Magnetic Resonance Materials in Physics, Biology and Medicine最新文献

Subject motion is a long-standing problem of magnetic resonance imaging (MRI), which can seriously deteriorate the image quality. Various prospective and retrospective methods have been proposed for MRI motion correction, among which deep learning approaches have achieved state-of-the-art motion correction performance. This survey paper aims to provide a comprehensive review of deep learning-based MRI motion correction methods. Neural networks used for motion artifacts reduction and motion estimation in the image domain or frequency domain are detailed. Furthermore, besides motion-corrected MRI reconstruction, how estimated motion is applied in other downstream tasks is briefly introduced, aiming to strengthen the interaction between different research areas. Finally, we identify current limitations and point out future directions of deep learning-based MRI motion correction.

Objective

Conventional single-target field control for matrix gradient coils will add control complexity in MRI spatial encoding, such as designing specialized fields and sequences. This complexity can be reduced by multi-target field control, which is realized by optimizing the coil structure according to target fields.

Methods

Based on the principle of multi-target field control, the X, Y and Z gradient fields can be set as target fields, and all coil elements can then be divided into three groups to generate these fields. An improved simulated annealing algorithm is proposed to optimize the coil element distribution of each group to generate the corresponding target field. In the improved simulated annealing process, two swapping modes are presented, and randomly selected with certain probabilities that are set to 0.25, 0.5 and 0.75, respectively. The flexibility of the final designed structure is demonstrated by a spherical harmonic basis up to the full second order with single-target field control. An experimental platform is built to measure the gradient fields generated by the designed structure with multi-target target control.

Results

With three probabilities of swapping modes, three similar coil element distributions are optimized, and their maximum magnetic field errors for generating X, Y and Z gradients are all below 5%. The structure selected for the final design is the one with a probability of 0.75, considering the coil performance and structural symmetry. The maximum error for all target fields generated by single-target field control is also below 5%. The experimental results show that the measured gradient fields along the axes have enough strength and high linearity.

Conclusions

With the proposed improved simulated annealing algorithm and swapping modes, multi-target field control for matrix gradient coils is verified and achieved in this study by optimizing the coil element distribution. Moreover, this study provides a solution to simplify the complexity of controlling the matrix gradient coil in spatial encoding.

Purpose

Propeller fast-spin-echo diffusion magnetic resonance imaging (FSE-dMRI) is essential for the diagnosis of Cholesteatoma. However, at clinical 1.5 T MRI, its signal-to-noise ratio (SNR) remains relatively low. To gain sufficient SNR, signal averaging (number of excitations, NEX) is usually used with the cost of prolonged scan time. In this work, we leveraged the benefits of Locally Low Rank (LLR) constrained reconstruction to enhance the SNR. Furthermore, we enhanced both the speed and SNR by employing Convolutional Neural Networks (CNNs) for the accelerated PROPELLER FSE-dMRI on a 1.5 T clinical scanner.

Methods

Residual U-Net (RU-Net) was found to be efficient for propeller FSE-dMRI data. It was trained to predict 2-NEX images obtained by Locally Low Rank (LLR) constrained reconstruction and used 1-NEX images obtained via simplified reconstruction as the inputs. The brain scans from healthy volunteers and patients with cholesteatoma were performed for model training and testing. The performance of trained networks was evaluated with normalized root-mean-square-error (NRMSE), structural similarity index measure (SSIM), and peak SNR (PSNR).

Results

For 4 × under-sampled with 7 blades data, online reconstruction appears to provide suboptimal images—some small details are missing due to high noise interferences. Offline LLR enables suppression of noises and discovering some small structures. RU-Net demonstrated further improvement compared to LLR by increasing 18.87% of PSNR, 2.11% of SSIM, and reducing 53.84% of NRMSE. Moreover, RU-Net is about 1500 × faster than LLR (0.03 vs. 47.59 s/slice).

Conclusion

The LLR remarkably enhances the SNR compared to online reconstruction. Moreover, RU-Net improves propeller FSE-dMRI as reflected in PSNR, SSIM, and NRMSE. It requires only 1-NEX data, which allows a 2 × scan time reduction. In addition, its speed is approximately 1500 times faster than that of LLR-constrained reconstruction.

The acquisition of images minutes or even hours after intravenous extracellular gadolinium-based contrast agents (GBCA) administration (“Late/Delayed Gadolinium Enhancement” imaging; in this review, further termed LGE) has gained significant prominence in recent years in magnetic resonance imaging. The major limitation of LGE is the long examination time; thus, it becomes necessary to understand when it is worth waiting time after the intravenous injection of GBCA and which additional information comes from LGE. LGE can potentially be applied to various anatomical sites, such as heart, arterial vessels, lung, brain, abdomen, breast, and the musculoskeletal system, with different pathophysiological mechanisms. One of the most popular clinical applications of LGE regards the assessment of myocardial tissue thanks to its ability to highlight areas of acute myocardial damage and fibrotic tissues. Other frequently applied clinical contexts involve the study of the urinary tract with magnetic resonance urography and identifying pathological abdominal processes characterized by high fibrous stroma, such as biliary tract tumors, autoimmune pancreatitis, or intestinal fibrosis in Crohn’s disease. One of the current areas of heightened research interest revolves around the possibility of non-invasively studying the dynamics of neurofluids in the brain (the glymphatic system), the disruption of which could underlie many neurological disorders.

Objective: To find a possible quantitative relation between activation-induced fast (< 10 s) changes in the γ-aminobutyric acid (GABA) level and the amplitude of a blood oxygen level-dependent contrast (BOLD) response (according to magnetic resonance spectroscopy [MRS] and functional magnetic resonance imaging [fMRI]).

Materials and methods: fMRI data and MEGA-PRESS magnetic resonance spectra [echo time (TE)/repetition time (TR) = 68 ms/1500 ms] of an activated area in the visual cortex of 33 subjects were acquired using a 3 T MR scanner. Stimulation was performed by presenting an image of a flickering checkerboard for 3 s, repeated with an interval of 13.5 s. The time course of GABA and creatine (Cr) concentrations and the width and height of resonance lines were obtained with a nominal time resolution of 1.5 s. Changes in the linewidth and height of n-acetylaspartate (NAA) and Cr signals were used to determine the BOLD effect.

Results: In response to the activation, the BOLD-corrected GABA + /Cr ratio increased by 5.0% (q = 0.027) and 3.8% (q = 0.048) at 1.6 and 3.1 s, respectively, after the start of the stimulus. Time courses of Cr and NAA signal width and height reached a maximum change at the 6th second (~ 1.2-1.5%, q < 0.05).

Conclusion: The quick response of the observed GABA concentration to the short stimulus is most likely due to a release of GABA from vesicles followed by its packaging back into vesicles.

Contemporary whole-body low-field MRI scanners (< 1 T) present new and exciting opportunities for improved body imaging. The fundamental reason is that the reduced off-resonance and reduced SAR provide substantially increased flexibility in the design of MRI pulse sequences. Promising body applications include lung parenchyma imaging, imaging adjacent to metallic implants, cardiac imaging, and dynamic imaging in general. The lower cost of such systems may make MRI favorable for screening high-risk populations and population health research, and the more open configurations allowed may prove favorable for obese subjects and for pregnant women. This article summarizes promising body applications for contemporary whole-body low-field MRI systems, with a focus on new platforms developed within the past 5 years. This is an active area of research, and one can expect many improvements as MRI physicists fully explore the landscape of pulse sequences that are feasible, and as clinicians apply these to patient populations.

Objective: To evaluate the repeatability of cartilage volume and thickness values at 1.5 T MRI using a fully automatic cartilage segmentation method and reproducibility of the method between 1.5 T and 3 T data.

Methods: The study included 20 knee joints from 10 healthy subjects with each subject having undergone double-knee MRI. All knees were scanned at 1.5 T and 3 T MR scanners using a three-dimensional (3D) high-resolution dual-echo in steady state (DESS) sequence. Cartilage volume and thickness of 21 subregions were quantified using a fully automatic cartilage segmentation research application (MR Chondral Health, version 3.0, Siemens Healthcare, Erlangen, Germany). The volume and thickness values derived from fully automatically computed segmentation masks were analyzed for the scan-rescan data from the same volunteers. The accuracy of the automatic segmentation of the cartilage in 1.5 T images was evaluated by the dice similarity coefficient (DSC) and Hausdorff distance (HD) using the manually corrected segmentation as a reference. The volume and thickness values calculated from 1.5 T and 3 T were also compared.

Results: No statistically significant differences were found for cartilage thickness or volume across all subregions between the scan-rescanned data at 1.5 T (P > 0.05). The mean DSC between the fully automatic and manually corrected knee cartilage segmentation contours at 1.5 T was 0.9946. The average value of HD was 2.41 mm. Overall, there was no statistically significant difference in the cartilage volume or thickness in most-subregions between the two field strengths (P > 0.05) except for the medial region of femur and tibia. Bland-Altman plot and intraclass correlation coefficient (ICC) showed high consistency between results obtained based on same and different scanning sequences.

Conclusion: The cartilage segmentation software had high repeatability for DESS images obtained from the same device. In addition, the overall reproducibility of the images obtained from equipment of two different field strengths was satisfactory.

Objective: The prospect of being able to gain relevant information from cardiovascular magnetic resonance (CMR) image analysis automatically opens up new potential to assist the evaluating physician. For machine-learning-based classification of complex congenital heart disease, only few studies have used CMR.

Materials and methods: This study presents a tailor-made neural network architecture for detection of 7 distinctive anatomic landmarks in CMR images of patients with hypoplastic left heart syndrome (HLHS) in Fontan circulation or healthy controls and demonstrates the potential of the spatial arrangement of the landmarks to identify HLHS. The method was applied to the axial SSFP CMR scans of 46 patients with HLHS and 33 healthy controls.

Results: The displacement between predicted and annotated landmark had a standard deviation of 8-17 mm and was larger than the interobserver variability by a factor of 1.1-2.0. A high overall classification accuracy of 98.7% was achieved.

Discussion: Decoupling the identification of clinically meaningful anatomic landmarks from the actual classification improved transparency of classification results. Information from such automated analysis could be used to quickly jump to anatomic positions and guide the physician more efficiently through the analysis depending on the detected condition, which may ultimately improve work flow and save analysis time.

Objectives: One challenge in arterial spin labeling (ASL) is the high variability of arterial transit times (ATT), which causes associated arterial transit delay (ATD) artifacts. In patients with pathological changes, these artifacts occur when post-labeling delay (PLD) and bolus durations are not optimally matched to the subject, resulting in difficult quantification of cerebral blood flow (CBF) and ATT. This is also true for the free lunch approach in Hadamard-encoded pseudocontinuous ASL (H-pCASL).

Material and methods: Five healthy volunteers were scanned with a 3 T MR-system. pCASL-subbolus timing was adjusted individually by the developed adaptive Walsh-ordered pCASL sequence and an automatic feedback algorithm. The quantification results for CBF and ATT and the respective standard deviations were compared with results obtained using recommended timings and intentionally suboptimal timings.

Results: The algorithm individually adjusted the pCASL-subbolus PLD for each subject within the range of recommended timing for healthy subjects, with a mean intra-subject adjustment deviation of 47.15 ms for single-shot and 44.5 ms for segmented acquisition in three repetitions.

Discussion: A first positive assessment of the results was performed on healthy volunteers. The extent to which the results can be transferred to patients and are of benefit must be investigated in follow-up studies.

Objective: First implementation of dynamic oxygen-17 (¹⁷O) MRI at 7 Tesla (T) during neuronal stimulation in the human brain.

Methods: Five healthy volunteers underwent a three-phase ¹⁷O gas (¹⁷O₂) inhalation experiment. Combined right-side visual stimulus and right-hand finger tapping were used to achieve neuronal stimulation in the left cerebral hemisphere. Data analysis included the evaluation of the relative partial volume (PV)-corrected time evolution of absolute ¹⁷O water (H₂¹⁷O) concentration and of the relative signal evolution without PV correction. Statistical analysis was performed using a one-tailed paired t test. Blood oxygen level-dependent (BOLD) experiments were performed to validate the stimulation paradigm.

Results: The BOLD maps showed significant activity in the stimulated left visual and sensorimotor cortex compared to the non-stimulated right side. PV correction of ¹⁷O MR data resulted in high signal fluctuations with a noise level of 10% due to small regions of interest (ROI), impeding further quantitative analysis. Statistical evaluation of the relative H₂¹⁷O signal with PV correction (p = 0.168) and without (p = 0.382) did not show significant difference between the stimulated left and non-stimulated right sensorimotor ROI.

Discussion: The change of cerebral oxygen metabolism induced by sensorimotor and visual stimulation is not large enough to be reliably detected with the current setup and methodology of dynamic ¹⁷O MRI at 7 T.