DeepQSM - using deep learning to solve the dipole inversion for quantitative susceptibility mapping

A subject-specific unsupervised deep learning method for quantitative susceptibility mapping using implicit neural representation

Quantitative susceptibility mapping (QSM) is an MRI-based technique that estimates the underlying tissue magnetic susceptibility based on phase signal. Deep learning (DL)-based methods have shown promise in handling the challenging ill-posed inverse problem for QSM reconstruction. However, they require extensive paired training data that are typically unavailable and suffer from generalization problems. Recent model-incorporated DL approaches also overlook the non-local effect of the tissue phase in applying the source-to-field forward model due to patch-based training constraint, resulting in a discrepancy between the prediction and measurement and subsequently suboptimal QSM reconstruction. This study proposes an unsupervised and subject-specific DL method for QSM reconstruction based on implicit neural representation (INR), referred to as INR-QSM. INR has emerged as a powerful framework for learning a high-quality continuous representation of the signal (image) by exploiting its internal information without training labels. In INR-QSM, the desired susceptibility map is represented as a continuous function of the spatial coordinates, parameterized by a fully-connected neural network. The weights are learned by minimizing a loss function that includes a data fidelity term incorporated by the physical model and regularization terms. Additionally, a novel phase compensation strategy is proposed for the first time to account for the non-local effect of tissue phase in data consistency calculation to make the physical model more accurate. Our experiments show that INR-QSM outperforms traditional established QSM reconstruction methods and the compared unsupervised DL method both qualitatively and quantitatively, and is competitive against supervised DL methods under data perturbations.

Clinical applications of deep learning in neuroinflammatory diseases: A scoping review

BACKGROUND

Deep learning (DL) is an artificial intelligence technology that has aroused much excitement for predictive medicine due to its ability to process raw data modalities such as images, text, and time series of signals.

OBJECTIVES

Here, we intend to give the clinical reader elements to understand this technology, taking neuroinflammatory diseases as an illustrative use case of clinical translation efforts. We reviewed the scope of this rapidly evolving field to get quantitative insights about which clinical applications concentrate the efforts and which data modalities are most commonly used.

METHODS

We queried the PubMed database for articles reporting DL algorithms for clinical applications in neuroinflammatory diseases and the radiology.healthairegister.com website for commercial algorithms.

RESULTS

The review included 148 articles published between 2018 and 2024 and five commercial algorithms. The clinical applications could be grouped as computer-aided diagnosis, individual prognosis, functional assessment, the segmentation of radiological structures, and the optimization of data acquisition. Our review highlighted important discrepancies in efforts. The segmentation of radiological structures and computer-aided diagnosis currently concentrate most efforts with an overrepresentation of imaging. Various model architectures have addressed different applications, relatively low volume of data, and diverse data modalities. We report the high-level technical characteristics of the algorithms and synthesize narratively the clinical applications. Predictive performances and some common a priori on this topic are finally discussed.

CONCLUSION

The currently reported efforts position DL as an information processing technology, enhancing existing modalities of paraclinical investigations and bringing perspectives to make innovative ones actionable for healthcare.

Plug-and-Play latent feature editing for orientation-adaptive quantitative susceptibility mapping neural networks

Quantitative susceptibility mapping (QSM) is a post-processing technique for deriving tissue magnetic susceptibility distribution from MRI phase measurements. Deep learning (DL) algorithms hold great potential for solving the ill-posed QSM reconstruction problem. However, a significant challenge facing current DL-QSM approaches is their limited adaptability to magnetic dipole field orientation variations during training and testing. In this work, we propose a novel Orientation-Adaptive Latent Feature Editing (OA-LFE) module to learn the encoding of acquisition orientation vectors and seamlessly integrate them into the latent features of deep networks. Importantly, it can be directly Plug-and-Play (PnP) into various existing DL-QSM architectures, enabling reconstructions of QSM from arbitrary magnetic dipole orientations. Its effectiveness is demonstrated by combining the OA-LFE module into our previously proposed phase-to-susceptibility single-step instant QSM (iQSM) network, which was initially tailored for pure-axial acquisitions. The proposed OA-LFE-empowered iQSM, which we refer to as iQSM+, is trained in a simulated-supervised manner on a specially-designed simulation brain dataset. Comprehensive experiments are conducted on simulated and in vivo human brain datasets, encompassing subjects ranging from healthy individuals to those with pathological conditions. These experiments involve various MRI platforms (3T and 7T) and aim to compare our proposed iQSM+ against several established QSM reconstruction frameworks, including the original iQSM. The iQSM+ yields QSM images with significantly improved accuracies and mitigates artifacts, surpassing other state-of-the-art DL-QSM algorithms. The PnP OA-LFE module's versatility was further demonstrated by its successful application to xQSM, a distinct DL-QSM network for dipole inversion. In conclusion, this work introduces a new DL paradigm, allowing researchers to develop innovative QSM methods without requiring a complete overhaul of their existing architectures.

Machine learning-based amide proton transfer imaging using partially synthetic training data

PURPOSE

Machine learning (ML) has been increasingly used to quantify CEST effect. ML models are typically trained using either measured data or fully simulated data. However, training with measured data often lacks sufficient training data, whereas training with fully simulated data may introduce bias because of limited simulations pools. This study introduces a new platform that combines simulated and measured components to generate partially synthetic CEST data, and to evaluate its feasibility for training ML models to predict amide proton transfer (APT) effect.

METHODS

Partially synthetic CEST signals were created using an inverse summation of APT effects from simulations and the other components from measurements. Training data were generated by varying APT simulation parameters and applying scaling factors to adjust the measured components, achieving a balance between simulation flexibility and fidelity. First, tissue-mimicking CEST signals along with ground truth information were created using multiple-pool model simulations to validate this method. Second, an ML model was trained individually on partially synthetic data, in vivo data, and fully simulated data, to predict APT effect in rat brains bearing 9 L tumors.

RESULTS

Experiments on tissue-mimicking data suggest that the ML method using the partially synthetic data is accurate in predicting APT. In vivo experiments suggest that our method provides more accurate and robust prediction than the training using in vivo data and fully synthetic data.

CONCLUSION

Partially synthetic CEST data can address the challenges in conventional ML methods.

Quantitative susceptibility mapping through model-based deep image prior (MoDIP)

The data-driven approach of supervised learning methods has limited applicability in solving dipole inversion in Quantitative Susceptibility Mapping (QSM) with varying scan parameters across different objects. To address this generalization issue in supervised QSM methods, we propose a novel training-free model-based unsupervised method called MoDIP (Model-based Deep Image Prior). MoDIP comprises a small, untrained network and a Data Fidelity Optimization (DFO) module. The network converges to an interim state, acting as an implicit prior for image regularization, while the optimization process enforces the physical model of QSM dipole inversion. Experimental results demonstrate MoDIP's excellent generalizability in solving QSM dipole inversion across different scan parameters. It exhibits robustness against pathological brain QSM, achieving over 32 % accuracy improvement than supervised deep learning methods. It is also 33 % more computationally efficient and runs 4 times faster than conventional DIP-based approaches, enabling 3D high-resolution image reconstruction in under 4.5 min.

Recommended implementation of quantitative susceptibility mapping for clinical research in the brain: A consensus of the ISMRM electro-magnetic tissue properties study group

This article provides recommendations for implementing QSM for clinical brain research. It is a consensus of the International Society of Magnetic Resonance in Medicine, Electro-Magnetic Tissue Properties Study Group. While QSM technical development continues to advance rapidly, the current QSM methods have been demonstrated to be repeatable and reproducible for generating quantitative tissue magnetic susceptibility maps in the brain. However, the many QSM approaches available have generated a need in the neuroimaging community for guidelines on implementation. This article outlines considerations and implementation recommendations for QSM data acquisition, processing, analysis, and publication. We recommend that data be acquired using a monopolar 3D multi-echo gradient echo (GRE) sequence and that phase images be saved and exported in Digital Imaging and Communications in Medicine (DICOM) format and unwrapped using an exact unwrapping approach. Multi-echo images should be combined before background field removal, and a brain mask created using a brain extraction tool with the incorporation of phase-quality-based masking. Background fields within the brain mask should be removed using a technique based on SHARP or PDF, and the optimization approach to dipole inversion should be employed with a sparsity-based regularization. Susceptibility values should be measured relative to a specified reference, including the common reference region of the whole brain as a region of interest in the analysis. The minimum acquisition and processing details required when reporting QSM results are also provided. These recommendations should facilitate clinical QSM research and promote harmonized data acquisition, analysis, and reporting.

DF-QSM: Data Fidelity based Hybrid Approach for Improved Quantitative Susceptibility Mapping of the Brain

Quantitative Susceptibility Mapping (QSM) is an advanced magnetic resonance imaging (MRI) technique to quantify the magnetic susceptibility of the tissue under investigation. Deep learning methods have shown promising results in deconvolving the susceptibility distribution from the measured local field obtained from the MR phase. Although existing deep learning based QSM methods can produce high-quality reconstruction, they are highly biased toward training data distribution with less scope for generalizability. This work proposes a hybrid two-step reconstruction approach to improve deep learning based QSM reconstruction. The susceptibility map prediction obtained from the deep learning methods has been refined in the framework developed in this work to ensure consistency with the measured local field. The developed method was validated on existing deep learning and model-based deep learning methods for susceptibility mapping of the brain. The developed method resulted in improved reconstruction for MRI volumes obtained with different acquisition settings, including deep learning models trained on constrained (limited) data settings.

Neuroimaging of Parkinson's disease by quantitative susceptibility mapping

Parkinson's disease (PD) is a common neurodegenerative disease, and apart from a few rare genetic causes, its pathogenesis remains largely unclear. Recent scientific interest has been captured by the involvement of iron biochemistry and the disruption of iron homeostasis, particularly within the brain regions specifically affected in PD. The advent of Quantitative Susceptibility Mapping (QSM) has enabled non-invasive quantification of brain iron in vivo by MRI, which has contributed to the understanding of iron-associated pathogenesis and has the potential for the development of iron-based biomarkers in PD. This review elucidates the biochemical underpinnings of brain iron accumulation, details advancements in iron-sensitive MRI technologies, and discusses the role of QSM as a biomarker of iron deposition in PD. Despite considerable progress, several challenges impede its clinical application after a decade of QSM studies. The initiation of multi-site research is warranted for developing robust, interpretable, and disease-specific biomarkers for monitoring PD disease progression.

Incorporating a-priori information in deep learning models for quantitative susceptibility mapping via adaptive convolution

Quantitative susceptibility mapping (QSM) has attracted considerable interest for tissue characterization (e.g., iron and calcium accumulation, myelination, venous vasculature) in the human brain and relies on extensive data processing of gradient-echo MRI phase images. While deep learning-based field-to-susceptibility inversion has shown great potential, the acquisition parameters applied in clinical settings such as image resolution or image orientation with respect to the magnetic field have not been fully accounted for. Furthermore, the lack of comprehensive training data covering a wide range of acquisition parameters further limits the current QSM deep learning approaches. Here, we propose the integration of a priori information of imaging parameters into convolutional neural networks with our approach, adaptive convolution, that learns the mapping between the additional presented information (acquisition parameters) and the changes in the phase images associated with these varying acquisition parameters. By associating a-priori information with the network parameters itself, the optimal set of convolution weights is selected based on data-specific attributes, leading to generalizability towards changes in acquisition parameters. Moreover, we demonstrate the feasibility of pre-training on synthetic data and transfer learning to clinical brain data to achieve substantial improvements in the computation of susceptibility maps. The adaptive convolution 3D U-Net demonstrated generalizability in acquisition parameters on synthetic and in-vivo data and outperformed models lacking adaptive convolution or transfer learning. Further experiments demonstrate the impact of the side information on the adaptive model and assessed susceptibility map computation on simulated pathologic data sets and measured phase data.

A spatially adaptive regularization based three-dimensional reconstruction network for quantitative susceptibility mapping

Objective.Quantitative susceptibility mapping (QSM) is a new imaging technique for non-invasive characterization of the composition and microstructure ofin vivotissues, and it can be reconstructed from local field measurements by solving an ill-posed inverse problem. Even for deep learning networks, it is not an easy task to establish an accurate quantitative mapping between two physical quantities of different units, i.e. field shift in Hz and susceptibility value in ppm for QSM.Approach. In this paper, we propose a spatially adaptive regularization based three-dimensional reconstruction network SAQSM. A spatially adaptive module is specially designed and a set of them at different resolutions are inserted into the network decoder, playing a role of cross-modality based regularization constraint. Therefore, the exact information of both field and magnitude data is exploited to adjust the scale and shift of feature maps, and thus any information loss or deviation occurred in previous layers could be effectively corrected. The network encoding has a dynamic perceptual initialization, which enables the network to overcome receptive field intervals and also strengthens its ability to detect features of various sizes.Main results. Experimental results on the brain data of healthy volunteers, clinical hemorrhage and simulated phantom with calcification demonstrate that SAQSM can achieve more accurate reconstruction with less susceptibility artifacts, while perform well on the stability and generalization even for severe lesion areas.Significance. This proposed framework may provide a valuable paradigm to quantitative mapping or multimodal reconstruction.

A latent code based multi-variable modulation network for susceptibility mapping

Quantitative susceptibility mapping (QSM) is a technique for obtaining quantitative information on tissue susceptibility and has shown promising potential for clinical applications, in which the magnetic susceptibility is calculated by solving an ill-posed inverse problem. Recently, deep learning-based methods are proposed to address this issue, but the diversity of data distribution was not well considered, and thus the model generalization is limited in clinical applications. In this paper, we propose a Latent Code based Multi-Variable modulation network for QSM reconstruction (LCMnet). Particularly, a specific modulation module is exploited to incorporate three variables, i.e., field map, magnitude image, and initial susceptibility. The latent code in the modulated convolution is learned from feature maps of the field data using the encoder-decoder framework. The susceptibility map pre-estimated from simple thresholding is the constant input of the module, thereby enhancing the network stability and accelerating training convergence. As another input, multi-level features generated by a cross-fusion block integrate the information of field and magnitude data effectively. Experimental results on in vivo human brain data, challenge data, clinical data and synthetic data demonstrate that the proposed method LCMnet can achieve outstanding performance on accurate susceptibility measurement and also excellent generalization.

Training Universal Deep-Learning Networks for Electromagnetic Medical Imaging Using a Large Database of Randomized Objects

Deep learning has become a powerful tool for solving inverse problems in electromagnetic medical imaging. However, contemporary deep-learning-based approaches are susceptible to inaccuracies stemming from inadequate training datasets, primarily consisting of signals generated from simplified and homogeneous imaging scenarios. This paper introduces a novel methodology to construct an expansive and diverse database encompassing domains featuring randomly shaped structures with electrical properties representative of healthy and abnormal tissues. The core objective of this database is to enable the training of universal deep-learning techniques for permittivity profile reconstruction in complex electromagnetic medical imaging domains. The constructed database contains 25,000 unique objects created by superimposing from 6 to 24 randomly sized ellipses and polygons with varying electrical attributes. Introducing randomness in the database enhances training, allowing the neural network to achieve universality while reducing the risk of overfitting. The representative signals in the database are generated using an array of antennas that irradiate the imaging domain and capture scattered signals. A custom-designed U-net is trained by using those signals to generate the permittivity profile of the defined imaging domain. To assess the database and confirm the universality of the trained network, three distinct testing datasets with diverse objects are imaged using the designed U-net. Quantitative assessments of the generated images show promising results, with structural similarity scores consistently exceeding 0.84, normalized root mean square errors remaining below 14%, and peak signal-to-noise ratios exceeding 33 dB. These results demonstrate the practicality of the constructed database for training deep learning networks that have generalization capabilities in solving inverse problems in medical imaging without the need for additional physical assistant algorithms.

DeepSTI: Towards tensor reconstruction using fewer orientations in susceptibility tensor imaging

Susceptibility tensor imaging (STI) is an emerging magnetic resonance imaging technique that characterizes the anisotropic tissue magnetic susceptibility with a second-order tensor model. STI has the potential to provide information for both the reconstruction of white matter fiber pathways and detection of myelin changes in the brain at mm resolution or less, which would be of great value for understanding brain structure and function in healthy and diseased brain. However, the application of STI in vivo has been hindered by its cumbersome and time-consuming acquisition requirement of measuring susceptibility induced MR phase changes at multiple head orientations. Usually, sampling at more than six orientations is required to obtain sufficient information for the ill-posed STI dipole inversion. This complexity is enhanced by the limitation in head rotation angles due to physical constraints of the head coil. As a result, STI has not yet been widely applied in human studies in vivo. In this work, we tackle these issues by proposing an image reconstruction algorithm for STI that leverages data-driven priors. Our method, called DeepSTI, learns the data prior implicitly via a deep neural network that approximates the proximal operator of a regularizer function for STI. The dipole inversion problem is then solved iteratively using the learned proximal network. Experimental results using both simulation and in vivo human data demonstrate great improvement over state-of-the-art algorithms in terms of the reconstructed tensor image, principal eigenvector maps and tractography results, while allowing for tensor reconstruction with MR phase measured at much less than six different orientations. Notably, promising reconstruction results are achieved by our method from only one orientation in human in vivo, and we demonstrate a potential application of this technique for estimating lesion susceptibility anisotropy in patients with multiple sclerosis.

Quantitative susceptibility mapping using multi-channel convolutional neural networks with dipole-adaptive multi-frequency inputs

Quantitative susceptibility mapping (QSM) quantifies the distribution of magnetic susceptibility and shows great potential in assessing tissue contents such as iron, myelin, and calcium in numerous brain diseases. The accuracy of QSM reconstruction was challenged by an ill-posed field-to-susceptibility inversion problem, which is related to the impaired information near the zero-frequency response of the dipole kernel. Recently, deep learning methods demonstrated great capability in improving the accuracy and efficiency of QSM reconstruction. However, the construction of neural networks in most deep learning-based QSM methods did not take the intrinsic nature of the dipole kernel into account. In this study, we propose a dipole kernel-adaptive multi-channel convolutional neural network (DIAM-CNN) method for the dipole inversion problem in QSM. DIAM-CNN first divided the original tissue field into high-fidelity and low-fidelity components by thresholding the dipole kernel in the frequency domain, and it then inputs the two components as additional channels into a multichannel 3D Unet. QSM maps from the calculation of susceptibility through multiple orientation sampling (COSMOS) were used as training labels and evaluation reference. DIAM-CNN was compared with two conventional model-based methods [morphology enabled dipole inversion (MEDI) and improved sparse linear equation and least squares (iLSQR) and one deep learning method (QSMnet)]. High-frequency error norm (HFEN), peak signal-to-noise-ratio (PSNR), normalized root mean squared error (NRMSE), and the structural similarity index (SSIM) were reported for quantitative comparisons. Experiments on healthy volunteers demonstrated that the DIAM-CNN results had superior image quality to those of the MEDI, iLSQR, or QSMnet results. Experiments on data with simulated hemorrhagic lesions demonstrated that DIAM-CNN produced fewer shadow artifacts around the bleeding lesion than the compared methods. This study demonstrates that the incorporation of dipole-related knowledge into the network construction has a potential to improve deep learning-based QSM reconstruction.

msQSM: Morphology-based Self-supervised Deep Learning for Quantitative Susceptibility Mapping

Quantitative susceptibility mapping (QSM) has been applied to the measurement of iron deposition and the auxiliary diagnosis of neurodegenerative disease. There still exists a dipole inversion problem in QSM reconstruction. Recently, deep learning approaches have been proposed to resolve this problem. However, most of these approaches are supervised methods that need pairs of the input phase and ground-truth. It remains a challenge to train a model for all resolutions without using the ground-truth and only using one resolution data. To address this, we proposed a self-supervised QSM deep learning method based on morphology. It consists of a morphological QSM builder to decouple the dependency of the QSM on acquisition resolution, and a morphological loss to reduce artifacts effectively and save training time efficiently. The proposed method can reconstruct arbitrary resolution QSM on both human data and animal data, regardless of whether the resolution is higher or lower than that of the training set. Our method outperforms the previous best unsupervised method with a 3.6% higher peak signal-to-noise ratio, 16.2% lower normalized root mean square error, and 22.1% lower high-frequency error norm. The morphological loss reduces training time by 22.1% with respect to the cycle gradient loss used in the previous unsupervised methods. Experimental results show that the proposed method accurately measures QSM with arbitrary resolutions, and achieves state-of-the-art results among unsupervised deep learning methods. Research on applications in neurodegenerative diseases found that our method is robust enough to measure significant increase in striatal magnetic susceptibility in patients during Alzheimer's disease progression, as well as significant increase in substantia nigra susceptibility in Parkinson's disease patients, and can be used as an auxiliary differential diagnosis tool for Alzheimer's disease and Parkinson's disease.

Deep learning-regularized, single-step quantitative susceptibility mapping quantification

The purpose of the current study was to develop deep learning-regularized, single-step quantitative susceptibility mapping (QSM) quantification, directly generating QSM from the total phase map. A deep learning-regularized, single-step QSM quantification model, named SS-POCSnet, was trained with datasets created using the QSM synthesis approach in QSM reconstruction challenge 2.0. In SS-POCSnet, a data fidelity term based on a single-step model was iteratively applied that combined the spherical mean value kernel and dipole model. Meanwhile, SS-POCSnet regularized susceptibility maps, avoiding underestimating susceptibility values. We evaluated the SS-POCSnet on 10 synthetic datasets, 24 clinical datasets with lesions of cerebral microbleed (CMB) and calcification, and 10 datasets with multiple sclerosis (MS).On synthetic datasets, SS-POCSnet showed the best performance among the methods evaluated, with a normalized root mean squared error of 37.3% ± 4.2%, susceptibility-tuned structured similarity index measure of 0.823 ± 0.02, high-frequency error norm of 37.0 ± 5.7, and peak signal-to-noise ratio of 42.8 ± 1.1. SS-POCSnet also reduced the underestimations of susceptibility values in deep brain nuclei compared with those from the other models evaluated. Furthermore, SS-POCSnet was sensitive to CMB/calcification and MS lesions, demonstrating its clinical applicability. Our method also supported variable imaging parameters, including matrix size and resolution. It was concluded that deep learning-regularized, single-step QSM quantification can mitigate underestimating susceptibility values in deep brain nuclei.

Affine transformation edited and refined deep neural network for quantitative susceptibility mapping

Deep neural networks have demonstrated great potential in solving dipole inversion for Quantitative Susceptibility Mapping (QSM). However, the performances of most existing deep learning methods drastically degrade with mismatched sequence parameters such as acquisition orientation and spatial resolution. We propose an end-to-end AFfine Transformation Edited and Refined (AFTER) deep neural network for QSM, which is robust against arbitrary acquisition orientation and spatial resolution up to 0.6 mm isotropic at the finest. The AFTER-QSM neural network starts with a forward affine transformation layer, followed by a Unet for dipole inversion, then an inverse affine transformation layer, followed by a Residual Dense Network (RDN) for QSM refinement. Simulation and in-vivo experiments demonstrated that the proposed AFTER-QSM network architecture had excellent generalizability. It can successfully reconstruct susceptibility maps from highly oblique and anisotropic scans, leading to the best image quality assessments in simulation tests and suppressed streaking artifacts and noise levels for in-vivo experiments compared with other methods. Furthermore, ablation studies showed that the RDN refinement network significantly reduced image blurring and susceptibility underestimation due to affine transformations. In addition, the AFTER-QSM network substantially shortened the reconstruction time from minutes using conventional methods to only a few seconds.

NeXtQSM-A complete deep learning pipeline for data-consistent Quantitative Susceptibility Mapping trained with hybrid data

Deep learning based Quantitative Susceptibility Mapping (QSM) has shown great potential in recent years, obtaining similar results to established non-learning approaches. Many current deep learning approaches are not data consistent, require in vivo training data or solve the QSM problem in consecutive steps resulting in the propagation of errors. Here we aim to overcome these limitations and developed a framework to solve the QSM processing steps jointly. We developed a new hybrid training data generation method that enables the end-to-end training for solving background field correction and dipole inversion in a data-consistent fashion using a variational network that combines the QSM model term and a learned regularizer. We demonstrate that NeXtQSM overcomes the limitations of previous deep learning methods. NeXtQSM offers a new deep learning based pipeline for computing quantitative susceptibility maps that integrates each processing step into the training and provides results that are robust and fast.

Accelerated 4D-flow MRI with 3-point encoding enabled by machine learning

PURPOSE

To investigate the acceleration of 4D-flow MRI using a convolutional neural network (CNN) that produces three directional velocities from three flow encodings, without requiring a fourth reference scan measuring background phase.

METHODS

A fully 3D CNN using a U-net architecture was trained in a block-wise fashion to take complex images from three flow encodings and to produce three real-valued images for each velocity component. Using neurovascular 4D-flow scans (n = 144), the CNN was trained to predict velocities computed from four flow encodings by standard reconstruction including correction for residual background phase offsets. Methods to optimize loss functions were investigated, including magnitude, complex difference, and uniform velocity weightings. Subsequently, 3-point encoding was evaluated using cross validation of pixelwise correlation, flow measurements in major arteries, and in experiments with data at differing acceleration rates than the training data.

RESULTS

The CNN-produced 3-point velocities showed excellent agreements with the 4-point velocities, both qualitatively in velocity images, in flow rate measures, and quantitatively in regression analysis (slope = 0.96, R²  = 0.992). Optimizing the training to focus on vessel velocities rather than the global velocity error and improved the correlation of velocity within vessels themselves. The lowest error was observed when the loss function used uniform velocity weighting, in which the magnitude-weighted inverse of the velocity frequency uniformly distributed weighting across all velocity ranges. When applied to highly accelerated data, the 3-point network maintained a high correlation with ground truth data and demonstrated a denoising effect.

CONCLUSION

The 4D-flow MRI can be accelerated using machine learning requiring only three flow encodings to produce three-directional velocity maps with small errors.

Deep Learning Staging of Liver Iron Content From Multiecho MR Images

BACKGROUND

MRI represents the most established liver iron content (LIC) evaluation approach by estimation of liver T2* value, but it is dependent on the choice of the measurement region and the software used for image analysis.

PURPOSE

To develop a deep-learning method for unsupervised classification of LIC from magnitude T2* multiecho MR images.

STUDY TYPE

Retrospective.

POPULATION/SUBJECTS

A total of 1069 thalassemia major patients enrolled in the core laboratory of the Myocardial Iron Overload in Thalassemia (MIOT) network, which were included in the training (80%) and test (20%) sets. Twenty patients from different MRI vendors included in the external test set.

FIELD STRENGTH/SEQUENCE

A5 T, T2* multiecho magnitude images.

ASSESSMENT

Four deep-learning convolutional neural networks (HippoNet-2D, HippoNet-3D, HippoNet-LSTM, and an ensemble network HippoNet-Ensemble) were used to achieve unsupervised staging of LIC using five classes (normal, borderline, middle, moderate, severe). The training set was employed to construct the deep-learning model. The performance of the LIC staging model was evaluated in the test set and in the external test set. The model's performances were assessed by evaluating the accuracy, sensitivity, and specificity with respect to the ground truth labels obtained by T2* measurements and by comparison with operator-induced variability originating from different region of interest (ROI) placements.

STATISTICAL TESTS

The network's performances were evaluated by single-class accuracy, specificity, and sensitivity and compared by one-way repeated measures analysis of variance (ANOVA) and one-way ANOVA.

RESULTS

HippoNet-Ensemble reached an accuracy significantly higher than the other networks, and a sensitivity and specificity higher than HippoNet-LSTM. Accuracy, sensitivity, and specificity values for the LIC stages were: normal: 0.96/0.93/0.97, borderline: 0.95/0.85/0.98, mild: 0.96/0.88/0.98, moderate: 0.95/0.89/0.97, severe: 0.97/0.95/0.98. Correctly staging of cases was in the range of 85%-95%, depending on the LIC class. Multiclass accuracy was 0.90 against 0.92 for the interobserver variability.

DATA CONCLUSION

The proposed HippoNet-Ensemble network can perform unsupervised LIC staging and achieves good prognostic performance.

EVIDENCE LEVEL

4 TECHNICAL EFFICACY: Stage 2.

Comparison of quantitative susceptibility mapping methods for iron-sensitive susceptibility imaging at 7T: An evaluation in healthy subjects and patients with Huntington's disease

Quantitative susceptibility mapping (QSM) is a promising tool for investigating iron dysregulation in neurodegenerative diseases, including Huntington's disease (HD). Many diverse methods have been proposed to generate accurate and robust QSM images. In this study, we evaluated the performance of different dipole inversion algorithms for iron-sensitive susceptibility imaging at 7T on healthy subjects of a large age range and patients with HD. We compared an iterative least-squares-based method (iLSQR), iterative methods that use regularization, single-step approaches, and deep learning-based techniques. Their performance was evaluated by comparing: (1) deviations from a multiple-orientation QSM reference; (2) visual appearance of QSM maps and the presence of artifacts; (3) susceptibility in subcortical brain regions with age; (4) regional brain susceptibility with published postmortem brain iron quantification; and (5) susceptibility in HD-affected basal ganglia regions between HD subjects and healthy controls. We found that single-step QSM methods with either total variation or total generalized variation constraints (SSTV/SSTGV) and the single-step deep learning method iQSM generally provided the best performance in terms of correlation with iron deposition and were better at differentiating between healthy controls and premanifest HD individuals, while deep learning QSM methods trained with multiple-orientation susceptibility data created QSM maps that were most similar to the multiple orientation reference and with the best visual scores.

Incomplete spectrum QSM using support information

Introduction

Reconstructing a bounded object from incomplete k-space data is a well posed problem, and it was recently shown that this incomplete spectrum approach can be used to reconstruct undersampled MRI images with similar quality to compressed sensing approaches. Here, we apply this incomplete spectrum approach to the field-to-source inverse problem encountered in quantitative magnetic susceptibility mapping (QSM). The field-to-source problem is an ill-posed problem because of conical regions in frequency space where the dipole kernel is zero or very small, which leads to the kernel's inverse being ill-defined. These "ill-posed" regions typically lead to streaking artifacts in QSM reconstructions. In contrast to compressed sensing, our approach relies on knowledge of the image-space support, more commonly referred to as the mask, of our object as well as the region in k-space with ill-defined values. In the QSM case, this mask is usually available, as it is required for most QSM background field removal and reconstruction methods.

Methods

We tuned the incomplete spectrum method (mask and band-limit) for QSM on a simulated dataset from the most recent QSM challenge and validated the QSM reconstruction results on brain images acquired in five healthy volunteers, comparing incomplete spectrum QSM to current state-of-the art-methods: FANSI, nonlinear dipole inversion, and conventional thresholded k-space division.

Results

Without additional regularization, incomplete spectrum QSM performs slightly better than direct QSM reconstruction methods such as thresholded k-space division (PSNR of 39.9 vs. 39.4 of TKD on a simulated dataset) and provides susceptibility values in key iron-rich regions similar or slightly lower than state-of-the-art algorithms, but did not improve the PSNR in comparison to FANSI or nonlinear dipole inversion. With added (ℓ1-wavelet based) regularization the new approach produces results similar to compressed sensing based reconstructions (at sufficiently high levels of regularization).

Discussion

Incomplete spectrum QSM provides a new approach to handle the "ill-posed" regions in the frequency-space data input to QSM.

Quantitative susceptibility mapping using plug-and-play alternating direction method of multipliers

Quantitative susceptibility mapping employs regularization to reduce artifacts, yet many recent denoisers are unavailable for reconstruction. We developed a plug-and-play approach to QSM reconstruction (PnP QSM) and show its flexibility using several patch-based denoisers. We developed PnP QSM using alternating direction method of multiplier framework and applied collaborative filtering denoisers. We apply the technique to the 2016 QSM Challenge and in 10 glioblastoma multiforme datasets. We compared its performance with four published QSM techniques and a multi-orientation QSM method. We analyzed magnetic susceptibility accuracy using brain region-of-interest measurements, and image quality using global error metrics. Reconstructions on glioblastoma data were analyzed using ranked and semiquantitative image grading by three neuroradiologist observers to assess image quality (IQ) and sharpness (IS). PnP-BM4D QSM showed good correlation (β = 0.84, R² = 0.98, p < 0.05) with COSMOS and no significant bias (bias = 0.007 ± 0.012). PnP-BM4D QSM achieved excellent quality when assessed using structural similarity index metric (SSIM = 0.860), high frequency error norm (HFEN = 58.5), cross correlation (CC = 0.804), and mutual information (MI = 0.475) and also maintained good conspicuity of fine features. In glioblastoma datasets, PnP-BM4D QSM showed higher performance (IQ_Grade = 2.4 ± 0.4, IS_Grade = 2.7 ± 0.3, IQ_Rank = 3.7 ± 0.3, IS_Rank = 3.9 ± 0.3) compared to MEDI (IQ_Grade = 2.1 ± 0.5, IS_Grade = 2.1 ± 0.6, IQ_Rank = 2.4 ± 0.6, IS_Rank = 2.9 ± 0.2) and FANSI-TGV (IQ_Grade = 2.2 ± 0.6, IS_Grade = 2.1 ± 0.6, IQ_Rank = 2.7 ± 0.3, IS_Rank = 2.2 ± 0.2). We illustrated the modularity of PnP QSM by interchanging two additional patch-based denoisers. PnP QSM reconstruction was feasible, and its flexibility was shown using several patch-based denoisers. This technique may allow rapid prototyping and validation of new denoisers for QSM reconstruction for an array of useful clinical applications.

Towards in vivo ground truth susceptibility for single-orientation deep learning QSM: a multi-orientation gradient-echo MRI dataset

Recently, deep neural networks have shown great potential for solving dipole inversion of quantitative susceptibility mapping (QSM) with improved results. However, these studies utilized their limited dataset for network training and inference, which may lead to untrustworthy conclusions. Thus, a common dataset is needed for a fair comparison between different QSM reconstruction networks. Additionally, finding an in vivo reference susceptibility map that matches acquired single-orientation phase data remains an open problem. Susceptibility tensor imaging (STI) χ₃₃ and Calculation of Susceptibility through Multiple Orientation Sampling (COSMOS) are considered reference susceptibility candidates. However, a large number of multi-orientation GRE data for both STI and COSMOS reconstruction are now unavailable for training supervised neural networks for QSM. In this study, we reported the largest multi-orientation dataset, to the best of our knowledge in the QSM research field, with a total of 144 scans from 8 healthy subjects collected using a 3D GRE sequence from the same MR scanner. In addition, the parcellation of deep gray matter is also provided for automatically extracting susceptibility values. Five recently developed deep neural networks, i.e., xQSM, QSMnet, autoQSM, LPCNN, and MoDL-QSM were performed on this dataset. This potential data source could provide a common framework and labels to test the accuracy and robustness of deep neural networks for QSM reconstruction. This dataset has the potential to provide a benchmark of reference susceptibility for the deep learning-based QSM methods. Additionally, the trained COSMOS-labeled and χ₃₃-labeled networks were tested on the pathological data to explore their potential applications. The data together with deep gray matter parcellation maps are now publicly available via an open repository at https://osf.io/yfms7/, and the raw multi-orientation GRE data were also available at https://osf.io/y6rc3/.

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 seconds using the trained iQFM and iQSM neural networks.

Deep learning-based quantitative susceptibility mapping (QSM) in the presence of fat using synthetically generated multi-echo phase training data

PURPOSE

To enable a fast and automatic deep learning-based QSM reconstruction of tissues with diverse chemical shifts, relevant to most regions outside the brain.

METHODS

A UNET was trained to reconstruct susceptibility maps using synthetically generated, unwrapped, multi-echo phase data as input. The RMS error with respect to synthetic validation data was computed. The method was tested on two in vivo knee and two pelvis data sets. Comparisons were made to a conventional fat-water separation pipeline by applying a commonly used graph-cut algorithm, both without and with an extended mask for background field removal (FWS-CONV-QSM and FWS-MASK-CONV-QSM, respectively). Several regions of interest were segmented and compared. Furthermore, the approach was tested on a prostate cancer patient receiving low-dose-rate brachytherapy, to detect and localize the seeds by MRI.

RESULTS

The RMS error was 0.292 ppm with FWS-CONV-QSM and 0.123 ppm for the UNET approach. Susceptibility maps were reconstructed much faster (< 10 s) and completely automatically (no background masking needed) by the UNET compared with the other applied techniques (5 min 51 s and 22 min 44 s for CONV-QSM and FWS-MASK-CONV-QSM, respectively. Background artifacts, fat-water swaps, and hypointense artifacts between I-125 seeds of a patient receiving low-dose brachytherapy in the prostate were largely reduced in the UNET approach.

CONCLUSIONS

Deep learning-based QSM reconstruction, trained solely with synthetic data, is well-suited to rapidly reconstructing high-quality susceptibility maps in the presence of fat without needing masking for background field removal.

Overview of quantitative susceptibility mapping using deep learning: Current status, challenges and opportunities

Quantitative susceptibility mapping (QSM) has gained broad interest in the field by extracting bulk tissue magnetic susceptibility, predominantly determined by myelin, iron and calcium from magnetic resonance imaging (MRI) phase measurements in vivo. Thereby, QSM can reveal pathological changes of these key components in a variety of diseases. QSM requires multiple processing steps such as phase unwrapping, background field removal and field-to-source inversion. Current state-of-the-art techniques utilize iterative optimization procedures to solve the inversion and background field correction, which are computationally expensive and require a careful choice of regularization parameters. With the recent success of deep learning using convolutional neural networks for solving ill-posed reconstruction problems, the QSM community also adapted these techniques and demonstrated that the QSM processing steps can be solved by efficient feed forward multiplications not requiring either iterative optimization or the choice of regularization parameters. Here, we review the current status of deep learning-based approaches for processing QSM, highlighting limitations and potential pitfalls, and discuss the future directions the field may take to exploit the latest advances in deep learning for QSM.

Rapid MR relaxometry using deep learning: An overview of current techniques and emerging trends

Quantitative mapping of MR tissue parameters such as the spin-lattice relaxation time (T₁ ), the spin-spin relaxation time (T₂ ), and the spin-lattice relaxation in the rotating frame (T_1ρ ), referred to as MR relaxometry in general, has demonstrated improved assessment in a wide range of clinical applications. Compared with conventional contrast-weighted (eg T₁ -, T₂ -, or T_1ρ -weighted) MRI, MR relaxometry provides increased sensitivity to pathologies and delivers important information that can be more specific to tissue composition and microenvironment. The rise of deep learning in the past several years has been revolutionizing many aspects of MRI research, including image reconstruction, image analysis, and disease diagnosis and prognosis. Although deep learning has also shown great potential for MR relaxometry and quantitative MRI in general, this research direction has been much less explored to date. The goal of this paper is to discuss the applications of deep learning for rapid MR relaxometry and to review emerging deep-learning-based techniques that can be applied to improve MR relaxometry in terms of imaging speed, image quality, and quantification robustness. The paper is comprised of an introduction and four more sections. Section 2 describes a summary of the imaging models of quantitative MR relaxometry. In Section 3, we review existing "classical" methods for accelerating MR relaxometry, including state-of-the-art spatiotemporal acceleration techniques, model-based reconstruction methods, and efficient parameter generation approaches. Section 4 then presents how deep learning can be used to improve MR relaxometry and how it is linked to conventional techniques. The final section concludes the review by discussing the promise and existing challenges of deep learning for rapid MR relaxometry and potential solutions to address these challenges.

S2Q-Net: Mining the High-Pass Filtered Phase Data in Susceptibility Weighted Imaging for Quantitative Susceptibility Mapping

Susceptibility weighted imaging (SWI) is a routine magnetic resonance imaging (MRI) sequence that combines the magnitude and high-pass filtered phase images to qualitatively enhance the image contrasts related to tissue susceptibility. Tremendous amounts of the high-pass filtered phase data with low signal to noise ratio and incomplete background field removal have thus been collected under default clinical settings. Since SWI cannot quantitatively estimate the susceptibility, it is thus non-trivial to derive quantitative susceptibility mapping (QSM) directly from these redundant phase data, which effectively promotes the mining of the SWI data collected previously and even provides potentials for synchronous imaging of both SWI and QSM based on single SWI scanning in future. To this end, a novel deep learning based SWI-to-QSM-Net (S2Q-Net) is proposed for QSM reconstruction from SWI high-pass filtered phase data. S2Q-Net firstly estimates the edge maps of QSM to integrate edge prior into features, which benefits the network to reconstruct QSM with realistic and clear tissue boundaries. Furthermore, a novel Second-order Cross Dense Block is proposed in S2Q-Net, which can capture rich inter-region interactions to provide rich non-local phase information related to local tissue susceptibility. Experimental results on both simulated and in-vivo datasets demonstrate its superiority over all the compared QSM reconstruction methods, including conventional methods and the state-of-the-art DL-based algorithms. Our results suggest the potentials of S2Q-Net to reconstruct promising QSM from the high-pass filtered phase obtained in clinical SWI sequences.

Learn Less, Infer More: Learning in the Fourier Domain for Quantitative Susceptibility Mapping

Quantitative susceptibility mapping (QSM) aims to evaluate the distribution of magnetic susceptibility from magnetic resonance phase measurements by solving the ill-conditioned dipole inversion problem. Removing the artifacts and preserving the anisotropy of tissue susceptibility simultaneously is still a challenge in QSM. To deal with this issue, a novel k-QSM network is proposed to resolve dipole inversion issues in QSM reconstruction. The k-QSM network converts the results obtained by truncated k-space division (TKD) into the Fourier domain as inputs. After passing through several convolutional and residual blocks, the ill-posed signals of TKD are corrected by making the network output close to the calculation of susceptibility through multiple orientation sampling (COSMOS)-labeled QSM. To evaluate the superiority of k-QSM, comparisons with several state-of-the-art methods are performed in terms of QSM artifacts removing, anisotropy preserving, generalization ability, and clinical applications. Compared to existing methods, the k-QSM achieves a 22.31% lower normalized root mean square error, 10.30% higher peak signal-to-noise ratio (PSNR), 33.10% lower high-frequency error norm, and 1.06% higher structural similarity. In addition, the orientation-dependent susceptibility variation obtained by k-QSM is significant, verifying that k-QSM has the ability to preserve susceptibility anisotropy. When the trained models are tested on the dataset from different centers, our k-QSM shows a strong generalization ability with the highest PSNR. Moreover, by comparing the susceptibility maps between healthy controls and drug addicts with different methods, we found the proposed k-QSM is more sensitive to the susceptibility abnormality in the patients. The proposed k-QSM method learns less—only to fix the ill-posed signals of TKD, but infers more—both COSMOS-like and anisotropy-preserving QSM results. Its generalization ability and great sensitivity to susceptibility changes can make it a potential method for distinguishing some diseases.

Native-resolution myocardial principal Eulerian strain mapping using convolutional neural networks and Tagged Magnetic Resonance Imaging

BACKGROUND

Assessment of regional myocardial function at native pixel-level resolution can play a crucial role in recognizing the early signs of the decline in regional myocardial function. Extensive data processing in existing techniques limits the effective resolution and accuracy of the generated strain maps. The purpose of this study is to compute myocardial principal strain maps ε_p1 and ε_p2 from tagged MRI (tMRI) at the native image resolution using deep-learning local patch convolutional neural network (CNN) models (DeepStrain).

METHODS

For network training, validation, and testing, realistic tMRI datasets were generated and consisted of 53,606 cine images simulating the heart, the liver, blood pool, and backgrounds, including ranges of shapes, positions, motion patterns, noise, and strain. In addition, 102 in-vivo image datasets from three healthy subjects, and three Pulmonary Arterial Hypertension patients, were acquired and used to assess the network's in-vivo performance. Four convolutional neural networks were trained for mapping input tagging patterns to corresponding ground-truth principal strains using different cost functions. Strain maps using harmonic phase analysis (HARP) were obtained with various spectral filtering settings for comparison. CNN and HARP strain maps were compared at the pixel level versus the ground-truth and versus the least-loss in-vivo maps using Pearson correlation coefficients (R) and the median error and Inter-Quartile Range (IQR) histograms.

RESULTS

CNN-based local patch DeepStrain maps at a phantom resolution of 1.1mm × 1.1 mm and in-vivo resolution of 2.1mm × 1.6 mm were artifact-free with multiple fold improvement with ε_p1 ground-truth median error of 0.009(0.007) vs. 0.32(0.385) using HARP and ε_p2 ground-truth error of 0.016(0.021) vs. 0.181(0.08) using HARP. CNN-based strain maps showed substantially higher agreement with the ground-truth maps with correlation coefficients R > 0.91 for ε_p1 and ε_p2 compared to R < 0.21 and R < 0.82 for HARP-generated maps, respectively.

CONCLUSION

CNN-generated Eulerian strain mapping permits artifact-free visualization of myocardial function at the native image resolution.

Edge prior guided dictionary learning for quantitative susceptibility mapping reconstruction

BACKGROUND

Compared with conventional magnetic resonance imaging methods, the quantitative magnetic susceptibility mapping (QSM) technique can quantitatively measure the magnetic susceptibility distribution of tissues, which has an important clinical application value in the investigations of brain micro-bleeds, Parkinson's, and liver iron deposition, etc. However, the quantitative susceptibility mapping algorithm is an ill-posed inverse problem due to the near-zero value in the dipole kernel, and high-quality QSM reconstruction with effective streaking artifact suppression remains a challenge. In recent years, the performance of sparse representation has been well validated in improving magnetic resonance image (MRI) reconstruction.

METHODS

In this study, by incorporating feature learning into sparse representation, we propose an edge prior guided dictionary learning-based reconstruction method for the dipole inversion in quantitative susceptibility mapping reconstruction. The structure feature dictionary relies on magnitude images for susceptibility maps have similar structures with magnitude images, and this structure feature dictionary and edge prior information are used in the dipole inversion step.

RESULTS

The performance of the proposed algorithm is assessed through in vivo human brain clinical data, leading to high-quality susceptibility maps with improved streaking artifact suppression, structural recovery, and quantitative metrics.

CONCLUSIONS

The proposed edge prior guided dictionary learning method for dipole inversion in QSM achieves improved performance in streaking artifacts suppression, structural recovery and deep gray matter reconstruction.

Generation of quantification maps and weighted images from synthetic magnetic resonance imaging using deep learning network

OBJECTIVE

The generation of quantification maps and weighted images in synthetic MRI techniques is based on complex fitting equations. This process requires longer image generation times. The objective of this study is to evaluate the feasibility of deep learning method for fast reconstruction of synthetic MRI.

APPROACH

A total of 44 healthy subjects were recruited and random divided into a training set (30 subjects) and a testing set (14 subjects). A multiple-dynamic, multiple-echo (MDME) sequence was used to acquire synthetic MRI images. Quantification maps (T1, T2, and proton density (PD) maps) and weighted (T1W, T2W, and T2W FLAIR) images were created with MAGiC software and then used as the ground truth images in the deep learning (DL) model. An improved multichannel U-Net structure network was trained to generate quantification maps and weighted images from raw synthetic MRI imaging data (8 module images). Quantitative evaluation was performed on quantification maps. Quantitative evaluation metrics, as well as qualitative evaluation were used in weighted image evaluation. Nonparametric Wilcoxon signed-rank tests were performed in this study.

MAIN RESULTS

The results of quantitative evaluation show that the error between the generated quantification images and the reference images is small. For weighted images, no significant difference in overall image quality or SNR was identified between DL images and synthetic images. Notably, the DL images achieved improved image contrast with T2W images, and fewer artifacts were present on DL images than synthetic images acquired by T2W FLAIR.

SIGNIFICANCE

The DL algorithm provides a promising method for image generation in synthetic MRI techniques, in which every step of the calculation can be optimized and faster, thereby simplifying the workflow of synthetic MRI techniques.

Rigid motion-resolved B 1 + prediction using deep learning for real-time parallel-transmission pulse design

Purpose

Tailored parallel-transmit (pTx) pulses produce uniform excitation profiles at 7 T, but are sensitive to head motion. A potential solution is real-time pulse redesign. A deep learning framework is proposed to estimate pTx B₁⁺ distributions following within-slice motion, which can then be used for tailored pTx pulse redesign.

Methods

Using simulated data, conditional generative adversarial networks were trained to predict B₁⁺ distributions in the head following a displacement. Predictions were made for two virtual body models that were not included in training. Predicted maps were compared with groundtruth (simulated, following motion) B₁⁺ maps. Tailored pTx pulses were designed using B₁⁺ maps at the original position (simulated, no motion) and evaluated using simulated B₁⁺ maps at displaced position (ground-truth maps) to quantify motion-related excitation error. A second pulse was designed using predicted maps (also evaluated on ground-truth maps) to investigate improvement offered by the proposed method.

Results

Predicted B₁⁺ maps corresponded well with ground-truth maps. Error in predicted maps was lower than motion-related error in 99% and 67% of magnitude and phase evaluations, respectively. Worst-case flip-angle normalized RMS error due to motion (76% of target flip angle) was reduced by 59% when pulses were redesigned using predicted maps.

Conclusion

We propose a framework for predicting B₁⁺ maps online with deep neural networks. Predicted maps can then be used for real-time tailored pulse redesign, helping to overcome head motion–related error in pTx.

Upstream Machine Learning in Radiology

Machine learning (ML) and Artificial intelligence (AI) has the potential to dramatically improve radiology practice at multiple stages of the imaging pipeline. Most of the attention has been garnered by applications focused on improving the end of the pipeline: image interpretation. However, this article reviews how AI/ML can be applied to improve upstream components of the imaging pipeline, including exam modality selection, hardware design, exam protocol selection, data acquisition, image reconstruction, and image processing. A breadth of applications and their potential for impact is shown across multiple imaging modalities, including ultrasound, computed tomography, and MRI.

QQ-NET - using deep learning to solve quantitative susceptibility mapping and quantitative blood oxygen level dependent magnitude (QSM+qBOLD or QQ) based oxygen extraction fraction (OEF) mapping

PURPOSE

To improve accuracy and speed of quantitative susceptibility mapping plus quantitative blood oxygen level-dependent magnitude (QSM+qBOLD or QQ) -based oxygen extraction fraction (OEF) mapping using a deep neural network (QQ-NET).

METHODS

The 3D multi-echo gradient echo images were acquired in 34 ischemic stroke patients and 4 healthy subjects. Arterial spin labeling and diffusion weighted imaging (DWI) were also performed in the patients. NET was developed to solve the QQ model inversion problem based on Unet. QQ-based OEF maps were reconstructed with previously introduced temporal clustering, tissue composition, and total variation (CCTV) and NET. The results were compared in simulation, ischemic stroke patients, and healthy subjects using a two-sample Kolmogorov-Smirnov test.

RESULTS

In the simulation, QQ-NET provided more accurate and precise OEF maps than QQ-CCTV with 150 times faster reconstruction speed. In the subacute stroke patients, OEF from QQ-NET had greater contrast-to-noise ratio (CNR) between DWI-defined lesions and their unaffected contralateral normal tissue than with QQ-CCTV: 1.9 ± 1.3 vs 6.6 ± 10.7 (p = 0.03). In healthy subjects, both QQ-CCTV and QQ-NET provided uniform OEF maps.

CONCLUSION

QQ-NET improves the accuracy of QQ-based OEF with faster reconstruction.

Accelerating Whole-Body Diffusion-weighted MRI with Deep Learning-based Denoising Image Filters

Purpose

To use deep learning to improve the image quality of subsampled images (number of acquisitions = 1 [NOA₁]) to reduce whole-body diffusion-weighted MRI (WBDWI) acquisition times.

Materials and Methods

Both retrospective and prospective patient groups were used to develop a deep learning–based denoising image filter (DNIF) model. For initial model training and validation, 17 patients with metastatic prostate cancer with acquired WBDWI NOA₁ and NOA₉ images (acquisition period, 2015–2017) were retrospectively included. An additional 22 prospective patients with advanced prostate cancer, myeloma, and advanced breast cancer were used for model testing (2019), and the radiologic quality of DNIF-processed NOA₁ (NOA_1-DNIF) images were compared with NOA₁ images and clinical NOA₁₆ images by using a three-point Likert scale (good, average, or poor; statistical significance was calculated by using a Wilcoxon signed ranked test). The model was also retrained and tested in 28 patients with malignant pleural mesothelioma (MPM) who underwent lung MRI (2015–2017) to demonstrate feasibility in other body regions.

Results

The model visually improved the quality of NOA₁ images in all test patients, with the majority of NOA_1-DNIF and NOA₁₆ images being graded as either “average” or “good” across all image-quality criteria. From validation data, the mean apparent diffusion coefficient (ADC) values within NOA_1-DNIF images of bone disease deviated from those within NOA₉ images by an average of 1.9% (range, 1.1%–2.6%). The model was also successfully applied in the context of MPM; the mean ADCs from NOA_1-DNIF images of MPM deviated from those measured by using clinical-standard images (NOA₁₂) by 3.7% (range, 0.2%–10.6%).

Conclusion

Clinical-standard images were generated from subsampled images by using a DNIF.

Keywords: Image Postprocessing, MR-Diffusion-weighted Imaging, Neural Networks, Oncology, Whole-Body Imaging, Supervised Learning, MR-Functional Imaging, Metastases, Prostate, Lung

Supplemental material is available for this article.

Published under a CC BY 4.0 license.

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning

PURPOSE

Supervised machine learning (ML) provides a compelling alternative to traditional model fitting for parameter mapping in quantitative MRI. The aim of this work is to demonstrate and quantify the effect of different training data distributions on the accuracy and precision of parameter estimates when supervised ML is used for fitting.

METHODS

We fit a two- and three-compartment biophysical model to diffusion measurements from in-vivo human brain, as well as simulated diffusion data, using both traditional model fitting and supervised ML. For supervised ML, we train several artificial neural networks, as well as random forest regressors, on different distributions of ground truth parameters. We compare the accuracy and precision of parameter estimates obtained from the different estimation approaches using synthetic test data.

RESULTS

When the distribution of parameter combinations in the training set matches those observed in healthy human data sets, we observe high precision, but inaccurate estimates for atypical parameter combinations. In contrast, when training data is sampled uniformly from the entire plausible parameter space, estimates tend to be more accurate for atypical parameter combinations but may have lower precision for typical parameter combinations.

CONCLUSION

This work highlights that estimation of model parameters using supervised ML depends strongly on the training-set distribution. We show that high precision obtained using ML may mask strong bias, and visual assessment of the parameter maps is not sufficient for evaluating the quality of the estimates.

Emerging deep learning techniques using magnetic resonance imaging data applied in multiple sclerosis and clinical isolated syndrome patients (Review)

Computer-aided diagnosis systems aim to assist clinicians in the early identification of abnormal signs in order to optimize the interpretation of medical images and increase diagnostic precision. Multiple sclerosis (MS) and clinically isolated syndrome (CIS) are chronic inflammatory, demyelinating diseases affecting the central nervous system. Recent advances in deep learning (DL) techniques have led to novel computational paradigms in MS and CIS imaging designed for automatic segmentation and detection of areas of interest and automatic classification of anatomic structures, as well as optimization of neuroimaging protocols. To this end, there are several publications presenting artificial intelligence-based predictive models aiming to increase diagnostic accuracy and to facilitate optimal clinical management in patients diagnosed with MS and/or CIS. The current study presents a thorough review covering DL techniques that have been applied in MS and CIS during recent years, shedding light on their current advances and limitations.

QSM reconstruction challenge 2.0: Design and report of results

PURPOSE

The aim of the second quantitative susceptibility mapping (QSM) reconstruction challenge (Oct 2019, Seoul, Korea) was to test the accuracy of QSM dipole inversion algorithms in simulated brain data.

METHODS

A two-stage design was chosen for this challenge. The participants were provided with datasets of multi-echo gradient echo images synthesized from two realistic in silico head phantoms using an MR simulator. At the first stage, participants optimized QSM reconstructions without ground truth data available to mimic the clinical setting. At the second stage, ground truth data were provided for parameter optimization. Submissions were evaluated using eight numerical metrics and visual ratings.

RESULTS

A total of 98 reconstructions were submitted for stage 1 and 47 submissions for stage 2. Iterative methods had the best quantitative metric scores, followed by deep learning and direct inversion methods. Priors derived from magnitude data improved the metric scores. Algorithms based on iterative approaches and total variation (and its derivatives) produced the best overall results. The reported results and analysis pipelines have been made public to allow researchers to compare new methods to the current state of the art.

CONCLUSION

The synthetic data provide a consistent framework to test the accuracy and robustness of QSM algorithms in the presence of noise, calcifications and minor voxel dephasing effects. Total Variation-based algorithms produced the best results among all metrics. Future QSM challenges should assess whether this good performance with synthetic datasets translates to more realistic scenarios, where background fields and dipole-incompatible phase contributions are included.

Visualization of iron-rich subcortical structures in non-human primates in vivo by quantitative susceptibility mapping at 3T MRI

Magnetic resonance imaging (MRI) is now an essential tool in the field of neuroscience involving non-human primates (NHP). Structural MRI scanning using T1-weighted (T1w) or T2-weighted (T2w) images provides anatomical information, particularly for experiments involving deep structures such as the basal ganglia and cerebellum. However, for certain subcortical structures, T1w and T2w image contrasts are insufficient for their detection of important anatomical details. To better visualize such structures in the macaque brain, we applied a relatively new method called quantitative susceptibility mapping (QSM), which enhances tissue contrast based on the local tissue magnetic susceptibility. The QSM significantly improved the visualization of important structures, including the ventral pallidum (VP), globus pallidus external and internal segments (GPe and GPi), substantia nigra (SN), subthalamic nucleus (STN) in the basal ganglia and the dentate nucleus (DN) in the cerebellum. We quantified this the contrast enhancement by systematically comparing of contrast-to-noise ratios (CNRs) of QSM images relative to the corresponding T1w and T2w images. In addition, QSM values of some structures were correlated to the age of the macaque subjects. These results identify the QSM method as a straightforward and useful tool for clearly visualizing details of subcortical structures that are invisible with more traditional scanning sequences.

Deep grey matter quantitative susceptibility mapping from small spatial coverages using deep learning

INTRODUCTION

Quantitative Susceptibility Mapping (QSM) is generally acquired with full brain coverage, even though many QSM brain-iron studies focus on the deep grey matter (DGM) region only. Reducing the spatial coverage to the DGM vicinity can substantially shorten the scan time or enhance the spatial resolution without increasing scan time; however, this may lead to significant DGM susceptibility underestimation.

METHOD

A recently proposed deep learning-based QSM method, namely xQSM, is investigated to assess the accuracy of dipole inversion on reduced brain coverages. The xQSM method is compared with two conventional dipole inversion methods using simulated and in vivo experiments from 4 healthy subjects at 3T. Pre-processed magnetic field maps are extended symmetrically from the centre of globus pallidus in the coronal plane to simulate QSM acquisitions of difference spatial coverages, ranging from 100% (∼32mm) to 400% (∼128mm) of the actual DGM physical size.

RESULTS

The proposed xQSM network led to the lowest DGM contrast loss in both simulated and in vivo subjects, with the smallest susceptibility variation range across all spatial coverages. For the digital brain phantom simulation, xQSM improved the DGM susceptibility underestimation more than 20% in small spatial coverages, as compared to conventional methods. For the in vivo acquisition, less than 5% DGM susceptibility error was achieved in 48mm axial slabs using the xQSM network, while a minimum of 112mm coverage was required for conventional methods. It is also shown that the background field removal process performed worse in reduced brain coverages, which further deteriorated the subsequent dipole inversion.

CONCLUSION

The recently proposed deep learning-based xQSM method significantly improves the accuracy of DGM QSM from small spatial coverages as compared with conventional QSM algorithms, which can shorten DGM QSM acquisition time substantially.

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.

MoDL-QSM: Model-based deep learning for quantitative susceptibility mapping

Quantitative susceptibility mapping (QSM) has demonstrated great potential in quantifying tissue susceptibility in various brain diseases. However, the intrinsic ill-posed inverse problem relating the tissue phase to the underlying susceptibility distribution affects the accuracy for quantifying tissue susceptibility. Recently, deep learning has shown promising results to improve accuracy by reducing the streaking artifacts. However, there exists a mismatch between the observed phase and the theoretical forward phase estimated by the susceptibility label. In this study, we proposed a model-based deep learning architecture that followed the STI (susceptibility tensor imaging) physical model, referred to as MoDL-QSM. Specifically, MoDL-QSM accounts for the relationship between STI-derived phase contrast induced by the susceptibility tensor terms (χ₁₃, χ₂₃ and χ₃₃) and the acquired single-orientation phase. The convolutional neural networks are embedded into the physical model to learn a regularization term containing prior information. χ₃₃ and phase induced by χ₁₃ and χ₂₃ terms were used as the labels for network training. Quantitative evaluation metrics were compared with recently developed deep learning QSM methods. The results showed that MoDL-QSM achieved superior performance, demonstrating its potential for future applications.

QSM reconstruction challenge 2.0: A realistic in silico head phantom for MRI data simulation and evaluation of susceptibility mapping procedures

PURPOSE

To create a realistic in silico head phantom for the second QSM reconstruction challenge and for future evaluations of processing algorithms for QSM.

METHODS

We created a digital whole-head tissue property phantom by segmenting and postprocessing high-resolution (0.64 mm isotropic), multiparametric MRI data acquired at 7 T from a healthy volunteer. We simulated the steady-state magnetization at 7 T using a Bloch simulator and mimicked a Cartesian sampling scheme through Fourier-based processing. Computer code for generating the phantom and performing the MR simulation was designed to facilitate flexible modifications of the phantom in the future, such as the inclusion of pathologies as well as the simulation of a wide range of acquisition protocols. Specifically, the following parameters and effects were implemented: TR and TE, voxel size, background fields, and RF phase biases. Diffusion-weighted imaging phantom data are provided, allowing future investigations of tissue-microstructure effects in phase and QSM algorithms.

RESULTS

The brain part of the phantom featured realistic morphology with spatial variations in relaxation and susceptibility values similar to the in vivo setting. We demonstrated some of the phantom's properties, including the possibility of generating phase data with nonlinear evolution over TE due to partial-volume effects or complex distributions of frequency shifts within the voxel.

CONCLUSION

The presented phantom and computer programs are publicly available and may serve as a ground truth in future assessments of the faithfulness of quantitative susceptibility reconstruction algorithms.

Non-Invasive Evaluation of Cerebral Microvasculature Using Pre-Clinical MRI: Principles, Advantages and Limitations

Alterations to the cerebral microcirculation have been recognized to play a crucial role in the development of neurodegenerative disorders. However, the exact role of the microvascular alterations in the pathophysiological mechanisms often remains poorly understood. The early detection of changes in microcirculation and cerebral blood flow (CBF) can be used to get a better understanding of underlying disease mechanisms. This could be an important step towards the development of new treatment approaches. Animal models allow for the study of the disease mechanism at several stages of development, before the onset of clinical symptoms, and the verification with invasive imaging techniques. Specifically, pre-clinical magnetic resonance imaging (MRI) is an important tool for the development and validation of MRI sequences under clinically relevant conditions. This article reviews MRI strategies providing indirect non-invasive measurements of microvascular changes in the rodent brain that can be used for early detection and characterization of neurodegenerative disorders. The perfusion MRI techniques: Dynamic Contrast Enhanced (DCE), Dynamic Susceptibility Contrast Enhanced (DSC) and Arterial Spin Labeling (ASL), will be discussed, followed by less established imaging strategies used to analyze the cerebral microcirculation: Intravoxel Incoherent Motion (IVIM), Vascular Space Occupancy (VASO), Steady-State Susceptibility Contrast (SSC), Vessel size imaging, SAGE-based DSC, Phase Contrast Flow (PC) Quantitative Susceptibility Mapping (QSM) and quantitative Blood-Oxygenation-Level-Dependent (qBOLD). We will emphasize the advantages and limitations of each strategy, in particular on applications for high-field MRI in the rodent's brain.

Improving phase-based conductivity reconstruction by means of deep learning-based denoising of B 1 + phase data for 3T MRI

PURPOSE

To denoise B 1 + phase using a deep learning method for phase-based in vivo electrical conductivity reconstruction in a 3T MR system.

METHODS

For B 1 + phase deep-learning denoising, a convolutional neural network (U-net) was chosen. Training was performed on data sets from 10 healthy volunteers. Input data were the real and imaginary components of single averaged spin-echo data (SNR = 45), which was used to approximate the B 1 + phase. For label data, multiple signal-averaged spin-echo data (SNR = 128) were used. Testing was performed on in silico and in vivo data. Reconstructed conductivity maps were derived using phase-based conductivity reconstructions. Additionally, we investigated the usability of the network to various SNR levels, imaging contrasts, and anatomical sites (ie, T₁ , T₂ , and proton density-weighted brain images and proton density-weighted breast images. In addition, conductivity reconstructions from deep learning-based denoised data were compared with conventional image filters, which were used for data denoising in electrical properties tomography (ie, the Gaussian filtering and the Savitzky-Golay filtering).

RESULTS

The proposed deep learning-based denoising approach showed improvement for B 1 + phase for both in silico and in vivo experiments with reduced quantitative error measures compared with other methods. Subsequently, this resulted in an improvement of reconstructed conductivity maps from the denoised B 1 + phase with deep learning.

CONCLUSION

The results suggest that the proposed approach can be used as an alternative preprocessing method to denoise B 1 + maps for phase-based conductivity reconstruction without relying on image filters or signal averaging.

Brain/MINDS beyond human brain MRI project: A protocol for multi-level harmonization across brain disorders throughout the lifespan

Psychiatric and neurological disorders are afflictions of the brain that can affect individuals throughout their lifespan. Many brain magnetic resonance imaging (MRI) studies have been conducted; however, imaging-based biomarkers are not yet well established for diagnostic and therapeutic use. This article describes an outline of the planned study, the Brain/MINDS Beyond human brain MRI project (BMB-HBM, FY2018 ~ FY2023), which aims to establish clinically-relevant imaging biomarkers with multi-site harmonization by collecting data from healthy traveling subjects (TS) at 13 research sites. Collection of data in psychiatric and neurological disorders across the lifespan is also scheduled at 13 sites, whereas designing measurement procedures, developing and analyzing neuroimaging protocols, and databasing are done at three research sites. A high-quality scanning protocol, Harmonization Protocol (HARP), was established for five high-quality 3 T scanners to obtain multimodal brain images including T1 and T2-weighted, resting-state and task functional and diffusion-weighted MRI. Data are preprocessed and analyzed using approaches developed by the Human Connectome Project. Preliminary results in 30 TS demonstrated cortical thickness, myelin, functional connectivity measures are comparable across 5 scanners, suggesting sensitivity to subject-specific connectome. A total of 75 TS and more than two thousand patients with various psychiatric and neurological disorders are scheduled to participate in the project, allowing a mixed model statistical harmonization. The HARP protocols are publicly available online, and all the imaging, demographic and clinical information, harmonizing database will also be made available by 2024. To the best of our knowledge, this is the first project to implement a prospective, multi-level harmonization protocol with multi-site TS data. It explores intractable brain disorders across the lifespan and may help to identify the disease-specific pathophysiology and imaging biomarkers for clinical practice.