The development of MRI systems operating at or above 7 T has provided researchers with a new window into the human body, yielding improved imaging speed, resolution and signal-to-noise ratio. In order to fully realise the potential of ultrahigh-field MRI, a range of technical hurdles must be overcome. The non-uniformity of the transmit field is one of such issues, as it leads to non-uniform images with spatially varying contrast. Parallel transmission (i.e. the use of multiple independent transmission channels) provides previously unavailable degrees of freedom that allow full spatial and temporal control of the radiofrequency (RF) fields. This review discusses the many ways in which these degrees of freedom can be used, ranging from making more uniform transmit fields to the design of subject-tailored RF pulses for both uniform excitation and spatial selection, and also the control of the specific absorption rate. © 2015 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd.© 2015 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd.

This study investigates whether two‐channel radiofrequency (RF) shimming can improve imaging without increasing specific absorption rate (SAR) for fetal MRI at 3T.

The accurate estimation of patient’s exposure to the radiofrequency (RF) electromagnetic field of magnetic resonance imaging (MRI) significantly depends on a precise individual anatomical model. In the study, we investigated the applicability of an efficient whole-body individual modelling method for the assessment of MRI RF exposure. The individual modelling method included a deformable human model and tissue simplification techniques. Besides its remarkable efficiency, this approach utilized only a low specific absorption rate (SAR) sequence or even no MRI scan to generate the whole-body individual model. Therefore, it substantially reduced the risk of RF exposure. The dosimetric difference of the individual modelling method was evaluated using the manually segmented human models. In addition, stochastic dosimetry using a surrogate model by polynomial chaos presented SAR variability due to body misalignment and tilt in the coil, which were frequently occurred in the practical scan. In conclusion, the dosimetric equivalence of the individual models was validated by both deterministic and stochastic dosimetry. The proposed individual modelling method allowed the physicians to quantify the patient-specific SAR while the statistical results enabled them to comprehensively weigh over the exposure risk and get the benefit of imaging enhancement by using the high-intensity scanners or the high-SAR sequences.

To investigate inter-subject variability of B, SAR and temperature rise in a database of human models using a local transmit array for 7 T cardiac imaging.Dixon images were acquired of 14 subjects and segmented in dielectric models with an eight-channel local transmit array positioned around the torso for cardiac imaging. EM simulations were done to calculate SAR distributions. Based on the SAR distributions, temperature simulations were performed for exposure times of 6 min and 30 min. Peak local SAR and temperature rise levels were calculated for different RF shim settings. A statistical analysis of the resulting peak local SAR and temperature rise levels was performed to arrive at safe power limits.For RF shim vectors with random phase and uniformly distributed power, a safe average power limit of 35.7 W was determined (first level controlled mode). When RF amplitude and phase shimming was performed on the heart, a safe average power limit of 35.0 W was found. According to Pennes' model, our numerical study suggests a very low probability of exceeding the absolute local temperature limit of 40 °C for a total exposure time of 6 min and a peak local SAR of 20 W/kg. For a 30 min exposure time at 20 W/kg, it was shown that the absolute temperature limit can be exceeded in the case where perfusion does not change with temperature.Safe power constraints were found for 7 T cardiac imaging with an eight-channel local transmit array, while considering the inter-subject variability of B, SAR and temperature rise.© 2021 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd.

Demonstrate ability to produce reasonable simulations of temperature using numerical models of the human body with a limited number of tissues.For both a male and female human body model, numerical simulations were used to calculate temperature distributions in three different models of the same human body: the original model with 35 tissues for the male model and 76 tissues for the female model, a simplified model having only three tissues (muscle, fat, and lung), and a simplified model having six tissues (muscle, fat, lung, bone, brain, and skin).Although a three-tissue model gave reasonable specific absorption rate estimates in comparison to an original with many more tissues, because of tissue-specific thermal and physiological properties that do not affect specific absorption rate, such as rate of perfusion by blood, the three-tissue model did not provide temperature distributions similar to those of the original model. Inclusion of a few additional tissues, as in the six-tissue model, produced results in much better agreement with those from the original model.Reasonable estimates of temperature can be simulated with a limited number of tissues, although this number is higher than the number of tissues required to produce reasonable simulations of specific absorption rate. For exposures primarily in the head and thorax, six tissues may be adequate for reasonable estimates of temperature.© 2021 International Society for Magnetic Resonance in Medicine.

Accurate prediction of specific absorption rate (SAR) for high field MRI is necessary to best exploit its potential and guarantee safe operation. To reduce the effort (time, complexity) of SAR simulations while maintaining robust results, the minimum requirements for the creation (segmentation, labeling) of human models and methods to reduce the time for SAR calculations for 7 Tesla MR-imaging are evaluated. The geometric extent of the model required for realistic head-simulations and the number of tissue types sufficient to form a reliable but simplified model of the human body are studied. Two models (male and female) of the virtual family are analyzed. Additionally, their position within the head-coil is taken into account. Furthermore, the effects of retuning the coils to different load conditions and the influence of a large bore radiofrequency-shield have been examined. The calculation time for SAR simulations in the head can be reduced by 50% without significant error for smaller model extent and simplified tissue structure outside the coil. Likewise, the model generation can be accelerated by reducing the number of tissue types. Local SAR can vary up to 14% due to position alone. This must be considered and sets a limit for SAR prediction accuracy. All these results are comparable between the two body models tested.Copyright © 2012 Wiley Periodicals, Inc.

\n Generative adversarial networks are a kind of artificial intelligence algorithm designed to solve the\n generative modeling\n problem. The goal of a generative model is to study a collection of training examples and learn the probability distribution that generated them. Generative Adversarial Networks (GANs) are then able to generate more examples from the estimated probability distribution. Generative models based on deep learning are common, but GANs are among the most successful generative models (especially in terms of their ability to generate realistic high-resolution images). GANs have been successfully applied to a wide variety of tasks (mostly in research settings) but continue to present unique challenges and research opportunities because they are based on game theory while most other approaches to generative modeling are based on optimization.\n

In multiphase coronary CT angiography (CTA), a series of CT images are taken at different levels of radiation dose during the examination. Although this reduces the total radiation dose, the image quality during the low-dose phases is significantly degraded. Recently, deep neural network approaches based on supervised learning technique have demonstrated impressive performance improvement over conventional model-based iterative methods for low-dose CT. However, matched low- and routine-dose CT image pairs are difficult to obtain in multiphase CT. To address this problem, we aim at developing a new deep learning framework.We propose an unsupervised learning technique that can remove the noise of the CT images in the low-dose phases by learning from the CT images in the routine dose phases. Although a supervised learning approach is not applicable due to the differences in the underlying heart structure in two phases, the images are closely related in two phases, so we propose a cycle-consistent adversarial denoising network to learn the mapping between the low- and high-dose cardiac phases.Experimental results showed that the proposed method effectively reduces the noise in the low-dose CT image while preserving detailed texture and edge information. Moreover, thanks to the cyclic consistency and identity loss, the proposed network does not create any artificial features that are not present in the input images. Visual grading and quality evaluation also confirm that the proposed method provides significant improvement in diagnostic quality.The proposed network can learn the image distributions from the routine-dose cardiac phases, which is a big advantage over the existing supervised learning networks that need exactly matched low- and routine-dose CT images. Considering the effectiveness and practicability of the proposed method, we believe that the proposed can be applied for many other CT acquisition protocols.© 2018 American Association of Physicists in Medicine.

Fully automatic tissue segmentation is an essential step to translate quantitative MRI techniques to clinical setting. The goal of this study was to develop a novel approach based on the generative adversarial networks for fully automatic segmentation of knee cartilage and meniscus.Defining proper loss function for semantic segmentation to enforce the learning of multiscale spatial constraints in an end-to-end training process is an open problem. In this work, we have used the conditional generative adversarial networks to improve segmentation performance of convolutional neural network, such as UNet alone by overcoming the problems caused by pixel-wise mapping based objective functions, and to capture cartilage features during the training of the network. Furthermore, the Dice coefficient and cross entropy losses were incorporated to the loss functions to improve the model performance. The model was trained and tested on 176, 3D DESS (double-echo steady-state) knee images from the Osteoarthritis Initiative data set.The proposed model provided excellent segmentation performance for cartilages with Dice coefficients ranging from 0.84 in patellar cartilage to 0.91 in lateral tibial cartilage, with an average Dice coefficient of 0.88. For meniscus segmentation, the model achieves 0.89 Dice coefficient for lateral meniscus and 0.87 Dice coefficient for medial meniscus. The results are superior to previously published automatic cartilage and meniscus segmentation methods based on deep learning models such as convolutional neural network.The proposed UNet-conditional generative adversarial networks based model demonstrated a fully automated segmentation method with high accuracy for knee cartilage and meniscus.© 2019 International Society for Magnetic Resonance in Medicine.

The management of local and global power deposition in human subjects (specific absorption rate, SAR) is a fundamental constraint to the application of parallel transmission (pTx) systems. Even though the pTx and single channel have to meet the same SAR requirements, the complex behavior of the spatial distribution of local SAR for transmission arrays poses problems that are not encountered in conventional single-channel systems and places additional requirements on pTx radio frequency pulse design. We propose a pTx pulse design method which builds on recent work to capture the spatial distribution of local SAR in numerical tissue models in a compressed parameterization in order to incorporate local SAR constraints within computation times that accommodate pTx pulse design during an in vivo magnetic resonance imaging scan. Additionally, the algorithm yields a protocol-specific ultimate peak in local SAR, which is shown to bound the achievable peak local SAR for a given excitation profile fidelity. The performance of the approach was demonstrated using a numerical human head model and a 7 Tesla eight-channel transmit array. The method reduced peak local 10 g SAR by 14-66% for slice-selective pTx excitations and 2D selective pTx excitations compared to a pTx pulse design constrained only by global SAR. The primary tradeoff incurred for reducing peak local SAR was an increase in global SAR, up to 34% for the evaluated examples, which is favorable in cases where local SAR constraints dominate the pulse applications.Copyright © 2011 Wiley Periodicals, Inc.