Magnetic Resonance Fingerprinting (MRF) is a new approach to quantitative magnetic resonance imaging that allows simultaneous measurement of multiple tissue properties in a single, time-efficient acquisition. The ability to reproducibly and quantitatively measure tissue properties could enable more objective tissue diagnosis, comparisons of scans acquired at different locations and time points, longitudinal follow-up of individual patients and development of imaging biomarkers. This review provides a general overview of MRF technology, current preclinical and clinical applications and potential future directions. MRF has been initially evaluated in brain, prostate, liver, cardiac, musculoskeletal imaging, and measurement of perfusion and microvascular properties through MR vascular fingerprinting.

Multiparametric quantitative imaging is gaining increasing interest due to its widespread advantages in clinical applications. Magnetic resonance fingerprinting is a recently introduced approach of fast multiparametric quantitative imaging. In this article, magnetic resonance fingerprinting acquisition, dictionary generation, reconstruction, and validation are reviewed.© International Society for Magnetic Resonance in Medicine.

Magnetic resonance fingerprinting (MRF) is a general framework to quantify multiple MR-sensitive tissue properties with a single acquisition. There have been numerous advances in MRF in the years since its inception. In this work we highlight some of the recent technical developments in MRF, focusing on sequence optimization, modifications for reconstruction and pattern matching, new methods for partial volume analysis, and applications of machine and deep learning. Level of Evidence: 2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2020;51:993-1007.© 2019 International Society for Magnetic Resonance in Medicine.

This study explores the possibility of using gradient echo-based sequences other than balanced steady-state free precession (bSSFP) in the magnetic resonance fingerprinting (MRF) framework to quantify the relaxation parameters.An MRF method based on a fast imaging with steady-state precession (FISP) sequence structure is presented. A dictionary containing possible signal evolutions with physiological range of T1 and T2 was created using the extended phase graph formalism according to the acquisition parameters. The proposed method was evaluated in a phantom and a human brain. T1, T2, and proton density were quantified directly from the undersampled data by the pattern recognition algorithm.T1 and T2 values from the phantom demonstrate that the results of MRF FISP are in good agreement with the traditional gold-standard methods. T1 and T2 values in brain are within the range of previously reported values.MRF-FISP enables a fast and accurate quantification of the relaxation parameters. It is immune to the banding artifact of bSSFP due to B0 inhomogeneities, which could improve the ability to use MRF for applications beyond brain imaging.© 2014 Wiley Periodicals, Inc.

MR fingerprinting (MRF) can be used for quantitative estimation of physical parameters in MRI. Here, we extend the method to incorporate B1 estimation.The acquisition is based on steady state free precession MR fingerprinting with a Cartesian trajectory. To increase the sensitivity to the B1 profile, abrupt changes in flip angle were introduced in the sequence. Slice profile and B1 effects were included in the dictionary and the results from two- and three-dimensional (3D) acquisitions were compared. Acceleration was demonstrated using retrospective undersampling in the phase encode directions of 3D data exploiting redundancy between MRF frames at the edges of k-space.Without B1 estimation, T2 and B1 were inaccurate by more than 20%. Abrupt changes in flip angle improved B1 maps. T1 and T2 values obtained with the new MRF methods agree with classical spin echo measurements and are independent of the B1 field profile. When using view sharing reconstruction, results remained accurate (error <10%) when sampling under 10% of k-space from the 3D data.The methods demonstrated here can successfully measure T1, T2, and B1. Errors due to slice profile can be substantially reduced by including its effect in the dictionary or acquiring data in 3D. Magn Reson Med 76:1127-1135, 2016. © 2015 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.© 2015 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine.

To introduce a quantitative tool that enables rapid forecasting of T and T parameter map errors due to normal and aliasing noise as a function of the MR fingerprinting (MRF) sequence, which can be used in sequence optimization.The variances of normal noise and aliasing artifacts in the collected signal are related to the variances in T and T maps through derived quality factors. This analytical result is tested against the results of a Monte-Carlo approach for analyzing MRF sequence encoding capability in the presence of aliasing noise, and verified with phantom experiments at 3 T. To further show the utility of our approach, our quality factors are used to find efficient MRF sequences for fewer repetitions.Experimental results verify the ability of our quality factors to rapidly assess the efficiency of an MRF sequence in the presence of both normal and aliasing noise. Quality factor assessment of MRF sequences is in agreement with the results of a Monte-Carlo approach. Analysis of MRF parameter map errors from phantom experiments is consistent with the derived quality factors, with T (T ) data yielding goodness of fit R ≥ 0.92 (0.80). In phantom and in vivo experiments, the efficient pulse sequence, determined through quality factor maximization, led to comparable or improved accuracy and precision relative to a longer sequence, demonstrating quality factor utility in MRF sequence design.The here introduced quality factor framework allows for rapid analysis and optimization of MRF sequence design through T and T error forecasting.© 2019 International Society for Magnetic Resonance in Medicine.

Sequence optimization and appropriate sequence selection is still an unmet need in magnetic resonance fingerprinting (MRF). The main challenge in MRF sequence design is the lack of an appropriate measure of the sequence's encoding capability. To find such a measure, three different candidates for judging the encoding capability have been investigated: local and global dot-product-based measures judging dictionary entry similarity as well as a Monte Carlo method that evaluates the noise propagation properties of an MRF sequence. Consistency of these measures for different sequence lengths as well as the capability to predict actual sequence performance in both phantom and in vivo measurements was analyzed. While the dot-product-based measures yielded inconsistent results for different sequence lengths, the Monte Carlo method was in a good agreement with phantom experiments. In particular, the Monte Carlo method could accurately predict the performance of different flip angle patterns in actual measurements. The proposed Monte Carlo method provides an appropriate measure of MRF sequence encoding capability and may be used for sequence optimization.Copyright © 2017 Elsevier Inc. All rights reserved.

Magnetic resonance fingerprinting (MRF) is a fast MRI-based technique that allows for multiparametric quantitative characterization of the tissues of interest in a single acquisition. In particular, it has gained attention in the field of cardiac imaging due to its ability to provide simultaneous and co-registered myocardial T1 and T2 mapping in a single breath-held cardiac MRF scan, in addition to other parameters. Initial results in small healthy subject groups and clinical studies have demonstrated the feasibility and potential of MRF imaging. Ongoing research is being conducted to improve the accuracy, efficiency, and robustness of cardiac MRF. However, these improvements usually increase the complexity of image reconstruction and dictionary generation and introduce the need for sequence optimization. Each of these steps increase the computational demand and processing time of MRF. The latest advances in artificial intelligence (AI), including progress in deep learning and the development of neural networks for MRI, now present an opportunity to efficiently address these issues. Artificial intelligence can be used to optimize candidate sequences and reduce the memory demand and computational time required for reconstruction and post-processing. Recently, proposed machine learning-based approaches have been shown to reduce dictionary generation and reconstruction times by several orders of magnitude. Such applications of AI should help to remove these bottlenecks and speed up cardiac MRF, improving its practical utility and allowing for its potential inclusion in clinical routine. This review aims to summarize the latest developments in artificial intelligence applied to cardiac MRF. Particularly, we focus on the application of machine learning at different steps of the MRF process, such as sequence optimization, dictionary generation and image reconstruction.

This work proposes new low rank approximation approaches with significant memory savings for large scale MR fingerprinting (MRF) problems.We introduce a compressed MRF with randomized singular value decomposition method to significantly reduce the memory requirement for calculating a low rank approximation of large sized MRF dictionaries. We further relax this requirement by exploiting the structures of MRF dictionaries in the randomized singular value decomposition space and fitting them to low-degree polynomials to generate high resolution MRF parameter maps. In vivo 1.5T and 3T brain scan data are used to validate the approaches.T, T, and off-resonance maps are in good agreement with that of the standard MRF approach. Moreover, the memory savings is up to 1000 times for the MRF-fast imaging with steady-state precession sequence and more than 15 times for the MRF-balanced, steady-state free precession sequence.The proposed compressed MRF with randomized singular value decomposition and dictionary fitting methods are memory efficient low rank approximation methods, which can benefit the usage of MRF in clinical settings. They also have great potentials in large scale MRF problems, such as problems considering multi-component MRF parameters or high resolution in the parameter space. Magn Reson Med 79:2392-2400, 2018. © 2017 International Society for Magnetic Resonance in Medicine.© 2017 International Society for Magnetic Resonance in Medicine.

MR fingerprinting (MRF) is a technique for quantitative tissue mapping using pseudorandom measurements. To estimate tissue properties such as T1, T2, proton density, and B0, the rapidly acquired data are compared against a large dictionary of Bloch simulations. This matching process can be a very computationally demanding portion of MRF reconstruction.We introduce a fast group matching algorithm (GRM) that exploits inherent correlation within MRF dictionaries to create highly clustered groupings of the elements. During matching, a group specific signature is first used to remove poor matching possibilities. Group principal component analysis (PCA) is used to evaluate all remaining tissue types. In vivo 3 Tesla brain data were used to validate the accuracy of our approach.For a trueFISP sequence with over 196,000 dictionary elements, 1000 MRF samples, and image matrix of 128 × 128, GRM was able to map MR parameters within 2s using standard vendor computational resources. This is an order of magnitude faster than global PCA and nearly two orders of magnitude faster than direct matching, with comparable accuracy (1-2% relative error).The proposed GRM method is a highly efficient model reduction technique for MRF matching and should enable clinically relevant reconstruction accuracy and time on standard vendor computational resources.© 2014 Wiley Periodicals, Inc.

Existing approaches for reconstruction of multiparametric maps with magnetic resonance fingerprinting (MRF) are currently limited by their estimation accuracy and reconstruction time. We aimed to address these issues with a novel combination of iterative reconstruction, fingerprint compression, additional regularization, and accelerated dictionary search methods. The pipeline described here, accelerated iterative reconstruction for magnetic resonance fingerprinting (AIR-MRF), was evaluated with simulations as well as phantom and in vivo scans. We found that the AIR-MRF pipeline provided reduced parameter estimation errors compared to non-iterative and other iterative methods, particularly at shorter sequence lengths. Accelerated dictionary search methods incorporated into the iterative pipeline reduced the reconstruction time at little cost of quality.Copyright © 2017 Elsevier Inc. All rights reserved.

Magnetic resonance fingerprinting (MRF) is a new technique for simultaneously quantifying multiple MR parameters using one temporally resolved MR scan. In MRF, MR signal is manipulated to have distinct temporal behavior with regard to different combinations of the underlying MR parameters and across spatial regions. The temporal behavior of acquired MR signal is then used as a key to find its unique counterpart in a MR signal dictionary. The dictionary generation and searching (DGS) process represents the most important part of MRF, which however can be intractable because of the disk space requirement and the computational demand exponentially increases with the number of MR parameters, spatial coverage, and spatial resolution. The goal of this paper was to develop a fast and space efficient MRF DGS algorithm.The optimal DGS algorithm: MRF ZOOM was designed based on the properties of the parameter matching objective function characterized with full dictionary simulations. Both synthetic data and in-vivo data were used to validate the method.MRF ZOOM can dramatically save MRF DGS time without sacrificing matching accuracy.MRF ZOOM can facilitate a wide range of MRF applications.

The purpose of this work is to evaluate methods from deep learning for application to Magnetic Resonance Fingerprinting (MRF). MRF is a recently proposed measurement technique for generating quantitative parameter maps. In MRF a non-steady state signal is generated by a pseudo-random excitation pattern. A comparison of the measured signal in each voxel with the physical model yields quantitative parameter maps. Currently, the comparison is done by matching a dictionary of simulated signals to the acquired signals. To accelerate the computation of quantitative maps we train a Convolutional Neural Network (CNN) on simulated dictionary data. As a proof of principle we show that the neural network implicitly encodes the dictionary and can replace the matching process.

A deep neural network was developed for magnetic resonance fingerprinting (MRF) quantification. This study aimed at extending previous studies of deep learning MRF to in vivo applications, allowing sub-second computation time for large-scale data.We applied the deep learning methodology based on our previously published multi-layer perceptron. The number of layers was four, which was optimized to balance the model capacity and noise robustness. The training sets were obtained from MRF dictionaries with 9000 to 28,000 atoms, depending on the desired T1 and T2 ranges. The simulated MRF undersampling artifact based on the k-space acquisition scheme and noise were both added to the training data to reduce the error in estimates.The neural network achieved high fidelity (R2 _ 0.98) as compared to the T1 and T2 values of the ISMRM standardized phantom. In brain MRF experiment, the model trained with simulated artifacts and noise showed less error compared to that without. The in vivo application of our neural network for liver and prostate were also demonstrated. For an MRF slice with 256 _ 256 image resolution, the computation time of our neural network was 0.12 s, compared with the _ 28 s-pre-slice for the conventional dictionary matching method.Our neural network achieved fast computation speed for MRF quantification. The model trained with simulated artifacts and noise showed less error and achieved optimal performance for phantom experiment and in vivo normal brain and liver, and prostate cancer patient.Copyright © 2020 Elsevier Inc. All rights reserved.

Conventional MRI can be limited in detecting subtle epileptic lesions or identifying active/epileptic lesions among widespread, multifocal lesions.We developed a high-resolution 3D MR fingerprinting (MRF) protocol to simultaneously provide quantitative T, T, proton density, and tissue fraction maps for detection and characterization of epileptic lesions.Prospective.National Institute of Standards and Technology (NIST) / International Society for Magnetic Resonance in Medicine (ISMRM) phantom, five healthy volunteers and 15 patients with medically intractable epilepsy undergoing presurgical evaluation with noninvasive or invasive electroclinical data.3D MRF scans and routine clinical epilepsy MR protocols were acquired at 3 T.The accuracy of the T and T values were first evaluated using the NIST/ISMRM phantom. The repeatability was then estimated with both phantom and volunteers based on the coefficient of variance (CV). For epilepsy patients, all the maps were qualitatively reviewed for lesion detection by three independent reviewers (S.E.J., M.L., I.N.) blinded to clinical data. Region of interest (ROI) analysis was performed on T and T maps to quantify the multiparametric signal differences between lesion and normal tissues. Findings from qualitative review and quantitative ROI analysis were compared with patients' electroclinical data to assess concordance.Phantom results were compared using R-squared, and patient results were compared using linear regression models.The phantom study showed high accuracy with the standard values, with an R of 0.99. The volunteer study showed high repeatability, with an average CV of 4.3% for T and T in various tissue regions. For the 15 patients, MRF showed additional findings in four patients, with the remaining 11 patients showing findings consistent with conventional MRI. The additional MRF findings were highly concordant with patients' electroclinical presentation.The 3D MRF protocol showed potential to identify otherwise inconspicuous epileptogenic lesions from the patients with negative conventional MRI diagnosis, as well as to correlate with different levels of epileptogenicity when widespread lesions were present.3. Technical Efficacy Stage: 3. J. Magn. Reson. Imaging 2019;49:1333-1346.© 2018 International Society for Magnetic Resonance in Medicine.

To develop a magnetic resonance (MR) "fingerprinting" technique for quantitative abdominal imaging.This HIPAA-compliant study had institutional review board approval, and informed consent was obtained from all subjects. To achieve accurate quantification in the presence of marked B0 and B1 field inhomogeneities, the MR fingerprinting framework was extended by using a two-dimensional fast imaging with steady-state free precession, or FISP, acquisition and a Bloch-Siegert B1 mapping method. The accuracy of the proposed technique was validated by using agarose phantoms. Quantitative measurements were performed in eight asymptomatic subjects and in six patients with 20 focal liver lesions. A two-tailed Student t test was used to compare the T1 and T2 results in metastatic adenocarcinoma with those in surrounding liver parenchyma and healthy subjects.Phantom experiments showed good agreement with standard methods in T1 and T2 after B1 correction. In vivo studies demonstrated that quantitative T1, T2, and B1 maps can be acquired within a breath hold of approximately 19 seconds. T1 and T2 measurements were compatible with those in the literature. Representative values included the following: liver, 745 msec ± 65 (standard deviation) and 31 msec ± 6; renal medulla, 1702 msec ± 205 and 60 msec ± 21; renal cortex, 1314 msec ± 77 and 47 msec ± 10; spleen, 1232 msec ± 92 and 60 msec ± 19; skeletal muscle, 1100 msec ± 59 and 44 msec ± 9; and fat, 253 msec ± 42 and 77 msec ± 16, respectively. T1 and T2 in metastatic adenocarcinoma were 1673 msec ± 331 and 43 msec ± 13, respectively, significantly different from surrounding liver parenchyma relaxation times of 840 msec ± 113 and 28 msec ± 3 (P <.0001 and P <.01) and those in hepatic parenchyma in healthy volunteers (745 msec ± 65 and 31 msec ± 6, P <.0001 and P =.021, respectively).A rapid technique for quantitative abdominal imaging was developed that allows simultaneous quantification of multiple tissue properties within one 19-second breath hold, with measurements comparable to those in published literature.

Purpose To develop a fast three-dimensional method for simultaneous T1 and T2 quantification for breast imaging by using MR fingerprinting. Materials and Methods In this prospective study, variable flip angles and magnetization preparation modules were applied to acquire MR fingerprinting data for each partition of a three-dimensional data set. A fast postprocessing method was implemented by using singular value decomposition. The proposed technique was first validated in phantoms and then applied to 15 healthy female participants (mean age, 24.2 years ± 5.1 [standard deviation]; range, 18-35 years) and 14 female participants with breast cancer (mean age, 55.4 years ± 8.8; range, 39-66 years) between March 2016 and April 2018. The sensitivity of the method to B field inhomogeneity was also evaluated by using the Bloch-Siegert method. Results Phantom results showed that accurate and volumetric T1 and T2 quantification was achieved by using the proposed technique. The acquisition time for three-dimensional quantitative maps with a spatial resolution of 1.6 × 1.6 × 3 mm was approximately 6 minutes. For healthy participants, averaged T1 and T2 relaxation times for fibroglandular tissues at 3.0 T were 1256 msec ± 171 and 46 msec ± 7, respectively. Compared with normal breast tissues, higher T2 relaxation time (68 msec ± 13) was observed in invasive ductal carcinoma (P <.001), whereas no statistical difference was found in T1 relaxation time (1183 msec ± 256; P =.37). Conclusion A method was developed for breast imaging by using the MR fingerprinting technique, which allows simultaneous and volumetric quantification of T1 and T2 relaxation times for breast tissues. © RSNA, 2018 Online supplemental material is available for this article.

To improve image quality and accelerate the acquisition of 3D MR fingerprinting (MRF).Building on the multi-axis spiral-projection MRF technique, a subspace reconstruction with locally low-rank constraint and a modified spiral-projection spatiotemporal encoding scheme called tiny golden-angle shuffling were implemented for rapid whole-brain high-resolution quantitative mapping. Reconstruction parameters such as the locally low-rank regularization parameter and the subspace rank were tuned using retrospective in vivo data and simulated examinations. B inhomogeneity correction using multifrequency interpolation was incorporated into the subspace reconstruction to further improve the image quality by mitigating blurring caused by off-resonance effect.The proposed MRF acquisition and reconstruction framework yields high-quality 1-mm isotropic whole-brain quantitative maps in 2 min at better quality compared with 6-min acquisitions of prior approaches. The proposed method was validated to not induce bias in T and T mapping. High-quality whole-brain MRF data were also obtained at 0.66-mm isotropic resolution in 4 min using the proposed technique, where the increased resolution was shown to improve visualization of subtle brain structures.The proposed tiny golden-angle shuffling, MRF with optimized spiral-projection trajectory and subspace reconstruction enables high-resolution quantitative mapping in ultrafast acquisition time.© 2022 International Society for Magnetic Resonance in Medicine.

Purpose To improve diagnosis of hippocampal sclerosis (HS) in patients with mesial temporal lobe epilepsy (MTLE) by using MR fingerprinting and compare with visual assessment of T1- and T2-weighted MR images. Materials and Methods For this prospective study performed between April and November 2016, T1 and T2 maps were obtained and tissue segmentation performed in consecutive patients with drug-resistant MTLE with unilateral or bilateral HS. T1 and T2 maps were compared between 33 patients with MTLE (23 women and 10 men; mean age, 32.6 years; age range, 16-60 years) and 30 healthy participants (20 women and 10 men; mean age, 28.8 years; age range, 18-40 years). Differences in individual bilateral hippocampi were compared by using a Wilcoxon signed rank test, whereas the Wilcoxon rank-sum test was used for difference analysis between healthy control participants and patients with MTLE. Results The diagnosis rate (ie, ratio of HS diagnosed on the basis of a 2.5-minute MR fingerprinting examination compared with standard methods: MRI, electroencephalography, and PET) was 32 of 33 (96.9%; 95% confidence interval: 84.9%, 100%), reflecting improved accuracy of diagnosis (P = 1.92 × 10) over routine MR examinations that had a diagnostic rate of 23 of 33 (69.7%; 95% confidence interval: 51.5%, 81.6%). The comparison between atrophic and normal-appearing hippocampus in 33 patients with MTLE and healthy control participants demonstrated that both T1 and T2 values in HS lesions were higher than those of normal hippocampal tissue of healthy participants (T1: 1361 msec ± 85 vs 1249 msec ± 59, respectively; T2: 135 msec ± 15 vs 104 msec ± 9, respectively; P <.0001). Conclusion MR fingerprinting allowed for multiparametric mapping of temporal lobe within 2.5 minutes and helped to identify lesions suspicious for HS in patients with MTLE with improved accuracy.© RSNA, 2018 Online supplemental material is available for this article.

Magnetic resonance imaging of the pancreas is increasingly used as an important diagnostic modality for characterisation of pancreatic lesions. Pancreatic MRI protocols are mostly qualitative due to time constraints and motion sensitivity. MR Fingerprinting is an innovative acquisition technique that provides qualitative data and quantitative parameter maps from a single free‐breathing acquisition with the potential to reduce exam times. This work investigates the feasibility of MRF parameter mapping for pancreatic imaging in the presence of free-breathing exam. Sixteen healthy participants were prospectively imaged using MRF framework. Regions-of-interest were drawn in multiple solid organs including the pancreas and T1 and T2 values determined. MRF T1 and T2 mapping was performed successfully in all participants (acquisition time:2.4–3.6 min). Mean pancreatic T1 values were 37–43% lower than those of the muscle, spleen, and kidney at both 1.5 and 3.0 T. For these organs, the mean pancreatic T2 values were nearly 40% at 1.5 T and &lt; 12% at 3.0 T. The feasibility of MRF at 1.5 T and 3 T was demonstrated in the pancreas. By enabling fast and free-breathing quantitation, MRF has the potential to add value during the clinical characterisation and grading of pathological conditions, such as pancreatitis or cancer.

Cardiovascular magnetic resonance is a versatile tool that enables noninvasive characterization of cardiac tissue structure and function. Parametric mapping techniques have allowed unparalleled differentiation of pathophysiological differences in the myocardium such as the delineation of myocardial fibrosis, hemorrhage, and edema. These methods are increasingly used as part of a tool kit to characterize disease states such as cardiomyopathies and coronary artery disease more accurately. Currently conventional mapping techniques require separate acquisitions for T and T mapping, the values of which may depend on specifics of the magnetic resonance imaging system hardware, pulse sequence implementation, and physiological variables including blood pressure and heart rate. The cardiac magnetic resonance fingerprinting (cMRF) technique has recently been introduced for simultaneous and reproducible measurement of T and T maps in a single scan. The potential for this technique to provide consistent tissue property values independent of variables including scanner, pulse sequence, and physiology could allow an unbiased framework for the assessment of intrinsic properties of cardiac tissue including structure, perfusion, and parameters such as extracellular volume without the administration of exogenous contrast agents. This review seeks to introduce the basics of the cMRF technique, including pulse sequence design, dictionary generation, and pattern matching. The potential applications of cMRF in assessing diseases such as nonischemic cardiomyopathy are also briefly discussed, and ongoing areas of research are described.Copyright © 2018 American College of Cardiology Foundation. Published by Elsevier Inc. All rights reserved.

Magnetic resonance fingerprinting (MRF) is a fast MRI-based technique that allows for multiparametric quantitative characterization of the tissues of interest in a single acquisition. In particular, it has gained attention in the field of cardiac imaging due to its ability to provide simultaneous and co-registered myocardial T1 and T2 mapping in a single breath-held cardiac MRF scan, in addition to other parameters. Initial results in small healthy subject groups and clinical studies have demonstrated the feasibility and potential of MRF imaging. Ongoing research is being conducted to improve the accuracy, efficiency, and robustness of cardiac MRF. However, these improvements usually increase the complexity of image reconstruction and dictionary generation and introduce the need for sequence optimization. Each of these steps increase the computational demand and processing time of MRF. The latest advances in artificial intelligence (AI), including progress in deep learning and the development of neural networks for MRI, now present an opportunity to efficiently address these issues. Artificial intelligence can be used to optimize candidate sequences and reduce the memory demand and computational time required for reconstruction and post-processing. Recently, proposed machine learning-based approaches have been shown to reduce dictionary generation and reconstruction times by several orders of magnitude. Such applications of AI should help to remove these bottlenecks and speed up cardiac MRF, improving its practical utility and allowing for its potential inclusion in clinical routine. This review aims to summarize the latest developments in artificial intelligence applied to cardiac MRF. Particularly, we focus on the application of machine learning at different steps of the MRF process, such as sequence optimization, dictionary generation and image reconstruction.

Background Only sparse literature investigates the reproducibility and repeatability of relaxometry methods in MRI. However, statistical data on reproducibility and repeatability of any quantitative method is essential for clinical application. Purpose To evaluate the reproducibility and repeatability of two-dimensional fast imaging with steady-state free precession MR fingerprinting in vivo in human brains. Materials and Methods Two-dimensional section-selective MR fingerprinting based on a steady-state free precession sequence with an external radiofrequency transmit field, or, correction was used to generate T1 and T2 maps. This prospective study was conducted between July 2017 and January 2018 with 10 scanners from a single manufacturer, including different models, at four different sites. T1 and T2 relaxation times and their variation across scanners (reproducibility) as well as across repetitions on a scanner (repeatability) were analyzed. The relative deviations of T1 and T2 to the average (95% confidence interval) were calculated for several brain compartments. Results Ten healthy volunteers (mean age ± standard deviation, 28.5 years ± 6.9; eight men, two women) participated in this study. Reproducibility and repeatability of T1 and T2 measures in the human brain varied across brain compartments (1.8%-20.9%) and were higher in solid tissues than in the cerebrospinal fluid. T1 measures in solid tissue brain compartments were more stable compared with T2 measures. The half-widths of the confidence intervals for relative deviations were 3.4% for mean T1 and 8.0% for mean T2 values across scanners. Intrascanner repeatability half-widths of the confidence intervals for relative deviations were in the range of 2.0%-3.1% for T1 and 3.1%-7.9% for T2. Conclusion This study provides values on reproducibility and repeatability of T1 and T2 relaxometry measured with fast imaging with steady-state free precession MR fingerprinting in brain tissues of healthy volunteers. Reproducibility and repeatability are considerably higher in solid brain compartments than in cerebrospinal fluid and are higher for T1 than for T2. © RSNA, 2019 See also the editorial by Barkhof and Parker in this issue.

To compare the absolute values and repeatability of magnetic resonance fingerprinting (MRF) with 3000 and 1500 echoes/slice acquired in 41 s and 20 s (MRF3k and MRF1.5k, respectively).