T1 Estimation Research Articles

A preparation pulse for fast steady state approach in Actual Flip angle Imaging.

Actual Flip angle Imaging (AFI) is a sequence used for B1 mapping, also embedded in the Variable flip angle with AFI for simultaneous estimation of T1 , B1 and equilibrium magnetization. To investigate the design of a preparation module for AFI to allow a fast approach to steady state (SS) without requiring the use of dummyacquisitions. The features of a preparation module with a B1 insensitive adiabatic pulse, spoiler gradients, and a recovery time were studied with simulations and validated via experiments and acquired with different k-space traveling strategies. The robustness of the flip angle of the preparation pulse on the acquired signal isstudied. When a 90° adiabatic pulse is used, the forthcoming can be expressed as a function of repetition times and AFI flip angle only as , where n represents the ratio between the two repetition times of AFI. The robustness of the method is demonstrated by showing that using the values further away from 90° still allows for a faster approach to SS than the use of dummypulses. The preparation module is particularly advantageous for low flip angles, as well as for AFI sequences that sample the center of the k-space early in the sequence, such as centric ordering acquisitions, and for ultrafast EPI-based AFI methods, thus allowing to reduce scanner overheadtime.

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

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. 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. 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. 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.

A deep learning approach for accurate and robust T1 estimation from cardiac magnetic resonance MOLLI sequences

Abstract Funding Acknowledgements Type of funding sources: None. Background T1 parametric maps of the myocardium are a recent development in cardiac Magnetic Resonance Imaging (MRI). They can be used to characterize the composition of the heart and identify alterations that can be indicative of a pathology. A common method of T1 estimation involves pixel-wise non-linear curve fitting from Modified Look-Locker Inversion Recovery (MOLLI) series [1]. Whilst relatively accurate, this method presents several drawbacks. Firstly, it assumes magnetization recovery to be a purely exponential process dependent on T1, thus neglecting other parameters such as T2 and B1 that can interfere in the process. Secondly, it does not consider that each acquisition during an inversion recovery experiment inverts the magnetization vector by a given angle and thus interferes with the recovery leading to the measurement of a shorter T1* rather than the actual T1 [2]. Additionally, since it considers each pixel individually, neglecting spatial information, it is sensitive to image quality [3]. Finally, since non-linear curve fitting is an iterative algorithm, it is computationally expensive and time-consuming. Methods To address these issues a convolutional neural network, DeepBLESS [4], and two spatially aware variants of it, were trained to predict T1, T2, and B1 with a dataset of MOLLI series simulated from experimental data. The simulation was performed using Bloch simulations and parameters extracted from a dataset composed of 27 single-slice MOLLI series acquired with the 5(3s)3 protocol using a b-SSFP readout at a main magnetic field strength of 3 Tesla. Thus, a dataset could be generated without relying on exponential fitting for ground truth generation. Results The exponential fitting of T1 was shown to consistently underestimate true T1 values by 138 +- 90 ms. In comparison, the spatially naïve DeepBLESS approach showed a mean absolute error of 14 +- 9 ms and the spatially aware DeepBLESS approach (Median-DeepBLESS) showed a mean absolute error of 13 +- 9 ms. As can be seen in Figure 1, Median-DeepBLESS also showed considerably more resilience to noise than exponential fitting and the spatially naïve version of DeepBLESS. Additionally, the deep learning approaches showcased promising results with regards to the estimation of T2 and B1 from MOLLI sequences (See Figure 2). Conclusions It was shown that the Deep Learning approaches eliminated the underestimation bias of exponential fitting with the spatially aware network Median-DeepBLESS being more robust to noise than both exponential fitting and the standard DeepBLESS network for myocardial T1 estimation. Relatively accurate estimation of T2 and B1 from MOLLI sequences was also accomplished through the Deep Learning approach, which may lead to simultaneous T1, T2, and B1 estimation in common clinical practice in the future.

Motion-corrected model-based reconstruction for 2D myocardial T1 mapping.

To allow for T1 mapping of the myocardium within 2.3 s for a 2D slice utilizing cardiac motion-corrected, model-based image reconstruction. Golden radial data acquisition is continuously carried out for 2.3 s after an inversion pulse. In a first step, dynamic images are reconstructed which show both contrast changes due to T1 recovery and anatomical changes due to the heartbeat. An image registration algorithm with a signal model for T1 recovery is applied to estimate non-rigid cardiac motion. In a second step, estimated motion fields are applied during an iterative model-based T1 reconstruction. The approach was evaluated in numerical simulations, phantom experiments and in in-vivo scans in healthy volunteers. The accuracy of cardiac motion estimation was shown in numerical simulations with an average motion field error of 0.7 ± 0.6 mm for a motion amplitude of 5.1 mm. The accuracy of T1 estimation was demonstrated in phantom experiments, with no significant difference (p = 0.13) in T1 estimated by the proposed approach compared to an inversion-recovery reference method. In vivo, the proposed approach yielded 1.3 × 1.3 mm T1 maps with no significant difference (p = 0.77) in T1 and SDs in comparison to a cardiac-gated approach requiring 16 s scan time (i.e., seven times longer than the proposed approach). Cardiac motion correction improved the precision of T1 maps, shown by a 40% reduced SD. We have presented an approach that provides T1 maps of the myocardium in 2.3 s by utilizing both cardiac motion correction and model-based T1 reconstruction.

Quantitative magnetization transfer MRI unbiased by on-resonance saturation and dipolar order contributions.

To demonstrate the bias in quantitative MT (qMT) measures introduced by the presence of dipolar order and on-resonance saturation (ONRS) effects using magnetization transfer (MT) spoiled gradient-recalled (SPGR) acquisitions, and propose changes to the acquisition and analysis strategies to remove these biases. The proposed framework consists of SPGR sequences prepared with simultaneous dual-offset frequency-saturation pulses to cancel out dipolar order and associated relaxation (T1D ) effects in Z-spectrum acquisitions, and a matched quantitative MT (qMT) mathematical model that includes ONRS effects of readout pulses. Variable flip angle and MT data were fitted jointly to simultaneously estimate qMT parameters (macromolecular proton fraction [MPF], T2,f , T2,b , R, and free pool T1 ). This framework is compared with standard qMT and investigated in terms of reproducibility, and then further developed to follow a joint single-point qMT methodology for combined estimation of MPF and T1 . Bland-Altman analyses demonstrated a systematic underestimation of MPF (-2.5% and -1.3%, on average, in white and gray matter, respectively) and overestimation of T1 (47.1 ms and 38.6 ms, on average, in white and gray matter, respectively) if both ONRS and dipolar order effects are ignored. Reproducibility of the proposed framework is excellent (ΔMPF = -0.03% and ΔT1 = -19.0 ms). The single-point methodology yielded consistent MPF and T1 values with respective maximum relative average bias of -0.15% and -3.5 ms found in white matter. The influence of acquisition strategy and matched mathematical model with regard to ONRS and dipolar order effects in qMT-SPGR frameworks has been investigated. The proposed framework holds promise for improved accuracy with reproducibility.

Rapid 3D T1 mapping using deep learning-assisted Look-Locker inversion recovery MRI.

Conventional 3D Look-Locker inversion recovery (LLIR) T1 mapping requires multi-repetition data acquisition to reconstruct images at different inversion times for T1 fitting. To ensure B1 robustness, sufficient time of delay (TD) is needed between repetitions, which prolongs scan time. This work proposes a novel deep learning-assisted LLIR MRI approach for rapid 3D T1 mapping without TD. The proposed approach is based on the fact that , the effective T1 in LLIR imaging, is independent of TD and can be estimated from both LLIR imaging with and without TD, while accurate conversion of to T1 requires TD. Therefore, deep learning can be used to learn the conversion of to T1 , which eliminates the need for TD. This idea was implemented for inversion-recovery-prepared Golden-angel RAdial Sparse Parallel T1 mapping (GraspT1 ). 39 GraspT1 datasets with a TD of 6 s (GraspT1 -TD6) were used for training, which also incorporates additional anatomical images. The trained network was applied for T1 estimation in 14 GraspT1 datasets without TD (GraspT1 -TD0). The robustness of the trained network was also tested. Deep learning-based T1 estimation from GraspT1 -TD0 is accurate compared to the reference. Incorporation of additional anatomical images improves the accuracy of T1 estimation. The technique is also robust against slight variation in spatial resolution, imaging orientation and scanner platform. Our approach eliminates the need for TD in 3D LLIR imaging without affecting the T1 estimation accuracy. It represents a novel use of deep learning towards more efficient and robust 3D LLIR T1 mapping.

Open-source myocardial T1 mapping with simultaneous multi-slice acceleration: Combining an auto-calibrated blipped-bSSFP readout with VERSE-MB pulses.

Enabling fast and accessible myocardial T1 mapping is crucial for extending its clinical application. We introduce Open-MOLLI-SMS combining simultaneous multi-slice (SMS) with auto-calibration and variable-rate selective excitation (VERSE)-multiband pulses to obtain all slices in a fast single-shot T1 mapping sequence. Open-MOLLI-SMS was developed by integrating SMS with the open-source method Open-MOLLI previously implemented in Pulseq. Three methods were integrated for Open-MOLLI-SMS: (1) auto-calibration blip patterns to ensure consistency between the data and coil information; (2) a blipped-balanced SSFP (bSSFP) readout to induce controlled aliasing in parallel imaging shifts without disturbing the bSSFP frequency response; and (3) a VERSE-multiband pulse for minimizing the achievable TR and the specific absortion rate (SAR) impact of SMS. Two (SMS2) or three (SMS3) slices were excited simultaneously and encoded with an in-plane acceleration factor of 2. Experiments were performed in the International Society for Magnetic Resonance in Medicine/National Institute of Standards and Technology phantom and five healthy volunteers. Phantom results show accurate T1 estimates for reference values between 400 to 2200 ms. Artifacts were visible for Open-MOLLI-SMS3 but not replicated in vivo. In vivo Open-MOLLI-SMS (T1 SMS2 = 993 ± 10 ms; T1 SMS3 = 1031 ± 17 ms) provided similar values to mean T1 single-band Open-MOLLI estimates (T1 Open-MOLLI = 1005 ± 47 ms). Open-MOLLI-SMS2 provided the closest estimates to the reference. This proof-of-principle implementation study demonstrates the feasibility of speeding up T1 -mapping acquisitions and increasing coverage by combining auto-calibration strategies with a blipped-bSFFP readout and VERSE multiband RF excitation pulses. The proposed methodology was built on the Open-MOLLI mapping sequence, which provides a fast means for prototyping and enables open-source sharing of the method.

Combined estimation of B1 and T1 for dynamic contrast-enhanced MRI by accounting for incomplete spoiling of transverse magnetization

Objective. The variable flip angle (VFA) method for longitudinal relaxation time (T 1) measurement is inherently sensitive to inaccuracies in the radiofRequency transmit field (B 1) and incomplete spoiling of transverse magnetization. The objective of this study is to devise a computational method that addresses the problems of incomplete spoiling and B 1 inhomogeneity in the estimation of T 1 using VFA method. Approach. Using an analytical expression of the gradient echo signal with account of incomplete spoiling, we first showed that ill-posedness in the simultaneous estimation of B 1 and T 1 can be lifted with the use of flip angles larger than the Ernst angle. We then devised a nonlinear optimization method based on this signal model of incomplete spoiling for simultaneous estimation of B 1 and T 1. Main results. We evaluated the proposed method on a graded-concentration phantom to show that the derived T 1 estimates offers an improvement over the regular VFA method and compares well with reference values measured by inversion recovery. Reduction of the number of flip angles from 17 to 5 yielded consistent results indicating that the proposed method is numerically stable. T 1 estimates derived from in-vivo brain imaging were consistent with literature values for gray and white matter tissues. Significance. Contrary to the common notion that B 1 correction in the VFA method for T 1 mapping should be performed separately, we show that combined estimation of B 1 and T 1 is feasible by the proposed method simply with the acquisition of 5 flip angles, as demonstrated on both phantom and in-vivo imaging data.

Feasibility of joint mapping of triglyceride saturation and water longitudinal relaxation in a single breath hold applied to high fat-fraction adipose depots in the periclavicular anatomy

IntroductionSimultaneous mapping of triglyceride (TAG) saturation and tissue water relaxation may improve the characterization of the structure and function of anatomies with significant adipose tissue. While several groups have demonstrated in vivo TAG saturation imaging using MRI, joint mapping of relaxation and TAG saturation is understudied. Such mappings may avoid bias from physiological motion, if they can be done within a single breath-hold, and also account for static and applied magnetic field heterogeneity. MethodsWe propose a transient-state/MR fingerprinting single breath-hold sequence at 3 T, a low-rank reconstruction, and a parameter estimation pipeline that jointly estimates the number of double bonds (NDB), number of methylene interrupted double bonds (NMIDB), and tissue water T1, while accounting for non-ideal radiofrequency transmit scaling and off-resonance effects. We test the proposed method in simulations, in phantom against MR spectroscopy (MRS), and in vivo regions in and around high fat fraction (FF) periclavicular adipose tissue. Partial volume and multi-peak transverse relaxation effects are explored. ResultsThe simulation results demonstrate accurate NDB, NMIDB, and water T1 estimates across a range of NDB, NMIDB, and T1 values. In phantoms, the proposed method's estimates of NDB and NMIDB correlate with those from MR spectroscopy (Pearson correlation ≥0.98), while the water T1 estimates are concordant with a standard phantom. The NDB and NMIDB are sensitive to partial volumes of water, showing increasing bias at FF < 40%. This bias is found to be due to noise and transverse relaxation effects. The in vivo periclavicular adipose tissue has high FF (>90%). The adipose tissue NDB and NMIDB, and muscle T1 estimates are comparable to those reported in the literature. ConclusionRobust estimation of NDB, NMIDB at high FF and water T1 across a broad range of FFs are feasible using the proposed methods. Further reduction of noise and model bias are needed to employ the proposed technique in low FF anatomies and pathologies.

Confounder-corrected T1 mapping in the liver through simultaneous estimation of T1 , PDFF, , and in a single breath-hold acquisition.

Quantitative volumetric T1 mapping in the liver has the potential to aid in the detection, diagnosis, and quantification of liver fibrosis, inflammation, and spatially resolved liver function. However, accurate measurement of hepatic T1 is confounded by the presence of fat and inhomogeneous excitation. Furthermore, scan time constraints related to respiratory motion require tradeoffs of reduced volumetric coverage and/or increased acquisition time. This work presents a novel 3D acquisition and estimation method for confounder-corrected T1 measurement over the entire liver within a single breath-hold through simultaneous estimation of T1 , fat and . The proposed method combines chemical shift encoded MRI and variable flip angle MRI with a mapping technique to enable confounder-corrected T1 mapping. The method was evaluated theoretically and demonstrated in both phantom and in vivo acquisitions at 1.5 and 3.0T. At 1.5T, the method was evaluated both pre- and post- contrast enhancement in healthy volunteers. The proposed method demonstrated excellent linear agreement with reference inversion-recovery spin-echo based T1 in phantom acquisitions at both 1.5 and 3.0T, with minimal bias (5.2 and 45 ms, respectively) over T1 ranging from 200-1200 ms. In vivo results were in general agreement with reference saturation-recovery based 2D T1 maps (SMART1 Map, GE Healthcare). The proposed 3D T1 mapping method accounts for fat and confounders through simultaneous estimation of T1 , , PDFF and . It demonstrates strong linear agreement with reference T1 measurements, with low bias and high precision, and can achieve full liver coverage in a single breath-hold.

IMPROVING ACCURACY AND REPEATABILITY OF T2 MAPPING IN THE OAI DATA THROUGH EXTEND PHASE GRAPH MODELING

IMPROVING ACCURACY AND REPEATABILITY OF T2 MAPPING IN THE OAI DATA THROUGH EXTEND PHASE GRAPH MODELING

Optimisation of data acquisition towards continuous cardiac Magnetic Resonance Fingerprinting applications

PurposeAssess and optimise acquisition parameters for continuous cardiac Magnetic Resonance Fingerprinting (MRF). MethodsDifferent acquisition schemes (flip angle amplitude, lobe size, T2-preparation pulses) for cardiac MRF were assessed in simulations and phantom and demonstrated in one healthy volunteer. Three different experimental designs were evaluated using central composite and fractional factorial designs. Relative errors for T1 and T2 were calculated for a wide range of realistic T1 and T2 value combinations. The effect of different designs on the accuracy of T1 and T2 was assessed using response surface modelling and Cohen’s f calculations. ResultsLarger flip angle amplitudes lead to an improvement of T2 accuracy and precision for simulations and phantom experiments. Similar effects could also be shown qualitatively in in-vivo scans. Accuracy and precision of T1 were robust to different design parameters with improved values for faster flip angle variation. Cohen’s f showed that T2-preparation pulses influence the accuracy of T2. The number of pulses used is the most important parameter. Without T2-preparation pulses, RMSE were 3.0 ± 8.09 % for T1 and 16.24 ± 14.47 % for T2. Using those pulses reduced the RMSE to 2.3 ± 8.4 % for T1 and 14.11 ± 13.46 % for T2. Nonetheless, even if the improvement is significant, RMSE are still too high for reliable quantification. ConclusionIn contrast to previous study using triggered MRF sequences using < 30° flip angles, large flip angle amplitudes led to better results for continuous cardiac MRF sequences. T2-preparation pulse can improve the accuracy of T2 estimation but lead to longer scan times.

Liver T1 Relaxation Quantification Using a 3-Dimensional Interleaved Look-Locker Acquisition With T2 Preparation Pulse Sequence (3D-QALAS): Comparison With Conventional 2-Dimensional MOLLI.

Changes in liver magnetic resonance imaging T1 relaxation times are associated with histologic inflammation and fibrosis. To compare liver T1 measurements obtained using a novel single-breath-hold 3-dimensional (3D) whole-liver T1 estimation method (3D-QALAS) to standard-of-care 2-dimensional (2D) modified Look-Locker (2D-MOLLI) measurements. With institutional review board approval, research magnetic resonance imaging examinations were performed in 19 participants at 1.5 T. T1 relaxometry of the liver was performed using a novel 3D whole-liver T1 estimation method (3D-QALAS) as well as a 2D modified Look-Locker (2D-MOLLI) method. The 3D method covered the entire liver in a single breath hold, whereas 2D imaging was performed at 4 anatomic levels in 4 consecutive breath holds. T1 measurements from parametric maps were obtained by a single operator, and region-of-interest area-weighted mean T1 values were calculated. Pearson correlation ( r ) was used to assess correlation between T1 estimation methods, and the paired t test and Bland-Altman analysis were used to compare agreement in T1 measurements. In 18 participants (1 participant was excluded from analysis because of respiratory motion artifacts on 3D-QALAS images), 2D-MOLLI and 3D-QALAS mean T1 measurements were strongly correlated ( r = 0.95, [95% CI: 0.87-0.98]; P < 0.0001). 2D-MOLLI T1 values were significantly longer than 3D-QALAS values (647.2 ± 87.3 milliseconds vs. 554.7 ± 75.8 milliseconds; P < 0.0001) with mean bias = 92.5 milliseconds (95% limits of agreement, 36.8, 148.2 milliseconds). Whole-liver T1 measurements obtained using a novel single-breath-hold 3D T1 estimation method correlate with a standard-of-care multiple consecutive-breath-hold 2D single-slice method but demonstrate systematic bias that should be considered or corrected when used in a clinical or research setting.

Bayesian Estimation of Human Population Toxicokinetics of PFOA, PFOS, PFHxS, and PFNA from Studies of Contaminated Drinking Water.

Setting health-protective standards for poly- and perfluoroalkyl substances (PFAS) exposure requires estimates of their population toxicokinetics, but existing studies have reported widely varying PFAS half-lives (T½) and volumes of distribution (Vd). We combined data from multiple studies to develop harmonized estimates of T½ and Vd, along with their interindividual variability, for four PFAS commonly found in drinking water: perfluorooctanoic acid (PFOA), perfluorooctane sulfonate (PFOS), perfluorononanoic acid (PFNA), and perfluorohexane sulfonate (PFHxS). We identified published data on PFAS concentrations in human serum with corresponding drinking water measurements, separated into training and testing data sets. We fit training data sets to a one-compartment model incorporating interindividual variability, time-dependent drinking water concentrations, and background exposures. Use of a hierarchical Bayesian approach allowed us to incorporate informative priors at the population level, as well as at the study level. We compared posterior predictions to testing data sets to evaluate model performance. Posterior median (95% CI) estimates of T½ (in years) for the population geometric mean were 3.14 (2.69, 3.73) for PFOA, 3.36 (2.52, 4.42) for PFOS, 2.35 (1.65, 3.16) for PFNA, and 8.30 (5.38, 13.5) for PFHxS, all of which were within the range of previously published values. The extensive individual-level data for PFOA allowed accurate estimation of population variability, with a population geometric standard deviation of 1.57 (95% CI: 1.42, 1.73); data from other PFAS were also consistent with this degree of population variability. Vd estimates ranged from 0.19 to across the four PFAS, which tended to be slightly higher than previously published estimates. These results have direct application in both risk assessment (quantitative interspecies extrapolation and uncertainty factors for interindividual variability) and risk communication (interpretation of monitoring data). In addition, this study provides a rigorous methodology for further refinement with additional data, as well as application to other PFAS. https://doi.org/10.1289/EHP10103.

T1 Mapping MOLLI 5(3)3 Acquisition Scheme Yields High Accuracy in 1.5 T Cardiac Magnetic Resonance.

Objectives: To systematically compare two modified Look-Locker inversion recovery (MOLLI) T1 mapping sequences and their impact on (1) myocardial T1 values native, (2) post-contrast and (3) extracellular volume (ECV). Methods: 200 patients were prospectively included for 1.5 T CMR for work-up of ischemic or non-ischemic cardiomyopathies. To determine native and post-contrast T1 for ECV calculation, two different T1 mapping MOLLI acquisition schemes, 5(3)3 (designed for native scans with long T1) and 4(1)3(1)2 (designed for post-contrast scans with short T1), were acquired in identical mid-ventricular short-axis slices. Both schemes were acquired in native and post-contrast scans. Results: Datasets from 163 patients were evaluated (age 55 ± 17 years; 38% female). Myocardial T1 native for 5(3)3 was 1017 ± 42 ms vs. 956 ± 40 ms for 4(1)3(1)2, with mean intraindividual difference -61 ms (p &lt; 0.0001). Post-contrast myocardial T1 in patients was similar for both acquisition schemes, with 494 ± 48 ms for 5(3)3 and 490 ± 45 ms for 4(1)3(1)2 and mean intraindividual difference -4 ms. Myocardial ECV for 5(3)3 was 27.6 ± 4% vs. 27 ± 4% for 4(1)3(1)2, with mean difference -0.6 percentage points (p &lt; 0.0001). Conclusions: The T1 MOLLI 5(3)3 acquisition scheme provides a reliable estimation of myocardial T1 for the clinically relevant range of long and short T1 values native and post-contrast. In contrast, the T1 MOLLI 4(1)3(1)2 acquisition scheme may only be used for post-contrast scans according to its designed purpose.

A Comparison of MRI Quantitative Susceptibility Mapping and TRUST-Based Measures of Brain Venous Oxygen Saturation in Sickle Cell Anaemia.

In recent years, interest has grown in the potential for magnetic resonance imaging (MRI) measures of venous oxygen saturation (Yv) to improve neurological risk prediction. T2-relaxation-under-spin-tagging (TRUST) is an MRI technique which has revealed changes in Yv in patients with sickle cell anemia (SCA). However, prior studies comparing Yv in patients with SCA relative to healthy controls have reported opposing results depending on whether the calibration model, developed to convert blood T2 to Yv, is based on healthy human hemoglobin (HbA), bovine hemoglobin (HbBV) or sickle hemoglobin (HbS). MRI Quantitative Susceptibility Mapping (QSM) is an alternative technique that may hold promise for estimating Yv in SCA as blood magnetic susceptibility is linearly dependent upon Yv, and no significant difference has been found between the magnetic susceptibility of HbA and HbS. Therefore, the aim of this study was to compare estimates of Yv using QSM and TRUST with five published calibration models in healthy controls and patients with SCA. 17 patients with SCA and 13 healthy controls underwent MRI. Susceptibility maps were calculated from a multi-parametric mapping acquisition and Yv was calculated from the mean susceptibility in a region of interest in the superior sagittal sinus. TRUST estimates of T2, within a similar but much smaller region, were converted to Yv using five different calibration models. Correlation and Bland-Altman analyses were performed to compare estimates of Yv between TRUST and QSM methods. For each method, t-tests were also used to explore group-wise differences between patients with SCA and healthy controls. In healthy controls, significant correlations were observed between QSM and TRUST measures of Yv, while in SCA, there were no such correlations. The magnitude and direction of group-wise differences in Yv varied with method. The TRUST-HbBV and QSM methods suggested decreased Yv in SCA relative to healthy controls, while the TRUST-HbS (p < 0.01) and TRUST-HbA models suggested increased Yv in SCA as in previous studies. Further validation of all MRI measures of Yv, relative to ground truth measures such as O15 PET and jugular vein catheterization, is required in SCA before QSM or TRUST methods can be considered for neurological risk prediction.

Improving accuracy of myocardial T1 estimation in MyoMapNet.

To improve the accuracy and robustness of T1 estimation by MyoMapNet, a deep learning-based approach using 4 inversion-recovery T1 -weighted images for cardiac T1 mapping. MyoMapNet is a fully connected neural network for T1 estimation of an accelerated cardiac T1 mapping sequence, which collects 4 T1 -weighted images by a single Look-Locker inversion-recovery experiment (LL4). MyoMapNet was originally trained using in vivo data from the modified Look-Locker inversion recovery sequence, which resulted in significant bias and sensitivity to various confounders. This study sought to train MyoMapNet using signals generated from numerical simulations and phantom MR data under multiple simulated confounders. The trained model was then evaluated by phantom data scanned using new phantom vials that differed from those used for training. The performance of the new model was compared with modified Look-Locker inversion recovery sequence and saturation-recovery single-shot acquisition for measuring native and postcontrast T1 in 25 subjects. In the phantom study, T1 values measured by LL4 with MyoMapNet were highly correlated with reference values from the spin-echo sequence. Furthermore, the estimated T1 had excellent robustness to changes in flip angle and off-resonance. Native and postcontrast myocardium T1 at 3Tesla measured by saturation-recovery single-shot acquisition, modified Look-Locker inversion recovery sequence, and MyoMapNet were 1483 ± 46.6ms and 791 ± 45.8ms, 1169 ± 49.0ms and 612 ± 36.0ms, and 1443 ± 57.5ms and 700 ± 57.5ms, respectively. The corresponding extracellular volumes were 22.90% ± 3.20%, 28.88% ± 3.48%, and 30.65% ± 3.60%, respectively. Training MyoMapNet with numerical simulations and phantom data will improve the estimation of myocardial T1 values and increase its robustness to confounders while also reducing the overall T1 mapping estimation time to only 4 heartbeats.

Impact of deep learning architectures on accelerated cardiac T1 mapping using MyoMapNet.

The objective of the current study was to investigate the performance of various deep learning (DL) architectures for MyoMapNet, a DL model for T1 estimation using accelerated cardiac T1 mapping from four T1 -weighted images collected after a single inversion pulse (Look-Locker 4 [LL4]). We implemented and tested three DL architectures for MyoMapNet: (a) a fully connected neural network (FC), (b) convolutional neural networks (VGG19, ResNet50), and (c) encoder-decoder networks with skip connections (ResUNet, U-Net). Modified Look-Locker inversion recovery (MOLLI) images from 749 patients at 3T were used for training, validation, and testing. The first four T1 -weighted images from MOLLI5(3)3 and/or MOLLI4(1)3(1)2 protocols were extracted to create accelerated cardiac T1 mapping data. We also prospectively collected data from 28 subjects using MOLLI and LL4 to further evaluate model performance. Despite rigorous training, conventional VGG19 and ResNet50 models failed to produce anatomically correct T1 maps, and T1 values had significant errors. While ResUNet yielded good quality maps, it significantly underestimated T1 . Both FC and U-Net, however, yielded excellent image quality with good T1 accuracy for both native (FC/U-Net/MOLLI =1217 ± 64/1208 ± 61/1199 ± 61 ms, all p< 0.05) and postcontrast myocardial T1 (FC/U-Net/MOLLI = 578 ± 57/567 ± 54/574 ± 55 ms, all p< 0.05). In terms of precision, the U-Net model yielded better T1 precision compared with the FC architecture (standard deviation of 61 vs. 67 ms for the myocardium for native [p< 0.05], and 31 vs. 38 ms [p< 0.05], for postcontrast). Similar findings were observed in prospectively collected LL4 data. It was concluded that U-Net and FC DL models in MyoMapNet enable fast myocardial T1 mapping using only four T1 -weighted images collected from a single LL sequence with comparable accuracy. U-Net also provides a slight improvement in precision.

Single-shot T2 mapping via multi-echo-train multiple overlapping-echo detachment planar imaging and multitask deep learning.

Quantitative magnetic resonance imaging provides robust biomarkers in clinics. Nevertheless, the lengthy scan time reduces imaging throughput and increases the susceptibility of imaging results to motion. In this context, a single-shot T2 mapping method based on multiple overlapping-echo detachment (MOLED) planar imaging was presented, but the relatively small echo time range limits its accuracy, especially in tissues with large T2 . In this work we proposed a novel single-shot method, Multi-Echo-Train Multiple OverLapping-Echo Detachment (METMOLED) planar imaging, to accommodate a large range of T2 quantification without additional measurements to rectify signal degeneration arisen from refocusing pulse imperfection. Multiple echo-train techniques were integrated into the MOLED sequence to capture larger TE information. Maps of T2 , B1 , and spin density were reconstructed synchronously from acquired METMOLED data via multitask deep learning. A typical U-Net was trained with 3000/600 synthetic data with geometric/brain patterns to learn the mapping relationship between METMOLED signals and quantitative maps. The refocusing pulse imperfection was settled through the inherent information of METMOLED data and auxiliary tasks. Experimental results on the digital brain (structural similarity (SSIM) index=0.975/0.991/0.988 for MOLED/METMOLED-2/METMOLED-3, hyphenated number denotes the number of echo-trains), physical phantom (the slope of linear fitting with reference T2 map=1.047/1.017/1.006 for MOLED/METMOLED-2/METMOLED-3), and human brain (Pearson's correlation coefficient (PCC)=0.9581/0.9760/0.9900 for MOLED/METMOLED-2/METMOLED-3) demonstrated that the METMOLED improved the quantitative accuracy and the tissue details in contrast to the MOLED. These improvements were more pronounced in tissues with large T2 and in application scenarios with high temporal resolution (PCC=0.8692/0.9465/0.9743 for MOLED/METMOLED-2/METMOLED-3). Moreover, the METMOLED could rectify the signal deviations induced by the non-ideal slice profiles of refocusing pulses without additional measurements. A preliminary measurement also demonstrated that the METMOLED is highly repeatable (mean coefficient of variation (CV)=1.65%). METMOLED breaks the restriction of echo-train length to TE and implements unbiased T2 estimates in an extensive range. Furthermore, it corrects the effect of refocusing pulse inaccuracy without additional measurements or signal post-processing, thus retaining its single-shot characteristic. This technique would be beneficial for accurate T2 quantification.

Deep Learning Staging of Liver Iron Content From Multiecho MR Images.

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. To develop a deep-learning method for unsupervised classification of LIC from magnitude T2* multiecho MR images. Retrospective. 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. A5 T, T2* multiecho magnitude images. 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. 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. 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. The proposed HippoNet-Ensemble network can perform unsupervised LIC staging and achieves good prognostic performance. 4 TECHNICAL EFFICACY: Stage 2.