Instant tissue field and magnetic susceptibility mapping from MRI raw phase using Laplacian enhanced deep neural networks.

Quantitative susceptibility mapping (QSM) is an MRI post-processing technique that produces spatially resolved magnetic susceptibility maps from phase data. However, the traditional QSM reconstruction pipeline involves multiple non-trivial steps, including phase unwrapping, background field removal, and dipole inversion. These intermediate steps not only increase the reconstruction time but accumulates errors. This study aims to overcome existing limitations by developing a Laplacian-of-Trigonometric-functions (LoT) enhanced deep neural network for near-instant quantitative field and susceptibility mapping (i.e., iQFM and iQSM) from raw MRI phase data. The proposed iQFM and iQSM methods were compared with established reconstruction pipelines on simulated and in vivo datasets. In addition, experiments on patients with intracranial hemorrhage and multiple sclerosis were also performed to test the generalization of the proposed neural networks. The proposed iQFM and iQSM methods in healthy subjects yielded comparable results to those involving the intermediate steps while dramatically improving reconstruction accuracies on intracranial hemorrhages with large susceptibilities. High susceptibility contrast between multiple sclerosis lesions and healthy tissue was also achieved using the proposed methods. Comparative studies indicated that the most significant contributor to iQFM and iQSM over conventional multi-step methods was the elimination of traditional Laplacian unwrapping. The reconstruction time on the order of minutes for traditional approaches was shortened to around 0.1 s using the trained iQFM and iQSM neural networks.

Exposure of Infants to Gradient Fields in a Baby MRI Scanner.

In pediatric magnetic resonance imaging (MRI), infants are exposed to rapid, time-varying gradient magnetic fields, leading to electric fields induced in the body of infants and potential safety risks (e.g. peripheral nerve stimulation). In this numerical study, the in situ electric fields in infants induced by small-sized gradient coils for a 1.5 T MRI scanner were evaluated. The gradient coil set was specially designed for the efficient imaging of infants within a small-bore (baby) scanner. The magnetic flux density and induced electric fields by the small x, y, z gradient coils in an infant model (8-week-old with a mass of 4.3 kg) were computed using the scalar potential finite differences method. The gradient coils were driven by a 1 kHz sinusoidal waveform and also a trapezoidal waveform with a 250 µs rise time. The model was placed at different scan positions, including the head area (position I), chest area (position II), and body center (position III). It was found that the induced electric fields in most tissues exceeded the basic restrictions of the ICNIRP 2010 guidelines for both waveforms. The electric fields were similar in the region of interest for all coil types and model positions but different outside the imaging region. The y-coil induced larger electric fields compared with the x- and z- coils. Bioelectromagnetics. 43:69-80, 2022. © 2021 Bioelectromagnetics Society.

Accelerating quantitative susceptibility and R2* mapping using incoherent undersampling and deep neural network reconstruction.

Quantitative susceptibility mapping (QSM) and R2* mapping are MRI post-processing methods that quantify tissue magnetic susceptibility and transverse relaxation rate distributions. However, QSM and R2* acquisitions are relatively slow, even with parallel imaging. Incoherent undersampling and compressed sensing reconstruction techniques have been used to accelerate traditional magnitude-based MRI acquisitions; however, most do not recover the full phase signal, as required by QSM, due to its non-convex nature. In this study, a learning-based Deep Complex Residual Network (DCRNet) is proposed to recover both the magnitude and phase images from incoherently undersampled data, enabling high acceleration of QSM and R2* acquisition. Magnitude, phase, R2*, and QSM results from DCRNet were compared with two iterative and one deep learning methods on retrospectively undersampled acquisitions from six healthy volunteers, one intracranial hemorrhage and one multiple sclerosis patients, as well as one prospectively undersampled healthy subject using a 7T scanner. Peak signal to noise ratio (PSNR), structural similarity (SSIM), root-mean-squared error (RMSE), and region-of-interest susceptibility and R2* measurements are reported for numerical comparisons. The proposed DCRNet method substantially reduced artifacts and blurring compared to the other methods and resulted in the highest PSNR, SSIM, and RMSE on the magnitude, R2*, local field, and susceptibility maps. Compared to two iterative and one deep learning methods, the DCRNet method demonstrated a 3.2% to 9.1% accuracy improvement in deep grey matter susceptibility when accelerated by a factor of four. The DCRNet also dramatically shortened the reconstruction time of single 2D brain images from 36-140 seconds using conventional approaches to only 15-70 milliseconds.

Optimizing multicontrast MRI reconstruction with shareable feature aggregation and selection.

This paper proposes a new method for optimizing feature sharing in deep neural network-based, rapid, multicontrast magnetic resonance imaging (MC-MRI). Using the shareable information of MC images for accelerated MC-MRI reconstruction, current algorithms stack the MC images or features without optimizing the sharing protocols, leading to suboptimal reconstruction results. In this paper, we propose a novel feature aggregation and selection scheme in a deep neural network to better leverage the MC features and improve the reconstruction results. First, we propose to extract and use the shareable information by mapping the MC images into multiresolution feature maps with multilevel layers of the neural network. In this way, the extracted features capture complementary image properties, including local patterns from the shallow layers and semantic information from the deep layers. Then, an explicit selection module is designed to compile the extracted features optimally. That is, larger weights are learned to incorporate the constructive, shareable features; and smaller weights are assigned to the unshareable information. We conduct comparative studies on publicly available T2-weighted and T2-weighted fluid attenuated inversion recovery brain images, and the results show that the proposed network consistently outperforms existing algorithms. In addition, the proposed method can recover the images with high fidelity under 16 times acceleration. The ablation studies are conducted to evaluate the effectiveness of the proposed feature aggregation and selection mechanism. The results and the visualization of the weighted features show that the proposed method does effectively improve the usage of the useful features and suppress useless information, leading to overall enhanced reconstruction results. Additionally, the selection module can zero-out repeated and redundant features and improve network efficiency.

xQSM: quantitative susceptibility mapping with octave convolutional and noise-regularized neural networks.

Quantitative susceptibility mapping (QSM) provides a valuable MRI contrast mechanism that has demonstrated broad clinical applications. However, the image reconstruction of QSM is challenging due to its ill-posed dipole inversion process. In this study, a new deep learning method for QSM reconstruction, namely xQSM, was designed by introducing noise regularization and modified octave convolutional layers into a U-net backbone and trained with synthetic and in vivo datasets, respectively. The xQSM method was compared with two recent deep learning (QSMnet+ and DeepQSM) and two conventional dipole inversion (MEDI and iLSQR) methods, using both digital simulations and in vivo experiments. Reconstruction error metrics, including peak signal-to-noise ratio, structural similarity, normalized root mean squared error and deep gray matter susceptibility measurements, were evaluated for comparison of the different methods. The results showed that the proposed xQSM network trained with in vivo datasets achieved the best reconstructions of all the deep learning methods. In particular, it led to, on average, 32.3%, 25.4% and 11.7% improvement in the accuracy of globus pallidus susceptibility estimation for digital simulations and 39.3%, 21.8% and 6.3% improvements for in vivo acquisitions compared with DeepQSM, QSMnet+ and iLSQR, respectively. It also exhibited the highest linearity against different susceptibility intensity scales and demonstrated the most robust generalization capability to various spatial resolutions of all the deep learning methods. In addition, the xQSM method also substantially shortened the reconstruction time from minutes using MEDI to only a few seconds.

A dedicated eight-channel receive RF coil array for monkey brain MRI at 9.4 T.

The neuroimaging of nonhuman primates (NHPs) realised with magnetic resonance imaging (MRI) plays an important role in understanding brain structures and functions, as well as neurodegenerative diseases and pathological disorders. Theoretically, an ultrahigh field MRI (≥7 T) is capable of providing a higher signal-to-noise ratio (SNR) for better resolution; however, the lack of appropriate radiofrequency (RF) coils for 9.4 T monkey MRI undermines the benefits provided by a higher field strength. In particular, the standard volume birdcage coil at 9.4 T generates typical destructive interferences in the periphery of the brain, which reduces the SNR in the neuroscience-focused cortex region. Also, the standard birdcage coil is not capable of performing parallel imaging. Consequently, extended scan durations may cause unnecessary damage due to overlong anaesthesia. In this work, assisted by numerical simulations, an eight-channel receive RF coil array was specially designed and manufactured for imaging NHPs at 9.4 T. The structure and geometry of the proposed receive array was optimised with numerical simulations, so that the SNR enhancement region was particularly focused on monkey brain. Validated with rhesus monkey and cynomolgus monkey brain images acquired from a 9.4 T MRI scanner, the proposed receive array outperformed standard birdcage coil with higher SNR, mean diffusivity and fractional anisotropy values, as well as providing better capability for parallel imaging.

Adaptive SAR mass-averaging framework to improve predictions of local RF heating near a hip implant for parallel transmit at 7 T.

PURPOSE: Magnetic resonance imaging is used increasingly to scan patients with hip prostheses. We evaluated the reliability of 10 g-averaged specific absorption rate (SAR10g ) to predict radiofrequency (RF) heating in tissues surrounding a hip implant at 7 T in an 8-channel pTx hip coil. A new adaptive SAR mass-averaging method is proposed to improve the correlation between the distribution of mass-averaged SAR and that of tissue temperature. METHODS: Currently, RF safety standards for implants are based on temperature instead of SAR, as SAR has not been introduced with regard to exposure scenarios with implants. In this manuscript, however, adaptive SAR is proposed for fast and reliable exposure evaluation with implants, after its correlation with tissue temperature is verified. A framework to calculate adaptive SAR mass-averaging was introduced, which uses a different averaging mass in tissues surrounding the implants and was designed to prevent the temperature from exceeding 39ºC. Predictions from SAR10g and adaptive SAR were compared with thermal simulations. RESULTS: The SAR10g method failed to predict both the location and amplitude of heating in tissue near the metal implants. In some cases, the temperature far exceeded 39ºC even when SAR10g was only 70% of the maximum allowed 10 W/kg. The distributions of adaptive SAR and temperature matched in most of the configurations, and the temperature remained below 39ºC when adaptive SAR was constrained. CONCLUSION: Adaptive SAR can accurately monitor RF heating and could be used for parallel transmit at 7 T to supplement current standards.

A numerical and experimental study of RF shimming in the presence of hip prostheses using adaptive SAR at 3 T.

PURPOSE: Parallel transmission techniques in MRI have the potential to improve the image quality near metal implants at 3 T. However, current testing of implants only evaluates the risk of radiofrequency (RF) heating in phantoms in circularly polarized mode. We investigate the influence of changing the transmission settings in a 2-channel body coil on the peak temperature near 2 CoCrMo hip prostheses, using adaptive specific absorption rate (SAR) as an estimate of RF heating. METHODS: Adaptive SAR is a SAR averaging method that is optimized to correlate with thermal simulations and limit the temperature to 39°C near hip implants. The simulated peak temperature was compared when using whole-body SAR, SAR10g , and adaptive SAR as a constraint for the maximum allowed input power. Adaptive SAR was used as a fast estimate of temperature to evaluate the trade-off between good image quality and low heating near the hip implants. Electromagnetic simulations were validated by simulating and measuring B1 maps and electric fields in a phantom at 3 T. RESULTS: Simulations and measurements showed excellent agreement. Limiting whole-body SAR to 2 W/kg and SAR10g to 10 W/kg resulted in temperatures up to 49.3°C and 40.7°C near the hip implants after 30 minutes of RF exposure, respectively. Predictions based on adaptive SAR limited the temperature to 39°C, and allowed to improve the B1 field distribution while preventing peak temperatures near the hip implants. CONCLUSION: Significant RF heating can occur at 3 T near hip implants when parallel transmission is used. Adaptive SAR can be integrated in RF shimming algorithms to improve the uniformity and reduce heating.

An open 8-channel parallel transmission coil for static and dynamic 7T MRI of the knee and ankle joints at multiple postures.

PURPOSE: We present the initial in vivo imaging results of an open architecture eight-channel parallel transmission (pTx) transceive radiofrequency (RF) coil array that was designed and constructed for static and dynamic 7T MRI of the knee and ankle joints. METHODS: The pTx coil has a U-shaped dual-row configuration (200 mm overall length longitudinally) that allows static and dynamic imaging of the knee and ankle joints at various postures and during active movements. This coil structure, in combination with B1 shimming, allows flexible configuration of B1 transmit profiles, with good homogeneity over 120-mm regions of interest. This coil enabled high-resolution gradient echo (e.g., 3D dual-echo steady state [DESS] and 3D multiecho data image combination [MEDIC]) and turbo spin echo (TSE) imaging (e.g., with proton density weighting [PDw], PDw with fat saturation, and T1 and T2 weightings) with local RF energy absorption rates well below regulatory limits. RESULTS: High-resolution 2D and 3D image series (e.g., 0.3 mm in-plane resolution for TSE, 0.47 mm isotropic for DESS and MEDIC) were obtained from the knee and ankle joints with excellent tissue contrast. Dynamic imaging during continuous knee and ankle flexion-extension cycles were successfully acquired. CONCLUSION: The new open pTx coil array provides versatility for high-quality static and dynamic MRI of the knee and ankle joints at 7T. Magn Reson Med 79:1804-1816, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Radial magnetic resonance imaging (MRI) using a rotating radiofrequency (RF) coil at 9.4 T.

The rotating radiofrequency coil (RRFC) has been developed recently as an alternative approach to multi-channel phased-array coils. The single-element RRFC avoids inter-channel coupling and allows a larger coil element with better B1 field penetration when compared with an array counterpart. However, dedicated image reconstruction algorithms require accurate estimation of temporally varying coil sensitivities to remove artefacts caused by coil rotation. Various methods have been developed to estimate unknown sensitivity profiles from a few experimentally measured sensitivity maps, but these methods become problematic when the RRFC is used as a transceiver coil. In this work, a novel and practical radial encoding method is introduced for the RRFC to facilitate image reconstruction without the measurement or estimation of rotation-dependent sensitivity profiles. Theoretical analyses suggest that the rotation-dependent sensitivities of the RRFC can be used to create a uniform profile with careful choice of sampling positions and imaging parameters. To test this new imaging method, dedicated electronics were designed and built to control the RRFC speed and hence positions in synchrony with imaging parameters. High-quality phantom and animal images acquired on a 9.4 T pre-clinical scanner demonstrate the feasibility and potential of this new RRFC method.

The coil array method for creating a dynamic imaging volume.

PURPOSE: Gradient strength and speed are limited by peripheral nerve stimulation (PNS) thresholds. The coil array method allows the gradient field to be moved across the imaging area. This can help reduce PNS and provide faster imaging for image-guided therapy systems such as the magnetic resonance imaging-guided linear accelerator (MRI-linac). THEORY: The coil array is designed such that many coils produce magnetic fields, which combine to give the desired gradient profile. The design of the coil array uses two methods: either the singular value decomposition (SVD) of a set of field profiles or the electromagnetic modes of the coil surface. METHODS: Two whole-body coils and one experimental coil were designed to investigate the method. The field produced by the experimental coil was compared to simulated results. RESULTS: The experimental coil region of uniformity (ROU) was moved along the z axis as shown in simulation. The highest observed field deviation was 16.9% at the edge of the ROU with a shift of 35 mm. The whole-body coils showed a median field deviation across all offsets below 5% with an eight-coil basis when using the SVD design method. CONCLUSION: Experimental results show the feasibility of a moving imaging region within an MRI with a low number of coils in the array. Magn Reson Med 78:784-793, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Automated analysis of hip joint cartilage combining MR T2 and three-dimensional fast-spin-echo images.

PURPOSE: To validate a fully automated scheme to extract biochemical information from the hip joint cartilages using MR T2 mapping images incorporating segmentation of co-registered three-dimensional Fast-Spin-Echo (3D-SPACE) images. METHODS: Manual analyses of unilateral hip (3 Tesla) MR images of 24 asymptomatic volunteers were used to validate a 3D deformable model method for automated cartilage segmentation of SPACE scans, partitioning of the individual femoral and acetabular cartilage plates into clinically defined sub-regions and propagating these results to T2 maps to calculate region-wise T2 value statistics. Analyses were completed on a desktop computer (∼ 10 min per case). RESULTS: The mean voxel overlap between automated A and manual M segmentations of the cartilage volumes in the (clinically based) SPACE images was 73% (100 × 2|A∩M|/[|A|+|M|]). The automated and manual analyses demonstrated a relative difference error <10% in the median "T2 average signal" for each cartilage plate. The automated and manual analyses showed consistent patterns between significant differences in T2 data across the hip cartilage sub-regions. CONCLUSION: The good agreement between the manual and automatic analyses of T2 values indicates the use of structural 3D-SPACE MR images with the proposed method provides a promising approach for automated quantitative T2 assessment of hip joint cartilages.

Personal exposure to static and time-varying magnetic fields during MRI procedures in clinical practice in the UK.

BACKGROUND: MRI has developed into one of the most important medical diagnostic imaging modalities, but it exposes staff to static magnetic fields (SMF) when present in the vicinity of the MR system, and to radiofrequency and switched gradient electromagnetic fields if they are present during image acquisition. We measured exposure to SMF and motion-induced time-varying magnetic fields (TVMF) in MRI staff in clinical practice in the UK to enable extensive assessment of personal exposure levels and variability, which enables comparison to other countries. METHODS: 8 MRI facilities across National Health Service sites in England, Wales and Scotland were included, and staff randomly selected during the days when measurements were performed were invited to wear a personal MRI-compatible dosimeter and keep a diary to record all procedures and tasks performed during the measured shift. RESULTS: 98 participants, primarily radiographers (71%) but also other healthcare staff, anaesthetists and other medical staff were included, resulting in 149 measurements. Average geometric mean peak SMF and TVMF exposures were 448 mT (range 20-2891) and 1083 mT/s (9-12 355 mT/s), and were highest for radiographers (GM=559 mT and GM=734 mT/s). Time-weighted exposures to SMF and TVMF (GM=16 mT (range 5-64) and GM=14 mT/s (range 9-105)) and exposed-time-weighted exposures to SMF and TVMF (GM=27 mT (range 11-89) and GM=17 mT/s (range 9-124)) were overall relative low-primarily because staff were not in the MRI suite for most of their shifts-and did not differ significantly between occupations. CONCLUSIONS: These results are comparable to the few data available from the UK but they differ from recent data collected in the Netherlands, indicating that UK staff are exposed for shorter periods but to higher levels. These data indicate that exposure to SMF and TVMF from MRI scanners cannot be extrapolated across countries.

Numerical prediction of temperature elevation induced around metallic hip prostheses by traditional, split, and uniplanar gradient coils.

PURPOSE: The paper presents a computational study for the estimation of the temperature elevation occurring in a human subject carrying metallic hip prostheses when exposed to the magnetic field produced by gradient coils. METHODS: The simulations are performed through validated numerical codes, which solve the electromagnetic and thermal equations applied to a high-resolution anatomical human model. Three different sets of gradient coils (traditional, split and uniplanar) are considered to evaluate the maximum steady-state temperature elevation in the human body. This result is then rescaled to take into account the waveform of the signal, the duty-cycle and the duration of the scan. RESULTS: Several exposure situations obtained by changing the patient's position are analyzed, finding temperature elevations on the order of some degrees. CONCLUSION: The results are of possible concern and provide evidence of the need for further specific investigations aimed at assuring the safety of potential patients carrying metallic hip implants. Magn Reson Med 74:272-279, 2015. © 2014 Wiley Periodicals, Inc.

Transient health symptoms of MRI staff working with 1.5 and 3.0 Tesla scanners in the UK.

OBJECTIVES: Recent studies have consistently shown that amongst staff working with MRI, transient symptoms directly attributable to the MRI system including dizziness, nausea, tinnitus, and concentration problems are reported. This study assessed symptom prevalence and incidence in radiographers and other staff working with MRI in healthcare in the UK. METHODS: One hundred and four volunteer staff from eight sites completed a questionnaire and kept a diary to obtain information on subjective symptoms and work practices, and wore a magnetic field dosimeter during one to three randomly selected working days. Incidence of MRI-related symptoms was obtained for all shifts and prevalence of MRI-related and reference symptoms was associated to explanatory factors using ordinal regression. RESULTS: Incident symptoms related to working with MRI were reported in 4% of shifts. Prevalence of MRI-related, but not reference symptoms were associated with number of hours per week working with MRI, shift length, and stress, but not with magnetic field strength (1.5 and 3 T) or measured magnetic field exposure. CONCLUSIONS: Reporting of prevalent symptoms was associated with longer duration of working in MRI departments, but not with measured field strength of exposure. Other factors related to organisation and stress seem to contribute to increased reporting of MRI-related symptoms. KEY POINTS: • Routine work with MRI is associated with increased reporting of transient symptoms • No link to the strength of the magnetic field was demonstrated. • Organisational factors and stress additionally contribute to reporting of MRI-related symptoms.

Distance informed Track-Weighted Imaging (diTWI): a framework for sensitising streamline information to neuropathology.

Track-Weighted Imaging (TWI), where voxel intensity is based on image metrics encoded along streamline trajectories, provides a mechanism to study white matter disease. However, with generalised streamline weighting, it is difficult to localise the precise anatomical source and spread of injury or neuropathology. This limitation can be overcome by modulating the voxel weight based on the distance of the voxel from a given anatomical location along the tract, which we term diTWI: distance informed Track-Weighted Imaging. The location of known neuropathology can be delineated on any given imaging modality (e.g. MRI or PET). To demonstrate the clinical utility of this approach, we measured tumour cell infiltration along WM fibre tracts in 13 patients with newly diagnosed glioblastoma and 1 patient with Anaplastic Astrocytoma. TWI and diTWI maps were generated using information obtained from dynamic contrast enhanced MRI (area under the curve, AUC) and diffusivity maps (ADC and FA) with tumour boundaries automatically extracted using a logistic regression classifier. The accuracy of the derived tumour volumes was compared to those generated using 3,4-dihydroxy-6-[(18)F]-fluoro-l-phenylalanine (FDOPA) PET imaging. The accuracy of the tumour volumes generated from the diTWI maps was superior to volumes derived from the TWI, geometric distance or baseline AUC, FA and ADC maps. The relative overlap and relative dissimilarity rates for the diTWI generated tumour volumes after classification were found to be 82.3±15.3% (range 69.1-91.9) and 16.9±8.8% (range 7.9-37.5), respectively. These findings show that diTWI maps provide a useful framework for localising neuropathological processes occurring along WM pathways.

Multilayer integral method for simulation of eddy currents in thin volumes of arbitrary geometry produced by MRI gradient coils.

PURPOSE: This article aims to present a fast, efficient and accurate multi-layer integral method (MIM) for the evaluation of complex spatiotemporal eddy currents in nonmagnetic and thin volumes of irregular geometries induced by arbitrary arrangements of gradient coils. METHODS: The volume of interest is divided into a number of layers, wherein the thickness of each layer is assumed to be smaller than the skin depth and where one of the linear dimensions is much smaller than the remaining two dimensions. The diffusion equation of the current density is solved both in time-harmonic and transient domain. RESULTS: The experimentally measured magnetic fields produced by the coil and the induced eddy currents as well as the corresponding time-decay constants were in close agreement with the results produced by the MIM. Relevant parameters such as power loss and force induced by the eddy currents in a split cryostat were simulated using the MIM. CONCLUSION: The proposed method is capable of accurately simulating the current diffusion process inside thin volumes, such as the magnet cryostat. The method permits the priori-calculation of optimal pre-emphasis parameters. The MIM enables unified designs of gradient coil-magnet structures for an optimal mitigation of deleterious eddy current effects.

Fully automatic lesion segmentation in breast MRI using mean-shift and graph-cuts on a region adjacency graph.

PURPOSE: To present and evaluate a fully automatic method for segmentation (i.e., detection and delineation) of suspicious tissue in breast MRI. MATERIALS AND METHODS: The method, based on mean-shift clustering and graph-cuts on a region adjacency graph, was developed and its parameters tuned using multimodal (T1, T2, DCE-MRI) clinical breast MRI data from 35 subjects (training data). It was then tested using two data sets. Test set 1 comprises data for 85 subjects (93 lesions) acquired using the same protocol and scanner system used to acquire the training data. Test set 2 comprises data for eight subjects (nine lesions) acquired using a similar protocol but a different vendor's scanner system. Each lesion was manually delineated in three-dimensions by an experienced breast radiographer to establish segmentation ground truth. The regions of interest identified by the method were compared with the ground truth and the detection and delineation accuracies quantitatively evaluated. RESULTS: One hundred percent of the lesions were detected with a mean of 4.5 ± 1.2 false positives per subject. This false-positive rate is nearly 50% better than previously reported for a fully automatic breast lesion detection system. The median Dice coefficient for Test set 1 was 0.76 (interquartile range, 0.17), and 0.75 (interquartile range, 0.16) for Test set 2. CONCLUSION: The results demonstrate the efficacy and accuracy of the proposed method as well as its potential for direct application across different MRI systems. It is (to the authors' knowledge) the first fully automatic method for breast lesion detection and delineation in breast MRI.

Exposure to static and time-varying magnetic fields from working in the static magnetic stray fields of MRI scanners: a comprehensive survey in the Netherlands.

Clinical and research staff who work around magnetic resonance imaging (MRI) scanners are exposed to the static magnetic stray fields of these scanners. Although the past decade has seen strong developments in the assessment of occupational exposure to electromagnetic fields from MRI scanners, there is insufficient insight into the exposure variability that characterizes routine MRI work practice. However, this is an essential component of risk assessment and epidemiological studies. This paper describes the results of a measurement survey of shift-based personal exposure to static magnetic fields (SMF) (B) and motion-induced time-varying magnetic fields (dB/dt) among workers at 15 MRI facilities in the Netherlands. With the use of portable magnetic field dosimeters, >400 full-shift and partial shift exposure measurements were collected among various jobs involved in clinical and research MRI. Various full-shift exposure metrics for B and motion-induced dB/dt exposure were calculated from the measurements, including instantaneous peak exposure and time-weighted average (TWA) exposures. We found strong correlations between levels of static (B) and time-varying (dB/dt) exposure (r = 0.88-0.92) and between different metrics (i.e. peak exposure, TWA exposure) to express full-shift exposure (r = 0.69-0.78). On average, participants were exposed to MRI-related SMFs during only 3.7% of their work shift. Average and peak B and dB/dt exposure levels during the work inside the MRI scanner room were highest among technical staff, research staff, and radiographers. Average and peak B exposure levels were lowest among cleaners, while dB/dt levels were lowest among anaesthesiology staff. Although modest exposure variability between workplaces and occupations was observed, variation between individuals of the same occupation was substantial, especially among research staff. This relatively large variability between workers with the same job suggests that exposure classification based solely on job title may not be an optimal grouping strategy for epidemiological purposes.

Automated anomaly-aware 3D segmentation of bones and cartilages in knee MR images from the Osteoarthritis Initiative.

In medical image analysis, automated segmentation of multi-component anatomical entities, with the possible presence of variable anomalies or pathologies, is a challenging task. In this work, we develop a multi-step approach using U-Net-based models to initially detect anomalies (bone marrow lesions, bone cysts) in the distal femur, proximal tibia and patella from 3D magnetic resonance (MR) images in individuals with varying grades of knee osteoarthritis. Subsequently, the extracted data are used for downstream tasks involving semantic segmentation of individual bone and cartilage volumes as well as bone anomalies. For anomaly detection, U-Net-based models were developed to reconstruct bone volume profiles of the femur and tibia in images via inpainting so anomalous bone regions could be replaced with close to normal appearances. The reconstruction error was used to detect bone anomalies. An anomaly-aware segmentation network, which was compared to anomaly-naïve segmentation networks, was used to provide a final automated segmentation of the individual femoral, tibial and patellar bone and cartilage volumes from the knee MR images which contain a spectrum of bone anomalies. The anomaly-aware segmentation approach provided up to 58% reduction in Hausdorff distances for bone segmentations compared to the results from anomaly-naïve segmentation networks. In addition, the anomaly-aware networks were able to detect bone anomalies in the MR images with greater sensitivity and specificity (area under the receiver operating characteristic curve [AUC] up to 0.896) compared to anomaly-naïve segmentation networks (AUC up to 0.874).