sparse reconstruction was necessary (SENSE x-t sparse). For acceleration factors of up to 5, both the low-rank and the compressed sensing methods were able to reconstruct the data comparably well (root mean squared errors [RMSEs] ≤ 10.5% for Cre). However, the reconstruction time of the low rank algorithm was drastically longer than compressed sensing. Using the optimized compressed sensing reconstruction, acceleration factors of 4 or 5 could be reached for the MRSI data with a matrix size of 64 × 64. For lower spatial resolutions, an acceleration factor of up to R∼4 was successfully achieved. CONCLUSION: By tailoring the reconstruction scheme to the nonlipid suppressed data through parameter optimization and performance evaluation, we present high resolution (97 µL voxel size) accelerated in vivo metabolite maps of the human brain acquired at 9.4T within scan times of 3 to 3.75 min.

High and ultra-high resolution metabolite mapping of the human brain using 1H FID MRSI at 9.4T.

Magnetic resonance spectroscopic imaging (MRSI) is a promising technique for mapping the spatial distribution of multiple metabolites in the human brain. These metabolite maps can be used as a diagnostic tool to gain insight into several biochemical processes and diseases in the brain. In comparison to lower field strengths, MRSI at ultra-high field strengths benefits from a higher signal to noise ratio (SNR) as well as higher chemical shift dispersion, and hence spectral resolution. This study combines the benefits of an ultra-high field magnet with the advantages of an ultra-short TE and TR single-slice FID-MRSI sequence (such as negligible J-evolution and loss of SNR due to T2 relaxation effects) and presents the first metabolite maps acquired at 9.4T in the healthy human brain at both high (voxel size of 97.6µL) and ultra-high (voxel size of 24.4µL) spatial resolutions in a scan time of 11 and 46min respectively. In comparison to lower field strengths, more anatomically-detailed maps with higher SNR from a larger number of metabolites are shown. A total of 12 metabolites including glutamate (Glu), glutamine (Gln), N-acetyl-aspartyl-glutamate (NAAG), Gamma-aminobutyric acid (GABA) and glutathione (GSH) are reliably mapped. Comprehensive description of the methodology behind these maps is provided.

Non-water-suppressed 1 H FID-MRSI at 3T and 9.4T.

PURPOSE: This study investigates metabolite concentrations using metabolite-cycled 1 H free induction decay (FID) magnetic resonance spectroscopic imaging (MRSI) at ultra-high fields. METHODS: A non-lipid-suppressed and slice-selective ultra-short echo time (TE) 1 H FID MRSI sequence was combined with a low-specific absorption rate (SAR) asymmetric inversion adiabatic pulse to enable non-water-suppressed metabolite mapping using metabolite-cycling at 9.4T. The results were compared to a water-suppressed FID MRSI sequence, and the same study was performed at 3T for comparison. The scan times for performing single-slice metabolite mapping with a nominal voxel size of 0.4 mL were 14 and 17.5 min on 3T and 9.4T, respectively. RESULTS: The low-SAR asymmetric inversion adiabatic pulse enabled reliable non-water-suppressed metabolite mapping using metabolite cycling at both 3T and 9.4T. The spectra and maps showed good agreement with the water-suppressed FID MRSI ones at both field strengths. A quantitative analysis of metabolite ratios with respect to N-acetyl aspartate (NAA) was performed. The difference in Cre/NAA was statistically significant, ∼0.1 higher for the non-water-suppressed case than for water suppression (from 0.73 to 0.64 at 3T and from 0.69 to 0.59 at 9.4T). The difference is likely because of chemical exchange effects of the water suppression pulses. Small differences in mI/NAA were also statistically significant, however, are they are less reliable because the metabolite peaks are close to the water peak that may be affected by the water suppression pulses or metabolite-cycling inversion pulse. CONCLUSION: We showed the first implementation of non-water-suppressed metabolite-cycled 1 H FID MRSI at ultra-high fields. An increase in Cre/NAA was seen for the metabolite-cycled case. The same methodology was further applied at 3T and similar results were observed. Magn Reson Med 80:442-451, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Constrained optimization for position calibration of an NMR field camera.

PURPOSE: Knowledge of the positions of field probes in an NMR field camera is necessary for monitoring the B0 field. The typical method of estimating these positions is by switching the gradients with known strengths and calculating the positions using the phases of the FIDs. We investigated improving the accuracy of estimating the probe positions and analyzed the effect of inaccurate estimations on field monitoring. METHODS: The field probe positions were estimated by 1) assuming ideal gradient fields, 2) using measured gradient fields (including nonlinearities), and 3) using measured gradient fields with relative position constraints. The fields measured with the NMR field camera were compared to fields acquired using a dual-echo gradient recalled echo B0 mapping sequence. Comparisons were done for shim fields from second- to fourth-order shim terms. RESULTS: The position estimation was the most accurate when relative position constraints were used in conjunction with measured (nonlinear) gradient fields. The effect of more accurate position estimates was seen when compared to fields measured using a B0 mapping sequence (up to 10%-15% more accurate for some shim fields). The models acquired from the field camera are sensitive to noise due to the low number of spatial sample points. CONCLUSION: Position estimation of field probes in an NMR camera can be improved using relative position constraints and nonlinear gradient fields. Magn Reson Med 80:380-390, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

A comparison of optimization algorithms for localized in vivo B0 shimming.

PURPOSE: To compare several different optimization algorithms currently used for localized in vivo B0 shimming, and to introduce a novel, fast, and robust constrained regularized algorithm (ConsTru) for this purpose. METHODS: Ten different optimization algorithms (including samples from both generic and dedicated least-squares solvers, and a novel constrained regularized inversion method) were implemented and compared for shimming in five different shimming volumes on 66 in vivo data sets from both 7 T and 9.4 T. The best algorithm was chosen to perform single-voxel spectroscopy at 9.4 T in the frontal cortex of the brain on 10 volunteers. RESULTS: The results of the performance tests proved that the shimming algorithm is prone to unstable solutions if it depends on the value of a starting point, and is not regularized to handle ill-conditioned problems. The ConsTru algorithm proved to be the most robust, fast, and efficient algorithm among all of the chosen algorithms. It enabled acquisition of spectra of reproducible high quality in the frontal cortex at 9.4 T. CONCLUSIONS: For localized in vivo B0 shimming, the use of a dedicated linear least-squares solver instead of a generic nonlinear one is highly recommended. Among all of the linear solvers, the constrained regularized method (ConsTru) was found to be both fast and most robust. Magn Reson Med 79:1145-1156, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Modeling real shim fields for very high degree (and order) B0 shimming of the human brain at 9.4 T.

PURPOSE: To describe the process of calibrating a B0 shim system using high-degree (or high order) spherical harmonic models of the measured shim fields, to provide a method that considers amplitude dependency of these models, and to show the advantage of very high-degree B0 shimming for whole-brain and single-slice applications at 9.4 Tesla (T). METHODS: An insert shim with up to fourth and partial fifth/sixth degree (order) spherical harmonics was used with a Siemens 9.4T scanner. Each shim field was measured and modeled as input for the shimming algorithm. Optimal shim currents can therefore be calculated in a single iteration. A range of shim currents was used in the modeling to account for possible amplitude nonlinearities. The modeled shim fields were used to compare different degrees of whole-brain B0 shimming on healthy subjects. RESULTS: The ideal shim fields did not correctly shim the subject brains. However, using the modeled shim fields improved the B0 homogeneity from 55.1 (second degree) to 44.68 Hz (partial fifth/sixth degree) on the whole brains of 9 healthy volunteers, with a total applied current of 0.77 and 6.8 A, respectively. CONCLUSIONS: The necessity of calibrating the shim system was shown. Better B0 homogeneity drastically reduces signal dropout and distortions for echo-planar imaging, and significantly improves the linewidths of MR spectroscopy imaging. Magn Reson Med 79:529-540, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

A system for automated, noninvasive, morphology-based evaluation of induced pluripotent stem cell cultures.

Due to the rapid adoption and use of human induced pluripotent stem cells (iPSCs) in recent years, there is a need for new technologies that standardize the evaluation of iPSCs to allow the objective comparison of results across different experiments and groups. In this article, we present a noninvasive, fully automated, and analytical system for morphology-based evaluation of iPSC cultures that consists of time-lapse microscopy and novel image analysis software. The presented system acquires low-light phase-contrast images of iPSC growth collected during a period of several days in culture, measures geometrical- and texture-based features of iPSC colonies throughout time, and derives a set of six biologically relevant features to automatically rank the quality of the cell culture. In a study of 94 iPSC cultures, we demonstrated the accuracy of the system by comparing the automated ranking with an independent expert evaluation based on visual review of the time-lapse movies. To our knowledge, this is the first demonstration of a fully automated and objective assessment of iPSC culture quality using noninvasive methods.