Deep learning–based MR fingerprinting ASL ReconStruction (DeepMARS)

Recommendations for quantitative cerebral perfusion MRI using multi-timepoint arterial spin labeling: Acquisition, quantification, and clinical applications

Accurate assessment of cerebral perfusion is vital for understanding the hemodynamic processes involved in various neurological disorders and guiding clinical decision-making. This guidelines article provides a comprehensive overview of quantitative perfusion imaging of the brain using multi-timepoint arterial spin labeling (ASL), along with recommendations for its acquisition and quantification. A major benefit of acquiring ASL data with multiple label durations and/or post-labeling delays (PLDs) is being able to account for the effect of variable arterial transit time (ATT) on quantitative perfusion values and additionally visualize the spatial pattern of ATT itself, providing valuable clinical insights. Although multi-timepoint data can be acquired in the same scan time as single-PLD data with comparable perfusion measurement precision, its acquisition and postprocessing presents challenges beyond single-PLD ASL, impeding widespread adoption. Building upon the 2015 ASL consensus article, this work highlights the protocol distinctions specific to multi-timepoint ASL and provides robust recommendations for acquiring high-quality data. Additionally, we propose an extended quantification model based on the 2015 consensus model and discuss relevant postprocessing options to enhance the analysis of multi-timepoint ASL data. Furthermore, we review the potential clinical applications where multi-timepoint ASL is expected to offer significant benefits. This article is part of a series published by the International Society for Magnetic Resonance in Medicine (ISMRM) Perfusion Study Group, aiming to guide and inspire the advancement and utilization of ASL beyond the scope of the 2015 consensus article.

Recurrent neural network-based simultaneous cardiac T1, T2, and T1ρ mapping

The purpose of the current study was to explore the feasibility of training a deep neural network to accelerate the process of generating T1, T2, and T1ρ maps for a recently proposed free-breathing cardiac multiparametric mapping technique, where a recurrent neural network (RNN) was utilized to exploit the temporal correlation among the multicontrast images. The RNN-based model was developed for rapid and accurate T1, T2, and T1ρ estimation. Bloch simulation was performed to simulate a dataset of more than 10 million signals and time correspondences with different noise levels for network training. The proposed RNN-based method was compared with a dictionary-matching method and a conventional mapping method to evaluate the model's effectiveness in phantom and in vivo studies at 3 T, respectively. In phantom studies, the RNN-based method and the dictionary-matching method achieved similar accuracy and precision in T1, T2, and T1ρ estimations. In in vivo studies, the estimated T1, T2, and T1ρ values obtained by the two methods achieved similar accuracy and precision for 10 healthy volunteers (T1: 1228.70 ± 53.80 vs. 1228.34 ± 52.91 ms, p > 0.1; T2: 40.70 ± 2.89 vs. 41.19 ± 2.91 ms, p > 0.1; T1ρ: 45.09 ± 4.47 vs. 45.23 ± 4.65 ms, p > 0.1). The RNN-based method can generate cardiac multiparameter quantitative maps simultaneously in just 2 s, achieving 60-fold acceleration compared with the dictionary-matching method. The RNN-accelerated method offers an almost instantaneous approach for reconstructing accurate T1, T2, and T1ρ maps, being much more efficient than the dictionary-matching method for the free-breathing multiparametric cardiac mapping technique, which may pave the way for inline mapping in clinical applications.

Simultaneous perfusion, diffusion, T2 *, and T1 mapping with MR fingerprinting

PURPOSE

Quantitative mapping of brain perfusion, diffusion, T2 *, and T1 has important applications in cerebrovascular diseases. At present, these sequences are performed separately. This study aims to develop a novel MRI technique to simultaneously estimate these parameters.

METHODS

This sequence to measure perfusion, diffusion, T2 *, and T1 mapping with magnetic resonance fingerprinting (MRF) was based on a previously reported MRF-arterial spin labeling (ASL) sequence, but the acquisition module was modified to include different TEs and presence/absence of bipolar diffusion-weighting gradients. We compared parameters derived from the proposed method to those derived from reference methods (i.e., separate sequences of MRF-ASL, conventional spin-echo DWI, and T2 * mapping). Test-retest repeatability and initial clinical application in two patients with stroke were evaluated.

RESULTS

The scan time of our proposed method was 24% shorter than the sum of the reference methods. Parametric maps obtained from the proposed method revealed excellent image quality. Their quantitative values were strongly correlated with those from reference methods and were generally in agreement with values reported in the literature. Repeatability assessment revealed that ADC, T2 *, T1 , and B1 + estimation was highly reliable, with voxelwise coefficient of variation (CoV) <5%. The CoV for arterial transit time and cerebral blood flow was 16% ± 3% and 25% ± 9%, respectively. The results from the two patients with stroke demonstrated that parametric maps derived from the proposed method can detect both ischemic and hemorrhagic stroke.

CONCLUSION

The proposed method is a promising technique for multi-parametric mapping and has potential use in patients with stroke.

Scanner-Independent MyoMapNet for Accelerated Cardiac MRI T1 Mapping Across Vendors and Field Strengths

BACKGROUND

In cardiac T1 mapping, a series of T1 -weighted (T1 w) images are collected and numerically fitted to a two or three-parameter model of the signal recovery to estimate voxel-wise T1 values. To reduce the scan time, one can collect fewer T1 w images, albeit at the cost of precision or/and accuracy. Recently, the feasibility of using a neural network instead of conventional two- or three-parameter fit modeling has been demonstrated. However, prior studies used data from a single vendor and field strength; therefore, the generalizability of the models has not been established.

PURPOSE

To develop and evaluate an accelerated cardiac T1 mapping approach based on MyoMapNet, a convolution neural network T1 estimator that can be used across different vendors and field strengths by incorporating the relevant scanner information as additional inputs to the model.

STUDY TYPE

Retrospective, multicenter.

POPULATION

A total of 1423 patients with known or suspected cardiac disease (808 male, 57 ± 16 years), from three centers, two vendors (Siemens, Philips), and two field strengths (1.5 T, 3 T). The data were randomly split into 60% training, 20% validation, and 20% testing.

FIELD STRENGTH/SEQUENCE

A 1.5 T and 3 T, Modified Look-Locker inversion recovery (MOLLI) for native and postcontrast T1 .

ASSESSMENT

Scanner-independent MyoMapNet (SI-MyoMapNet) was developed by altering the deep learning (DL) architecture of MyoMapNet to incorporate scanner vendor and field strength as inputs. Epicardial and endocardial contours and blood pool (by manually drawing a large region of interest in the blood pool) of the left ventricle were manually delineated by three readers, with 2, 8, and 9 years of experience, and SI-MyoMapNet myocardial and blood pool T1 values (calculated from four T1 w images) were compared with conventional MOLLI T1 values (calculated from 8 to 11 T1 w images).

STATISTICAL TESTS

Equivalency test with 95% confidence interval (CI), linear regression slope, Pearson correlation coefficient (r), Bland-Altman analysis.

RESULTS

The proposed SI-MyoMapNet successfully created T1 maps. Native and postcontrast T1 values measured from SI-MyoMapNet were strongly correlated with MOLLI, despite using only four T1 w images, at both field-strengths and vendors (all r > 0.86). For native T1 , SI-MyoMapNet and MOLLI were in good agreement for myocardial and blood T1 values in institution 1 (myocardium: 5 msec, 95% CI [3, 8]; blood: -10 msec, 95%CI [-16, -4]), in institution 2 (myocardium: 6 msec, 95% CI [0, 11]; blood: 0 msec, [-18, 17]), and in institution 3 (myocardium: 7 msec, 95% CI [-8, 22]; blood: 8 msec, [-14, 30]). Similar results were observed for postcontrast T1 .

DATA CONCLUSION

Inclusion of field strength and vendor as additional inputs to the DL architecture allows generalizability of MyoMapNet across different vendors or field strength.

EVIDENCE LEVEL

2.

TECHNICAL EFFICACY

Stage 2.

Deeply Accelerated Arterial Spin Labeling Perfusion MRI for Measuring Cerebral Blood Flow and Arterial Transit Time

Cerebral blood flow (CBF) indicates both vascular integrity and brain function. Regional CBF can be non-invasively measured with arterial spin labeling (ASL) perfusion MRI. By repeating the same ASL MRI sequence several times, each with a different post-labeling delay (PLD), another important neurovascular index, the arterial transit time (ATT) can be estimated by fitting the acquired ASL signal to a kinetic model. This process however faces two challenges: one is the multiplicatively prolonged scan time, making it impractically for clinical use due to the escalated risk of motions; the other is the reduced signal-to-noise-ratio (SNR) in the long PLD scans due to the T1 decay of the labeled spins. Increasing SNR needs more repetitions which will further increase the total scan time. Currently, there lacks a way to accurately estimate ATT from a parsimonious number of PLDs. In this paper, we proposed a deep learning-based algorithm to reduce the number of PLDs and to accurately estimate ATT and CBF. Two separate deep networks were trained: one is designed to estimate CBF and ATT from ASL data with a single PLD; the other is to estimate CBF and ATT from ASL data with two PLDs. The models were trained and tested using the large Human Connectome Project multiple-PLD ASL MRI. Performance of the DL-based approach was compared to the traditional full dataset-based data fitting approach. Our results showed that ATT and CBF can be reliably estimated using deep networks even with one PLD.

DeepFittingNet: A deep neural network-based approach for simplifying cardiac T1 and T2 estimation with improved robustness

PURPOSE

To develop and evaluate a deep neural network (DeepFittingNet) for T1 /T2 estimation of the most commonly used cardiovascular MR mapping sequences to simplify data processing and improve robustness.

THEORY AND METHODS

DeepFittingNet is a 1D neural network composed of a recurrent neural network (RNN) and a fully connected (FCNN) neural network, in which RNN adapts to the different number of input signals from various sequences and FCNN subsequently predicts A, B, and Tx of a three-parameter model. DeepFittingNet was trained using Bloch-equation simulations of MOLLI and saturation-recovery single-shot acquisition (SASHA) T1 mapping sequences, and T2 -prepared balanced SSFP (T2 -prep bSSFP) T2 mapping sequence, with reference values from the curve-fitting method. Several imaging confounders were simulated to improve robustness. The trained DeepFittingNet was tested using phantom and in-vivo signals, and compared to the curve-fitting algorithm.

RESULTS

In testing, DeepFittingNet performed T1 /T2 estimation of four sequences with improved robustness in inversion-recovery T1 estimation. The mean bias in phantom T1 and T2 between the curve-fitting and DeepFittingNet was smaller than 30 and 1 ms, respectively. Excellent agreements between both methods was found in the left ventricle and septum T1 /T2 with a mean bias <6 ms. There was no significant difference in the SD of both the left ventricle and septum T1 /T2 between the two methods.

CONCLUSION

DeepFittingNet trained with simulations of MOLLI, SASHA, and T2 -prep bSSFP performed T1 /T2 estimation tasks for all these most used sequences. Compared with the curve-fitting algorithm, DeepFittingNet improved the robustness for inversion-recovery T1 estimation and had comparable performance in terms of accuracy and precision.

Recent Technical Developments in ASL: A Review of the State of the Art

This review article provides an overview of a range of recent technical developments in advanced arterial spin labeling (ASL) methods that have been developed or adopted by the community since the publication of a previous ASL consensus paper by Alsop et al. It is part of a series of review/recommendation papers from the International Society for Magnetic Resonance in Medicine Perfusion Study Group. Here, we focus on advancements in readouts and trajectories, image reconstruction, noise reduction, partial volume correction, quantification of nonperfusion parameters, fMRI, fingerprinting, vessel selective ASL, angiography, deep learning, and ultrahigh field ASL. We aim to provide a high level of understanding of these new approaches and some guidance for their implementation, with the goal of facilitating the adoption of such advances by research groups and by MRI vendors. Topics outside the scope of this article that are reviewed at length in separate articles include velocity selective ASL, multiple-timepoint ASL, body ASL, and clinical ASL recommendations.

Artificial intelligence in multi-parametric magnetic resonance imaging: A review

Multi-parametric magnetic resonance imaging (mpMRI) is an indispensable tool in the clinical workflow for the diagnosis and treatment planning of various diseases. Machine learning-based artificial intelligence (AI) methods, especially those adopting the deep learning technique, have been extensively employed to perform mpMRI image classification, segmentation, registration, detection, reconstruction, and super-resolution. The current availability of increasing computational power and fast-improving AI algorithms have empowered numerous computer-based systems for applying mpMRI to disease diagnosis, imaging-guided radiotherapy, patient risk and overall survival time prediction, and the development of advanced quantitative imaging technology for magnetic resonance fingerprinting. However, the wide application of these developed systems in the clinic is still limited by a number of factors, including robustness, reliability, and interpretability. This survey aims to provide an overview for new researchers in the field as well as radiologists with the hope that they can understand the general concepts, main application scenarios, and remaining challenges of AI in mpMRI. This article is protected by copyright. All rights reserved.

Contrast-agent-free state-of-the-art MRI on cerebral small vessel disease-part 1. ASL, IVIM, and CVR

Cerebral small vessel disease (cSVD), a common cause of stroke and dementia, is traditionally considered the small vessel equivalent of large artery occlusion or rupture that leads to cortical and subcortical brain damage. Microvessel endothelial dysfunction can also contribute to it. Brain imaging, including MRI, is useful to show the presence of lesions of several types, although the association between conventional MRI measures and clinical features of cSVD is not always concordant. We assessed the additional contribution of contrast-agent-free, state-of-the-art MRI techniques such as arterial spin labeling (ASL), diffusion tensor imaging, functional MRI, and intravoxel incoherent motion (IVIM) applied to cSVD in the existing literature. We performed a review following the PICO Worksheet and Search Strategy, including original papers in English, published between 2000 and 2022. For each MRI method, we extracted information about their contributions, in addition to those established with traditional MRI methods and related information about the origins, pathology, markers, and clinical outcomes in cSVD. This paper presents the first part of the review, which includes 37 studies focusing on ASL, IVIM, and cerebrovascular reactivity (CVR) measures. In general, they have shown that, in addition to white matter hyperintensities, alterations in other neuroimaging parameters such as blood flow and CVR also indicate the presence of cSVD. Such quantitative parameters were also related to cSVD risk factors. Therefore, they are promising, noninvasive tools to explore questions that have not yet been clarified about this clinical condition. However, protocol standardization is essential to increase their clinical use.

Simultaneous Hemodynamic and Structural Imaging of Ischemic Stroke With Magnetic Resonance Fingerprinting Arterial Spin Labeling

BACKGROUND

Perfusion and structural imaging play an important role in ischemic stroke. Magnetic resonance fingerprinting (MRF) arterial spin labeling (ASL) is a novel noninvasive method of ASL perfusion that allows simultaneous estimation of cerebral blood flow (CBF), bolus arrival time (BAT), and tissue T₁ map in a single scan of <4 minutes. Here, we evaluated the utility of MRF-ASL in patients with ischemic stroke in terms of detecting hemodynamic and structural damage and predicting neurological deficits and disability.

METHODS

A total of 34 patients were scanned on 3T magnetic resonance imaging. MRF-ASL, standard single-delay pseudo-continuous ASL, T₂-weighted, and diffusion magnetic resonance imaging were performed. Regions of interest of lesion and contralateral normal tissues were manually delineated. CBF (with 2 different compartmental models), BAT, and tissue T₁ parameters were quantified. Cross-sectional linear regression analyses were performed to examine the relationship between MRF-ASL parameters and National Institutes of Health Stroke Scale (NIHSS) and modified Rankin Scale. Receiver operating characteristic analyses were performed to determine the utility of MRF-ASL in the classification of stroke lesion voxels.

RESULTS

MRF-ASL derived parameters revealed a significant difference between stroke lesion and contralateral normal regions of interest, in that lesion regions manifested a lower CBF_{1-compartment} (P<0.001), lower CBF_{2-compartment} (P<0.001), longer BAT (P=0.002), and longer T₁ (P<0.001) compared with normal regions of interest. NIHSS scores at acute stage revealed a strong association with lesion-normal differences in CBF_{1-compartment,diff} (β=-0.11, P=0.008), CBF_{2-compartment,diff} (β=-0.16, P=0.003), and T_1,diff (β=0.008, P=0.001). MRF-ASL parameters were also predictive of NIHSS score and modified Rankin Scale scale measured at a later stage, although the degree of the associations was weaker. These associations tended to be even stronger when the MRF-ASL data were acquired at the acute/subacute stage. Compared with standard pseudo-continuous ASL, the multiparametric capability of MRF-ASL yielded higher area under curve values in the receiver operating characteristic analyses of stroke voxel classifications.

CONCLUSIONS

MRF-ASL may provide a new approach for quantitative hemodynamic and structural imaging in ischemic stroke.

Automated generation of cerebral blood flow and arterial transit time maps from multiple delay arterial spin-labeled MRI

PURPOSE

Develop and evaluate a deep learning approach to estimate cerebral blood flow (CBF) and arterial transit time (ATT) from multiple post-labeling delay (PLD) ASL MRI.

METHODS

ASL MRI were acquired with 6 PLDs on a 1.5T or 3T GE system in adults with and without cognitive impairment (N = 99). Voxel-level CBF and ATT maps were quantified by training models with distinct convolutional neural network architectures: (1) convolutional neural network (CNN) and (2) U-Net. Models were trained and compared via 5-fold cross validation. Performance was evaluated using mean absolute error (MAE). Model outputs were trained on and compared against a reference ASL model fitting after data cleaning. Minimally processed ASL data served as another benchmark. Model output uncertainty was estimated using Monte Carlo dropout. The better-performing neural network was subsequently re-trained on inputs with missing PLDs to investigate generalizability to different PLD schedules.

RESULTS

Relative to the CNN, the U-Net yielded lower MAE on training data. On test data, the U-Net MAE was 8.4 ± 1.4 mL/100 g/min for CBF and 0.22 ± 0.09 s for ATT. A significant association was observed between MAE and Monte Carlo dropout-based uncertainty estimates. Neural network performance remained stable despite systematically reducing the number of input images (i.e., up to 3 missing PLD images). Mean processing time was 10.77 s for the U-Net neural network compared to 10 min 41 s for the reference pipeline.

CONCLUSION

It is feasible to generate CBF and ATT maps from 1.5T and 3T multi-PLD ASL MRI with a fast deep learning image-generation approach.

Macrohistory of Moyamoya Disease Analyzed Using Artificial Intelligence

INTRODUCTION

Moyamoya disease is characterized by progressive stenotic changes in the terminal segment of the internal carotid artery and the development of abnormal vascular networks called moyamoya vessels. The objective of this review was to provide a holistic view of the epidemiology, etiology, clinical findings, treatment, and pathogenesis of moyamoya disease. A literature search was performed in PubMed using the term "moyamoya disease," for articles published until 2021.

RESULTS

Artificial intelligence (AI) clustering was used to classify the articles into 5 clusters: (1) pathophysiology (23.5%); (2) clinical background (37.3%); (3) imaging (13.2%); (4) treatment (17.3%); and (5) genetics (8.7%). Many articles in the "clinical background" cluster were published from the 1970s. However, in the "treatment" and "genetics" clusters, the articles were published from the 2010s through 2021. In 2011, it was confirmed that a gene called Ringin protein 213 (RNF213) is a susceptibility gene for moyamoya disease. Since then, tremendous progress in genomic, transcriptomic, and epigenetic profiling (e.g., methylation profiling) has resulted in new concepts for classifying moyamoya disease. Our literature survey revealed that the pathogenesis involves aberrations of multiple signaling pathways through genetic mutations and altered gene expression.

CONCLUSION

We analyzed the content vectors in abstracts using AI, and reviewed the pathophysiology, clinical background, radiological features, treatments, and genetic peculiarity of moyamoya disease.

Non-contrast hemodynamic imaging of Moyamoya disease with MR fingerprinting ASL: A feasibility study

PURPOSE

MR Fingerprinting (MRF) Arterial Spin Labeling (ASL) is a non-contrast technique to estimate multiple brain hemodynamic and structural parameters in a single scan. The purpose of this study is to examine the feasibility and initial utility of MRF-ASL in Moyamoya disease.

METHODS

MRF-ASL, conventional single-delay ASL, Time-of-flight (TOF) MR angiography, and contrast-based dynamic-susceptibility-contrast (DSC) MRI were prospectively collected from a group of Moyamoya patients in North America (N = 21, 4 men and 17 women). Sixteen healthy subjects (7 men and 9 women) also underwent an MRF-ASL scan. Cerebral blood flow (CBF), bolus arrival time (BAT), and tissue T1 were compared between Moyamoya patients and healthy controls. Perfusion parameters from MRF-ASL were compared to those from other MRI sequences. Multi-linear regression was used for comparisons of parameter values between Moyamoya and control groups. Linear mixed-effects models was used when comparing MRF-ASL to PCASL and DSC parameters. Spearman's Rank Correlation Coefficient was calculated when comparing MRF-ASL to and MRA grades. A P value of 0.05 or less was considered significant.

RESULTS

BAT in stenotic internal carotid artery (ICA) territories was prolonged (P < 0.001) in Moyamoya patients, when compared with healthy controls. CBF in stenotic ICA territories of Moyamoya patients was not different from CBF in healthy controls; but in the PCA territories, CBF in Moyamoya patients was higher (P < 0.01) than controls. Quantitative T1 values in the stenotic ICA territories was longer (P < 0.05) than that in controls. Hemodynamic parameters estimated from MRF-ASL were significantly correlated with single-delay ASL and DSC. Longer BAT was associated with more severe intracranial artery stenosis in ICA.

CONCLUSIONS

MRF-ASL is a promising technique to assess perfusion and structural abnormalities in Moyamoya patients.

Data Loss Reconstruction Method for a Bridge Weigh-in-Motion System Using Generative Adversarial Networks

In the application of a bridge weigh-in-motion (WIM) system, the collected data may be temporarily or permanently lost due to sensor failure or system transmission failure. The high data loss rate weakens the distribution characteristics of the collected data and the ability of the monitoring system to conduct assessments on bridge condition. A deep learning-based model, or generative adversarial network (GAN), is proposed to reconstruct the missing data in the bridge WIM systems. The proposed GAN in this study can model the collected dataset and predict the missing data. Firstly, the data from stable measurements before the data loss are provided, and then the generator is trained to extract the retained features from the dataset and the data lost in the process are collected by using only the responses of the remaining functional sensors. The discriminator feeds back the recognition results to the generator in order to improve its reconstruction accuracy. In the model training, two loss functions, generation loss and confrontation loss, are used, and the general outline and potential distribution characteristics of the signal are well processed by the model. Finally, by applying the engineering data of the Hangzhou Jiangdong Bridge to the GAN model, this paper verifies the effectiveness of the proposed method. The results show that the final reconstructed dataset is in good agreement with the actual dataset in terms of total vehicle weight and axle weight. Furthermore, the approximate contour and potential distribution characteristics of the original dataset are reproduced. It is suggested that the proposed method can be used in real-life applications. This research can provide a promising method for the data reconstruction of bridge monitoring systems.

Accelerated cardiac T1 mapping in four heartbeats with inline MyoMapNet: a deep learning-based T1 estimation approach

PURPOSE

To develop and evaluate MyoMapNet, a rapid myocardial T1 mapping approach that uses fully connected neural networks (FCNN) to estimate T1 values from four T1-weighted images collected after a single inversion pulse in four heartbeats (Look-Locker, LL4).

METHOD

We implemented an FCNN for MyoMapNet to estimate T1 values from a reduced number of T1-weighted images and corresponding inversion-recovery times. We studied MyoMapNet performance when trained using native, post-contrast T1, or a combination of both. We also explored the effects of number of T1-weighted images (four and five) for native T1. After rigorous training using in-vivo modified Look-Locker inversion recovery (MOLLI) T1 mapping data of 607 patients, MyoMapNet performance was evaluated using MOLLI T1 data from 61 patients by discarding the additional T1-weighted images. Subsequently, we implemented a prototype MyoMapNet and LL4 on a 3 T scanner. LL4 was used to collect T1 mapping data in 27 subjects with inline T1 map reconstruction by MyoMapNet. The resulting T1 values were compared to MOLLI.

RESULTS

MyoMapNet trained using a combination of native and post-contrast T1-weighted images had excellent native and post-contrast T1 accuracy compared to MOLLI. The FCNN model using four T1-weighted images yields similar performance compared to five T1-weighted images, suggesting that four T1 weighted images may be sufficient. The inline implementation of LL4 and MyoMapNet enables successful acquisition and reconstruction of T1 maps on the scanner. Native and post-contrast myocardium T1 by MOLLI and MyoMapNet was 1170 ± 55 ms vs. 1183 ± 57 ms (P = 0.03), and 645 ± 26 ms vs. 630 ± 30 ms (P = 0.60), and native and post-contrast blood T1 was 1820 ± 29 ms vs. 1854 ± 34 ms (P = 0.14), and 508 ± 9 ms vs. 514 ± 15 ms (P = 0.02), respectively.

CONCLUSION

A FCNN, trained using MOLLI data, can estimate T1 values from only four T1-weighted images. MyoMapNet enables myocardial T1 mapping in four heartbeats with similar accuracy as MOLLI with inline map reconstruction.

Streamlined magnetic resonance fingerprinting: Fast whole-brain coverage with deep-learning based parameter estimation

Magnetic resonance fingerprinting (MRF) is a quantitative MRI (qMRI) framework that provides simultaneous estimates of multiple relaxation parameters as well as metrics of field inhomogeneity in a single acquisition. However, current challenges exist in the forms of (1) scan time; (2) need for custom image reconstruction; (3) large dictionary sizes; (4) long dictionary-matching time. This study aims to introduce a novel streamlined magnetic-resonance fingerprinting (sMRF) framework based on a single-shot echo-planar imaging (EPI) sequence to simultaneously estimate tissue T1, T2, and T2* with integrated B1⁺ correction. Encouraged by recent work on EPI-based MRF, we developed a method that combines spin-echo EPI with gradient-echo EPI to achieve T2 in addition to T1 and T2* quantification. To this design, we add simultaneous multi-slice (SMS) acceleration to enable full-brain coverage in a few minutes. Moreover, in the parameter-estimation step, we use deep learning to train a deep neural network (DNN) to accelerate the estimation process by orders of magnitude. Notably, due to the high image quality of the EPI scans, the training process can rely simply on Bloch-simulated data. The DNN also removes the need for storing large dictionaries. Phantom scans along with in-vivo multi-slice scans from seven healthy volunteers were acquired with resolutions of 1.1×1.1×3 mm³ and 1.7×1.7×3 mm³, and the results were validated against ground truth measurements. Excellent correspondence was found between our T1, T2, and T2* estimates and results obtained from standard approaches. In the phantom scan, a strong linear relationship (R = 1-1.04, R²>0.96) was found for all parameter estimates, with a particularly high agreement for T2 estimation (R²>0.99). Similar findings are reported for the in-vivo human data for all of our parameter estimates. Incorporation of DNN results in a reduction of parameter estimation time on the order of 1000 x and a reduction in storage requirements on the order of 2500 x while achieving highly similar results as conventional dictionary matching (%differences of 7.4 ± 0.4%, 3.6 ± 0.3% and 6.0 ± 0.4% error in T1, T2, and T2* estimation). Thus, sMRF has the potential to be the method of choice for future MRF studies by providing ease of implementation, fast whole-brain coverage, and ultra-fast T1/T2/T2* estimation.

Deep learning for brain disorders: from data processing to disease treatment

In order to reach precision medicine and improve patients' quality of life, machine learning is increasingly used in medicine. Brain disorders are often complex and heterogeneous, and several modalities such as demographic, clinical, imaging, genetics and environmental data have been studied to improve their understanding. Deep learning, a subpart of machine learning, provides complex algorithms that can learn from such various data. It has become state of the art in numerous fields, including computer vision and natural language processing, and is also growingly applied in medicine. In this article, we review the use of deep learning for brain disorders. More specifically, we identify the main applications, the concerned disorders and the types of architectures and data used. Finally, we provide guidelines to bridge the gap between research studies and clinical routine.

Multi-band MR fingerprinting (MRF) ASL imaging using artificial-neural-network trained with high-fidelity experimental data

PURPOSE

We aim to leverage the power of deep-learning with high-fidelity training data to improve the reliability and processing speed of hemodynamic mapping with MR fingerprinting (MRF) arterial spin labeling (ASL).

METHODS

A total of 15 healthy subjects were studied on a 3T MRI. Each subject underwent 10 runs of a multi-band multi-slice MRF-ASL sequence for a total scan time of approximately 40 min. MRF-ASL images were averaged across runs to yield a set of high-fidelity data. Training of a fully connected artificial neural network (ANN) was then performed using these data. The results from ANN were compared to those of dictionary matching (DM), ANN trained with single-run experimental data and with simulation data. Initial clinical performance of the technique was also demonstrated in a Moyamoya patient.

RESULTS

The use of ANN reduced the processing time of MRF-ASL data to 3.6 s, compared to DM of 3 h 12 min. Parametric values obtained with ANN and DM were strongly correlated (R² between 0.84 and 0.96). Results obtained from high-fidelity ANN were substantially more reliable compared to those from DM or single-run ANN. Voxel-wise coefficient of variation (CoV) of high-fidelity ANN, DM, and single-run ANN was 0.15 ± 0.08, 0.41 ± 0.20, 0.30 ± 0.16, respectively, for cerebral blood flow and 0.11 ± 0.06, 0.20 ± 0.19, 0.15 ± 0.10, respectively, for bolus arrival time. In vivo data trained ANN also outperformed ANN trained with simulation data. The superior performance afforded by ANN allowed more conspicuous depiction of hemodynamic abnormalities in Moyamoya patient.

CONCLUSION

Deep-learning-based parametric reconstruction improves the reliability of MRF-ASL hemodynamic maps and reduces processing time.

Numerical approximation to the general kinetic model for ASL quantification

PURPOSE

To develop a numerical approximation to the general kinetic model for arterial spin labeling (ASL) quantification that will enable greater flexibility in ASL acquisition methods.

THEORY

The Bloch-McConnell equations are extended to include the effects of single-compartment inflow and outflow on both the transverse and longitudinal magnetization. These can be solved using an extension of Jaynes' matrix formalism with piecewise constant approximation of incoming labeled arterial flow and a clearance operator for outgoing venous flow.

METHODS

The proposed numerical approximation is compared with the general kinetic model using simulations of pulsed labeling and pseudo-continuous labeling and a broad range of transit time and bolus duration for tissue blood flow of 0.6 mL/g/min. Accuracy of the approximation is studied as a function of the timestep using Monte-Carlo simulations. Three additional scenarios are demonstrated: (1) steady-pulsed ASL, (2) MR fingerprinting ASL, and (3) balanced SSFP and spoiled gradient-echo sequences.

RESULTS

The proposed approximation was found to be arbitrarily accurate for pulsed labeling and pseudo-continuous labeling. The pulsed labeling/pseudo-continuous labeling approximation error compared with the general kinetic model was less than 0.002% (<0.002%) and less than 0.05% (<0.05%) for timesteps of 3 ms and 35 ms, respectively. The proposed approximation matched well with customized signal expressions of steady-pulsed ASL and MR fingerprinting ASL. The simulations of simultaneous modeling of flow, T₂ , and magnetization transfer showed an increase in steady-state balanced SSFP and spoiled gradient signals.

CONCLUSION

We demonstrate a numerical approximation of the "Bloch-McConnell flow" equations that enables arbitrarily accurate modeling of pulsed ASL and pseudo-continuous labeling signals comparable to the general kinetic model. This enables increased flexibility in the experiment design for quantitative ASL.