Analysis and Evaluation of a Deep Learning Reconstruction Approach with Denoising for Orthopedic MRI.

PURPOSE: To evaluate two settings (noise reduction of 50% or 75%) of a deep learning (DL) reconstruction model relative to each other and to conventional MR image reconstructions on clinical orthopedic MRI datasets. MATERIALS AND METHODS: This retrospective study included 54 patients who underwent two-dimensional fast spin-echo MRI for hip (n = 22; mean age, 44 years ± 13 [standard deviation]; nine men) or shoulder (n = 32; mean age, 56 years ± 17; 17 men) conditions between March 2019 and June 2020. MR images were reconstructed with conventional methods and the vendor-provided and commercially available DL model applied with 50% and 75% noise reduction settings (DL 50 and DL 75, respectively). Quantitative analytics, including relative anatomic edge sharpness, relative signal-to-noise ratio (rSNR), and relative contrast-to-noise ratio (rCNR) were computed for each dataset. In addition, the image sets were randomized, blinded, and presented to three board-certified musculoskeletal radiologists for ranking based on overall image quality and diagnostic confidence. Statistical analysis was performed with a nonparametric hypothesis comparing derived quantitative metrics from each reconstruction approach. In addition, inter- and intrarater agreement analysis was performed on the radiologists' rankings. RESULTS: Both denoising settings of the DL reconstruction showed improved edge sharpness, rSNR, and rCNR relative to the conventional reconstructions. The reader rankings demonstrated strong agreement, with both DL reconstructions outperforming the conventional approach (Gwet agreement coefficient = 0.98). However, there was lower agreement between the readers on which DL reconstruction denoising setting produced higher-quality images (Gwet agreement coefficient = 0.31 for DL 50 and 0.35 for DL 75). CONCLUSION: The vendor-provided DL MRI reconstruction showed higher edge sharpness, rSNR, and rCNR in comparison with conventional methods; however, optimal levels of denoising may need to be further assessed.Keywords: MRI Reconstruction Method, Deep Learning, Image Analysis, Signal-to-Noise Ratio, MR-Imaging, Neural Networks, Hip, Shoulder, Physics, Observer Performance, Technology Assessment Supplemental material is available for this article. © RSNA, 2021.

SMS MUSSELS: A navigator-free reconstruction for simultaneous multi-slice-accelerated multi-shot diffusion weighted imaging.

PURPOSE: To introduce a novel reconstruction method for simultaneous multi-slice (SMS)-accelerated multi-shot diffusion weighted imaging (ms-DWI). METHODS: SMS acceleration using blipped-CAIPI schemes have been proposed to speed up the acquisition of ms-DWIs. The reconstruction of the data requires (a) phase compensation to combine data from different shots and (b) slice unfolding to separate the data of different slices. The traditional approaches first estimate the phase maps corresponding to each shot and slice which are then employed to iteratively recover the slice unfolded DWIs without phase artifacts. In contrast, the proposed reconstruction directly recovers the slice-unfolded k-space data of the multiple shots for each slice in a single-step recovery scheme. The proposed method is enabled by the low-rank property inherent in the k-space samples of ms-DW acquisition. This enabled to formulate a joint recovery scheme that simultaneously (a) unfolds the k-space data of each slice using a SENSE-based scheme and (b) recover the missing k-space samples in each slice of the multi-shot acquisition employing a structured low-rank matrix completion. Additional smoothness regularization is also utilized for higher acceleration factors. The proposed joint recovery is tested on simulated and in vivo data and compared to similar un-navigated methods. RESULTS: Our experiments show effective slice unfolding and successful recovery of DWIs with minimal phase artifacts using the proposed method. The performance is comparable to existing methods at low acceleration factors and better than existing methods for higher acceleration factors. CONCLUSIONS: For the slice accelerations considered in this study, the proposed method can successfully recover DWIs from SMS-accelerated ms-DWI acquisitions.

In-phase zero TE musculoskeletal imaging.

PURPOSE: To introduce a new method for in-phase zero TE (ipZTE) musculoskeletal MR imaging. METHODS: ZTE is a 3D radial imaging method, which is sensitive to chemical shift off-resonance signal interference, especially around fat-water tissue interfaces. The ipZTE method addresses this fat-water chemical shift artifact by acquiring each 3D radial spoke at least twice with varying readout gradient amplitude and hence varying effective sampling time. Using k-space-based chemical shift decomposition, the acquired data is then reconstructed into an in-phase ZTE image and an out-of-phase disturbance. RESULTS: The ipZTE method was tested for knee, pelvis, brain, and whole-body. The obtained images demonstrate exceptional soft-tissue uniformity free from out-of-phase disturbances apparent in the original ZTE images. The chemical shift decomposition was found to improve SNR at the cost of reduced image resolution. CONCLUSION: The ipZTE method can be used as an averaging mechanism to eliminate fat-water chemical shift artifacts and improve SNR. The method is expected to improve ZTE-based musculoskeletal imaging and pseudo CT conversion as required for PET/MR attenuation correction and MR-guided radiation therapy planning.

Ultrashort echo time and zero echo time MRI at 7T.

OBJECTIVE: Zero echo time (ZTE) and ultrashort echo time (UTE) pulse sequences for MRI offer unique advantages of being able to detect signal from rapidly decaying short-T2 tissue components. In this paper, we applied 3D ZTE and UTE pulse sequences at 7T to assess differences between these methods. MATERIALS AND METHODS: We matched the ZTE and UTE pulse sequences closely in terms of readout trajectories and image contrast. Our ZTE used the water- and fat-suppressed solid-state proton projection imaging method to fill the center of k-space. Images from healthy volunteers obtained at 7T were compared qualitatively, as well as with SNR and CNR measurements for various ultrashort, short, and long-T2 tissues. RESULTS: We measured nearly identical contrast-to-noise and signal-to-noise ratios (CNR/SNR) in similar scan times between the two approaches for ultrashort, short, and long-T2 components in the brain, knee and ankle. In our protocol, we observed gradient fidelity artifacts in UTE, and our chosen flip angle and readout also resulted in shading artifacts in ZTE due to inadvertent spatial selectivity. These can be corrected by advanced reconstruction methods or with different chosen protocol parameters. CONCLUSION: The applied ZTE and UTE pulse sequences achieved similar contrast and SNR efficiency for volumetric imaging of ultrashort-T2 components. Key differences include that ZTE is limited to volumetric imaging, but has substantially reduced acoustic noise levels during the scan. Meanwhile, UTE has higher acoustic noise levels and greater sensitivity to gradient fidelity, but offers more flexibility in image contrast and volume selection.

Field shaping arrays: a means to address shading in high field breast MRI.

PURPOSE: To develop a simple correction approach to mitigate shading in 3 Tesla (T) breast MRI. MATERIALS AND METHODS: A slightly modified breast receive (Rx) array, which we termed field shaping array (FSA), was shown to mitigate breast shading at 3T. In this FSA, one Rx element was selectively unblocked and tuned off the Larmor frequency during the transmit (Tx) phase. The current flowing in this element during Tx created a secondary transmit field; the vector addition of this field and the one created directly by the body coil resulted in a more uniform excitation profile over the entire breast area. The receive Rx element was returned to its intended tuning during the Rx phase, ensuring unperturbed signal reception. RESULTS: Using the FSA, improved Tx field uniformity, better fat suppression, increased image homogeneity and reduced power deposition was seen in all volunteers studied. CONCLUSION: A simple modification of a standard breast Rx array, converting it to a field shaping array, was shown to mitigate breast shading in all volunteers studied.

Nonuniform and multidimensional Shinnar-Le Roux RF pulse design method.

The Shinnar-Le Roux (SLR) radiofrequency (RF) pulse design algorithm is widely used for designing slice-selective RF pulses due to its intuitiveness, optimality, and speed. SLR is limited, however, in that it is only capable of designing one-dimensional pulses played along constant gradients. We present a nonuniform SLR RF pulse design framework that extends most of the capabilities of classical SLR to nonuniform gradient trajectories and multiple dimensions. Specifically, like classical SLR, the new method is a hard pulse approximation-based technique that uses filter design relationships to produce the lowest power RF pulse that satisfies target magnetization ripple levels. The new method is validated and compared with methods conventionally used for nonuniform and multidimensional large-tip-angle RF pulse design.

Designing multichannel, multidimensional, arbitrary flip angle RF pulses using an optimal control approach.

The vast majority of parallel transmission RF pulse designs so far are based on small-tip-angle (STA) approximation of the Bloch equation. These methods can design only excitation pulses with small flip angles (e.g., 30 degrees ). The linear class large-tip-angle (LCLTA) method is able to design large-tip-angle parallel transmission pulses through concatenating a sequence of small-excitation pulses when certain k-space trajectories are used. However, both STA and LCLTA are linear approximations of the nonlinear Bloch equation. Therefore, distortions from the ideal magnetization profiles due to the higher order terms can appear in the final magnetization profiles. This issue is addressed in this work by formulating the multidimensional multichannel RF pulse design as an optimal control problem with multiple controls based directly on the Bloch equation. Necessary conditions for the optimal solution are derived and a first-order gradient optimization algorithm is used to iteratively solve the optimal control problem, where an existing pulse is used as an initial "guess." A systematic design procedure is also presented. Bloch simulation and phantom experimental results using various parallel transmission pulses (excitation, inversion, and refocusing) are shown to illustrate the effectiveness of the optimal control method in improving the spatial localization or homogeneity of the magnetization profiles.

A noniterative method to design large-tip-angle multidimensional spatially-selective radio frequency pulses for parallel transmission.

Recently, theoretical and experimental work has shown that parallel transmission of RF pulses can be used to shorten the duration of multidimensional spatially-selective pulses and compensate for B(1) field inhomogeneity. However, all the existing noniterative methods can design only excitation pulses for parallel transmission with a small flip angle (e.g., 30 degrees , or at most 90 degrees ) and cannot design large-tip-angle inversion/refocusing pulses, because these methods are based on the small-tip-angle (STA) approximation of the Bloch equation. In this work, a method to design large-tip-angle multidimensional spatially-selective pulses for parallel transmission is proposed, based on an extension of the single-channel linear-class large-tip-angle (LCLTA) theory. Design examples of 2D refocusing and inversion parallel transmit pulses and magnetization profiles from Bloch equation simulations demonstrate the strength of the proposed method. A 2D spin-echo parallel transmission experiment on a slab phantom using a 180 degrees refocusing pulse with an eight-channel transmit-only array further validates the effectiveness of the proposed method.

SENSE optimization of a transceive surface coil array for MRI at 4 T.

Parallel imaging (PI) techniques employ the use of multiple radiofrequency (RF) channels to transmit and/or receive the NMR signal. In the current study we use a finite difference time domain (FDTD) method to simulate the electromagnetic fields of a RF coil array operating in transmit-receive (transceive) mode and receive-only mode. Optimization of these configurations for PI is studied as well. Our results suggest that a coil array can effectively be used for transceive or receive-only PI techniques. For a head coil configuration, the sensitivity encoding (SENSE) optimized coil array gap size and PI acceleration factor for MRI are shown to be a function of the physiological-to-intrinsic-noise ratio (PhINR) with a much stronger dependence on acceleration factor than gap size. The results provide a means to optimize any PI sequence by varying the acceleration factor based on the measured PhINR. In addition, an example design for an eight-element transceive coil array for heads at 4 T is given.

Radiofrequency power deposition utilizing thermal imaging.

Wavelength effects influence radiofrequency (RF) power deposition distributions and limit magnetic resonance (MR) medical applications at very high magnetic fields. The power depositions in spherical saline gel phantoms were deduced from proton resonance shift thermal maps at both 1.5 T and 3.0 T over a range of conductivities. Phase differences before and after RF heating were measured for both a quadrature head coil and a circular surface coil. A long echo time (TE) pulse sequence with a 3D phase unwrap algorithm provided increased thermal sensitivity. The measured thermal maps agreed with a model of eddy-current heating by circularly polarized oscillating RF fields in a conducting dielectric sphere. At 3.0 T, thermal maps were acquired with a <0.32 degrees C temperature rise at 4 W. Proton resonance shift thermal maps provided a measure of hot spots in very-high-field MR imaging (MRI), in which both the phase sensitivity and signal-to-noise ratio (SNR) were increased. The method provides a means of studying the heat distribution generated by RF coils excited by clinical pulse sequences.

Peripheral nerve stimulation properties of head and body gradient coils of various sizes.

Peripheral nerve stimulation (PNS) caused by time-varying magnetic fields has been studied both theoretically and experimentally. A human volunteer study performed on three different body-size gradient coils and one head-size gradient coil is presented in this work. The experimental results were used to generate average PNS threshold parameters for the tested gradient systems. It was found that the average stimulation threshold increases while gradient-region-of-uniformity size decreases. In addition, linear relationships between PNS parameters and diameter of homogeneous gradient spherical volume (DSV) were discovered: SR(min) and DeltaG(min) both vary inverse linearly with DSV. More importantly, the chronaxie value was found to vary inversely linearly with the DSV. This finding indicates that, contrary to the general understanding, the parameter "chronaxie" in the commonly accepted simple stimulation models cannot be considered to be a single-value, nerve-specific constant. A modified linear model for gradient-induced PNS based on these results was developed, which may permit, for the first time, the general prediction of nerve stimulation properties for gradient coils of arbitrary linear region dimension.

Characterization of benign and metastatic vertebral compression fractures with quantitative diffusion MR imaging.

BACKGROUND AND PURPOSE: Conventional imaging techniques cannot be used to unambiguously and reliably differentiate malignant from benign vertebral compression fractures. Our hypothesis is that these malignant and benign vertebral lesions can be better distinguished on the basis of tissue apparent diffusion coefficients (ADCs). The purpose of this study was to test this hypothesis by using a quantitative diffusion imaging technique. METHODS: Twenty-seven patients with known cancer and suspected metastatic vertebral lesions underwent 1.5-T conventional T1-weighted, T2-weighted, and contrast-enhanced T1-weighted imaging to identify the lesions. Diffusion-weighted images of the areas of interest were acquired by using a fast spin-echo diffusion pulse sequence with b values of 0-250 s/mm(2). The abnormal regions on the diffusion-weighted images were outlined by using the conventional images as guides, and the ADC values were calculated. On the basis of pathologic results and clinical findings, the cases were divided into two categories: benign compression fractures and metastatic lesions. The ADC values for each category were combined and plotted as histograms; this procedure was followed by statistical analysis. RESULTS: The patient group had 12 benign fractures and 15 metastases. The mean ADC values, as obtained from the histograms, were (1.9 +/- 0.3) x 10(-4) mm(2)/s and (3.2 +/- 0.5) x 10(-4) mm(2)/s for metastases and benign fractures, respectively. CONCLUSION: Our results indicate that quantitative ADC mapping, instead of qualitative diffusion-weighted imaging, can provide valuable information in differentiating benign vertebral fractures from metastatic lesions.