A Systematic Review and Identification of the Challenges of Deep Learning Techniques for Undersampled Magnetic Resonance Image Reconstruction.

Deep learning (DL) in magnetic resonance imaging (MRI) shows excellent performance in image reconstruction from undersampled k-space data. Artifact-free and high-quality MRI reconstruction is essential for ensuring accurate diagnosis, supporting clinical decision-making, enhancing patient safety, facilitating efficient workflows, and contributing to the validity of research studies and clinical trials. Recently, deep learning has demonstrated several advantages over conventional MRI reconstruction methods. Conventional methods rely on manual feature engineering to capture complex patterns and are usually computationally demanding due to their iterative nature. Conversely, DL methods use neural networks with hundreds of thousands of parameters and automatically learn relevant features and representations directly from the data. Nevertheless, there are some limitations to DL-based techniques concerning MRI reconstruction tasks, such as the need for large, labeled datasets, the possibility of overfitting, and the complexity of model training. Researchers are striving to develop DL models that are more efficient, adaptable, and capable of providing valuable information for medical practitioners. We provide a comprehensive overview of the current developments and clinical uses by focusing on state-of-the-art DL architectures and tools used in MRI reconstruction. This study has three objectives. Our main objective is to describe how various DL designs have changed over time and talk about cutting-edge tactics, including their advantages and disadvantages. Hence, data pre- and post-processing approaches are assessed using publicly available MRI datasets and source codes. Secondly, this work aims to provide an extensive overview of the ongoing research on transformers and deep convolutional neural networks for rapid MRI reconstruction. Thirdly, we discuss several network training strategies, like supervised, unsupervised, transfer learning, and federated learning for rapid and efficient MRI reconstruction. Consequently, this article provides significant resources for future improvement of MRI data pre-processing and fast image reconstruction.

Straightforward Magnetic Resonance Temperature Measurements Combined with High Frame Rate and Magnetic Susceptibility Correction.

Proton resonance frequency shift (PRFS) is an MRI-based simple temperature mapping method that exhibits higher spatial and temporal resolution than temperature mapping methods based on T1 relaxation time and diffusion. PRFS temperature measurements are validated against fiber-optic thermal sensors (FOSs). However, the use of FOSs may introduce temperature errors, leading to both underestimation and overestimation of PRFS measurements, primarily due to material susceptibility changes caused by the thermal sensors. In this study, we demonstrated susceptibility-corrected PRFS (scPRFS) with a high frame rate and accuracy for suitably distributed temperatures. A single-echo-based background removal technique was employed for phase variation correction, primarily owing to magnetic susceptibility, which enabled fast temperature mapping. The scPRFS was used to validate the temperature fidelity by comparing the temperatures of fiber-optic sensors and conventional PRFS through phantom-mimicked human and ex vivo experiments. This study demonstrates that scPRFS measurements in agar-gel are in good agreement with the thermal sensor readings, with a root mean square error (RMSE) of 0.33-0.36 °C in the phantom model and 0.12-0.16 °C in the ex vivo experiment. These results highlight the potential of scPRFS for precise thermal monitoring and ablation in both low- and high-temperature non-invasive therapies.

Proposed Safety Guidelines for Patient Assistants in an Open MRI Environment.

The wide-open side of an open magnetic resonance imaging (MRI) system allows a patient to easily contact the patient assistant during MRI scans. A wide-open-shaped magnet is highly effective when interventional procedures are necessary. Patient assistants can provide comfort by holding a part of the patient's body. Because current regulations or guidelines are concerned with only patient radio frequency (RF) safety, investigations on the safety of patient assistants exposed to high-magnetic field MRI (up to 1.2 T) are required. In this study, five different poses of patient assistants were numerically simulated at a 1.2 T open MRI system to determine the impact of poses on the RF exposure level. The 10-g averaged specific absorption rate (SAR) levels were analyzed for the poses of each patient assistant wearing gloves. Compared with the patient, up to 29.8% of the patient SAR was observed in the patient assistant. When the patient assistant wore latex gloves, a 63.7% reduction in the 10-g averaged SAR level was observed, which could be a remedy to minimize possible RF hazards. To prevent possible RF hazards during MRI scans, certain clauses regarding the patient assistant's poses or wearing gloves must be added to the existing MRI screening forms.

Evaluation of the Abdominal Aorta and External Iliac Arteries Using Three-Dimensional Time-of-Flight, Three Dimensional Electrocardiograph-Gated Fast Spin-Echo, and Contrast-Enhanced Magnetic Resonance Angiography in Clinically Healthy Cats.

Arterial thromboembolism is associated with high morbidity and mortality rates in cats. Definitive diagnosis requires advanced imaging modalities, such as computed tomography angiography (CTA) and contrast-enhanced (CE) magnetic resonance angiography (MRA). However, CTA involves exposure to a large amount of ionized radiation, and CE-MRA can cause systemic nephrogenic fibrosis. Non-contrast-enhanced (NE) MRA can help accurately diagnose vascular lesions without such limitations. In this study, we evaluated the ability of NE-MRA using three-dimensional electrocardiograph-gated fast spin-echo (3D ECG-FSE) and 3D time-of-flight (3D TOF) imaging to visualize the aorta and external iliac arteries in clinically healthy cats and compared the results with those obtained using CE-MRA. All 11 cats underwent 3D ECG-FSE, 3D TOF, and CE-MRA sequences. Relative signal intensity (rSI) for quantitative image analysis and image quality scores (IQS) for qualitative image analysis were assessed; the rSI values based on the 3D TOF evaluations were significantly lower than those obtained using 3D ECG-FSE (aorta 3D TOF: 0.57 ± 0.06, aorta 3D ECG-FSE: 0.83 ± 0.06, P < 0.001; external iliac arteries 3D TOF: 0.45 ± 0.06, external iliac arteries 3D ECG-FSE:0.80 ± 0.05, P < 0.001) and similar to those obtained using CE-MRA (aorta: 0.58 ± 0.05, external iliac arteries: 0.57 ± 0.03). Moreover, IQS obtained using 3D TOF were significantly higher than those obtained using 3D ECG-FSE (aorta 3D TOF: 3.95 ± 0.15, aorta 3D ECG-FSE: 2.32 ± 0.60, P < 0.001; external iliac arteries 3D ECG-FSE: 3.98 ± 0.08, external iliac arteries 3D ECG-FSE: 2.23 ± 0.56, P < 0.001) and similar to those obtained using CE-MRA (aorta: 3.61 ± 0.41, external iliac arteries: 3.57 ± 0.41). Thus, 3D TOF is more suitable and produces consistent image quality for visualizing the aorta and external iliac arteries in clinically healthy cats and this will be of great help in the diagnosis of FATE.

In Vivo Magnetic Resonance Thermometry for Brain and Body Temperature Variations in Canines under General Anesthesia.

The core body temperature tends to decrease under general anesthesia. Consequently, monitoring the core body temperature during procedures involving general anesthesia is essential to ensure patient safety. In veterinary medicine, rectal temperature is used as an indicator of the core body temperature, owing to the accuracy and convenience of this approach. Some previous studies involving craniotomy reported differences between the brain and core temperatures under general anesthesia. However, noninvasive imaging techniques are required to ascertain this because invasive brain temperature measurements can cause unintended temperature changes by inserting the temperature sensors into the brain or by performing the surgical operations. In this study, we employed in vivo magnetic resonance thermometry to observe the brain temperatures of patients under general anesthesia using the proton resonance frequency shift method. The rectal temperature was also recorded using a fiber optic thermometer during the MR thermometry to compare with the brain temperature changes. When the rectal temperature decreased by 1.4 ± 0.5 °C (mean ± standard deviation), the brain temperature (white matter) decreased by 4.8 ± 0.5 °C. Furthermore, a difference in the temperature reduction of the different types of brain tissue was observed; the reduction in the temperature of white matter exceeded that of gray matter mainly due to the distribution of blood vessels in the gray matter. We also analyzed and interpreted the core temperature changes with the body conditioning scores of subjects to see how the body weight affected the temperature changes.

Radio-Frequency Vector Magnetic Field Mapping in Magnetic Resonance Imaging.

A method is presented to measure the radio-frequency (RF) vector magnetic field inside an object using magnetic resonance imaging (MRI). Conventional " [Formula: see text] mapping" in MRI can measure the proton co-rotating component ( [Formula: see text] of the RF field produced by a transmit coil. Here we show that by repeating [Formula: see text] mapping on the same object and coil at multiple (8) specific orientations with respect to the main magnet, the magnitudes and relative phases of all (x, y, z) Cartesian components of the RF field can be determined unambiguously. We demonstrate the method on a circularly polarized volume coil and a loop coil tuned at 123.25 MHz in a 3 Tesla MRI scanner, with liquid phantoms. The volume coil measurement showed the axial component of the RF field, which is normally unmeasurable in MRI, away from the center of the coil. The measured RF vector field maps of both coils compared favorably with numerical simulation, with volumetric normalized root-mean-square difference in the range of 7~20%. While the proposed method cannot be applied to human imaging at present, applications to phantoms and small animals could provide a useful experimental tool to validate RF simulation and verify certain assumptions in [Formula: see text] map-based electrical properties tomography (EPT).

Quantitative magnetic susceptibility assessed by 7T magnetic resonance imaging in Alzheimer's disease caused by streptozotocin administration.

Streptozotocin treatment has emerged as an alternative model of sporadic Alzheimer's disease (SAD). Streptozotocin-induced alterations in iron and calcium levels reflect magnetic susceptibility changes, while susceptibility distribution in the cerebral regions has not been reported yet. This study aimed to investigate susceptibility distribution in the limbic system after streptozotocin administration to cynomolgus monkeys for exploring informative SAD biomarkers. Quantitative susceptibility mapping (QSM) using 7T magnetic resonance imaging (MRI) was utilized to quantitatively compare the susceptibility distributions in monkeys with sporadic Alzheimer disease and age-matched healthy controls. Compared to healthy controls, overall susceptibility values differed in the SAD models. Notable substantial susceptibility changes were observed in the hypothalamus with a 4.38-time decrease (AD: -47.45±12.19 ppb, healthy controls: 14.02±9.51 ppb) and in the posterior parts of the corpus callosum with a 2.83-times increase (AD: 31.49±15.90 ppb; healthy controls: 11.13±4.02 ppb). These susceptibility alterations may reflect neuronal death, and could serve as key biomarkers in the SAD. These results may be useful for specifying AD pathologies such as cognitive and non-cognitive symptoms.

Complex difference constrained compressed sensing reconstruction for accelerated PRF thermometry with application to MRI-induced RF heating.

PURPOSE: Introduce a novel compressed sensing reconstruction method to accelerate proton resonance frequency shift temperature imaging for MRI-induced radiofrequency heating evaluation. METHODS: A compressed sensing approach that exploits sparsity of the complex difference between postheating and baseline images is proposed to accelerate proton resonance frequency temperature mapping. The method exploits the intra-image and inter-image correlations to promote sparsity and remove shared aliasing artifacts. Validations were performed on simulations and retrospectively undersampled data acquired in ex vivo and in vivo studies by comparing performance with previously published techniques. RESULTS: The proposed complex difference constrained compressed sensing reconstruction method improved the reconstruction of smooth and local proton resonance frequency temperature change images compared to various available reconstruction methods in a simulation study, a retrospective study with heating of a human forearm in vivo, and a retrospective study with heating of a sample of beef ex vivo. CONCLUSION: Complex difference based compressed sensing with utilization of a fully sampled baseline image improves the reconstruction accuracy for accelerated proton resonance frequency thermometry. It can be used to improve the volumetric coverage and temporal resolution in evaluation of radiofrequency heating due to MRI, and may help facilitate and validate temperature-based methods for safety assurance.

Measurement of SAR-induced temperature increase in a phantom and in vivo with comparison to numerical simulation.

PURPOSE: To compare numerically simulated and experimentally measured temperature increase due to specific energy absorption rate from radiofrequency fields. METHODS: Temperature increase induced in both a phantom and in the human forearm when driving an adjacent circular surface coil was mapped using the proton resonance frequency shift technique of magnetic resonance thermography. The phantom and forearm were also modeled from magnetic resonance image data, and both specific energy absorption rate and temperature change as induced by the same coil were simulated numerically. RESULTS: The simulated and measured temperature increase distributions were generally in good agreement for the phantom. The relative distributions for the human forearm were very similar, with the simulations giving maximum temperature increase about 25% higher than measured. CONCLUSION: Although a number of parameters and uncertainties are involved, it should be possible to use numerical simulations to produce reasonably accurate and conservative estimates of temperature distribution to ensure safety in magnetic resonance imaging. R01 EB006563

Bloch-based MRI system simulator considering realistic electromagnetic fields for calculation of signal, noise, and specific absorption rate.

PURPOSE: To describe and introduce new software capable of accurately simulating MR signal, noise, and specific absorption rate (SAR) given arbitrary sample, sequence, static magnetic field distribution, and radiofrequency magnetic and electric field distributions for each transmit and receive coil. THEORY AND METHODS: Using fundamental equations for nuclear precession and relaxation, signal reception, noise reception, and calculation of SAR, a versatile MR simulator was developed. The resulting simulator was tested with simulation of a variety of sequences demonstrating several common imaging contrast types and artifacts. The simulation of intravoxel dephasing and rephasing with both tracking of the first order derivatives of each magnetization vector and multiple magnetization vectors was examined to ensure adequate representation of the MR signal. A quantitative comparison of simulated and experimentally measured SNR was also performed. RESULTS: The simulator showed good agreement with our expectations, theory, and experiment. CONCLUSION: With careful design, an MR simulator producing realistic signal, noise, and SAR for arbitrary sample, sequence, and fields has been created. It is hoped that this tool will be valuable in a wide variety of applications.

An approach to rapid calculation of temperature change in tissue using spatial filters to approximate effects of thermal conduction.

We present an approach to performing rapid calculations of temperature within tissue by interleaving, at regular time intervals, 1) an analytical solution to the Pennes (or other desired) bioheat equation excluding the term for thermal conduction and 2) application of a spatial filter to approximate the effects of thermal conduction. Here, the basic approach is presented with attention to filter design. The method is applied to a few different cases relevant to magnetic resonance imaging, and results are compared to those from a full finite-difference (FD) implementation of the Pennes bioheat equation. It is seen that results of the proposed method are in reasonable agreement with those of the FD approach, with about 15% difference in the calculated maximum temperature increase, but are calculated in a fraction of the time, requiring less than 2% of the calculation time for the FD approach in the cases evaluated.

Permittivity and performance of dielectric pads with sintered ceramic beads in MRI: early experiments and simulations at 3 T.

Passive dielectric materials have been used to improve aspects of MRI by affecting the distribution of radiofrequency electromagnetic fields. Recently, interest in such materials has increased with the number of high-field MRI sites. Here, we introduce a new material composed of sintered high-permittivity ceramic beads in deuterated water. This arrangement maintains the ability to create flexible pads for conforming to individual subjects. The properties of the material are measured and the performance of the material is compared to previously used materials in both simulation and experiment at 3 T. Results show that both permittivity of the beads and effect on signal-to-noise ratio and required transmit power in MRI are greater than those of materials consisting of ceramic powder in water. Importantly, use of beads results in both higher permittivity and lower conductivity than use of powder.

Quadrature RF Coil for In Vivo Brain MRI of a Macaque Monkey in a Stereotaxic Head Frame.

We present a quadrature volume coil designed for brain imaging of a macaque monkey fixed in a sphinx position (facing down the bore) within a stereotactic frame at 3 T, where the position of the monkey and presence of the frame preclude use of existing coils. Requirements include the ability to position and remove the coil without disturbing the position of the monkey in the frame. A saddle coil and a solenoid were combined on a modified cylindrical former and connected in quadrature as to produce a homogeneous circularly polarized field throughout the brain. To allow the loops of the saddle coil to encompass the ear posts, partial disassembly and reassembly were facilitated by embedding pin and socket contacts into separate pieces of the former. Coil design included simulation of the electromagnetic fields for the coil containing a 3D model of a monkey's head. The resulting coil produced adequate homogeneity and signal-to-noise ratio throughout the brain.

Improved homogeneity of the transmit field by simultaneous transmission with phased array and volume coil.

PURPOSE: To improve the homogeneity of transmit volume coils at high magnetic fields (> or =4 T). Due to radiofrequency (RF) field/tissue interactions at high fields, 4 T to 8 T, the transmit profile from head-sized volume coils shows a distinctive pattern with relatively strong RF magnetic field B(1) in the center of the brain. MATERIALS AND METHODS: In contrast to conventional volume coils at high field strengths, surface coil phased arrays can provide increased RF field strength peripherally. In theory, simultaneous transmission from these two devices could produce a more homogeneous transmission field. To minimize interactions between the phased array and the volume coil, counter rotating current (CRC) surface coils consisting of two parallel rings carrying opposite currents were used for the phased array. RESULTS: Numerical simulations and experimental data demonstrate that substantial improvements in transmit field homogeneity can be obtained. CONCLUSION: We have demonstrated the feasibility of using simultaneous transmission with human head-sized volume coils and CRC phased arrays to improve homogeneity of the transmit RF B(1) field for high-field MRI systems.

Experimental and numerical assessment of MRI-induced temperature change and SAR distributions in phantoms and in vivo.

It is important to accurately characterize the heating of tissues due to the radiofrequency energy applied during MRI. This has led to an increase in the use of numerical methods to predict specific energy absorption rate distributions for safety assurance in MRI. To ensure these methods are accurate for actual MRI coils, however, it is necessary to compare to experimental results. Here, we report results of some recent efforts to experimentally map temperature change and specific energy absorption rate in a phantom and in vivo where the only source of heat is the radiofrequency fields produced by the imaging coil. Results in a phantom match numerical simulation well, and preliminary results in vivo show measurable temperature increase. With further development, similar methods may be useful for verifying numerical methods for predicting specific energy absorption rate distributions and in some cases for directly measuring temperature changes and specific energy absorption rate induced by the radiofrequency fields in MRI experiments.