Whole-heart chemical shift encoded water-fat MRI

Accelerated chemical shift encoded cardiovascular magnetic resonance imaging with use of a resolution enhancement network

BACKGROUND

Cardiovascular magnetic resonance (CMR) chemical shift encoding (CSE) enables myocardial fat imaging. We sought to develop a deep learning network (fast chemical shift encoding [FastCSE]) to accelerate CSE.

METHODS

FastCSE was built on a super-resolution generative adversarial network extended to enhance complex-valued image sharpness. FastCSE enhances each echo image independently before water-fat separation. FastCSE was trained with retrospectively identified cines from 1519 patients (56 ± 16 years; 866 men) referred for clinical 3T CMR. In a prospective study of 16 participants (58 ± 19 years; 7 females) and 5 healthy individuals (32 ± 17 years; 5 females), dual-echo CSE images were collected with 1.5 × 1.5 mm2, 2.5 × 1.5 mm2, and 3.8 × 1.9 mm2 resolution using generalized autocalibrating partially parallel acquisition (GRAPPA). FastCSE was applied to images collected with resolution of 2.5 × 1.5 mm2 and 3.8 × 1.9 mm2 to restore sharpness. Fat images obtained from two-point Dixon reconstruction were evaluated using a quantitative blur metric and analyzed with a five-way analysis of variance.

RESULTS

FastCSE successfully reconstructed CSE images inline. FastCSE acquisition, with a resolution of 2.5 × 1.5 mm2 and 3.8 × 1.9 mm2, reduced the number of breath-holds without impacting visualization of fat by approximately 1.5-fold and 3-fold compared to GRAPPA acquisition with a resolution of 1.5 × 1.5 mm2, from 3.0 ± 0.8 breath-holds to 2.0 ± 0.2 and 1.1 ± 0.4 breath-holds, respectively. FastCSE improved image sharpness and removed ringing artifacts in GRAPPA fat images acquired with a resolution of 2.5 × 1.5 mm2 (0.32 ± 0.03 vs 0.35 ± 0.04, P < 0.001) and 3.8 × 1.9 mm2 (0.32 ± 0.03 vs 0.43 ± 0.06, P < 0.001). Blurring in FastCSE images was similar to blurring in images with 1.5 × 1.5 mm2 resolution (0.32 ± 0.03 vs 0.31 ± 0.03, P = 0.57; 0.32 ± 0.03 vs 0.31 ± 0.03, P = 0.66).

CONCLUSION

We showed that a deep learning-accelerated CSE technique based on complex-valued resolution enhancement can reduce the number of breath-holds in CSE imaging without impacting the visualization of fat. FastCSE showed similar image sharpness compared to a standardized parallel imaging method.

Trajectory correction enables free-running chemical shift encoded imaging for accurate cardiac proton-density fat fraction quantification at 3T

BACKGROUND

Metabolic diseases can negatively alter epicardial fat accumulation and composition, which can be probed using quantitative cardiac chemical shift encoded (CSE) cardiovascular magnetic resonance (CMR) by mapping proton-density fat fraction (PDFF). To obtain motion-resolved high-resolution PDFF maps, we proposed a free-running cardiac CSE-CMR framework at 3T. To employ faster bipolar readout gradients, a correction for gradient imperfections was added using the gradient impulse response function (GIRF) and evaluated on intermediate images and PDFF quantification.

METHODS

Ten minutes free-running cardiac 3D radial CSE-CMR acquisitions were compared in vitro and in vivo at 3T. Monopolar and bipolar readout gradient schemes provided 8 echoes (TE1/ΔTE = 1.16/1.96 ms) and 13 echoes (TE1/ΔTE = 1.12/1.07 ms), respectively. Bipolar-gradient free-running cardiac fat and water images and PDFF maps were reconstructed with or without GIRF correction. PDFF values were evaluated in silico, in vitro on a fat/water phantom, and in vivo in 10 healthy volunteers and 3 diabetic patients.

RESULTS

In monopolar mode, fat-water swaps were demonstrated in silico and confirmed in vitro. Using bipolar readout gradients, PDFF quantification was reliable and accurate with GIRF correction with a mean bias of 0.03% in silico and 0.36% in vitro while it suffered from artifacts without correction, leading to a PDFF bias of 4.9% in vitro and swaps in vivo. Using bipolar readout gradients, in vivo PDFF of epicardial adipose tissue was significantly lower compared to subcutaneous fat (80.4 ± 7.1% vs 92.5 ± 4.3%, P < 0.0001).

CONCLUSIONS

Aiming for an accurate PDFF quantification, high-resolution free-running cardiac CSE-MRI imaging proved to benefit from bipolar echoes with k-space trajectory correction at 3T. This free-breathing acquisition framework enables to investigate epicardial adipose tissue PDFF in metabolic diseases.

Motion-compensated fat-water imaging for 3D cardiac MRI at ultra-high fields

PURPOSE

Respiratory motion-compensated (MC) 3D cardiac fat-water imaging at 7T.

METHODS

Free-breathing bipolar 3D triple-echo gradient-recalled-echo (GRE) data with radial phase-encoding (RPE) trajectory were acquired in 11 healthy volunteers (7M\4F, 21-35 years, mean: 30 years) with a wide range of body mass index (BMI; 19.9-34.0 kg/m² ) and volunteer tailored B 1 + shimming. The bipolar-corrected triple-echo GRE-RPE data were binned into different respiratory phases (self-navigation) and were used for the estimation of non-rigid motion vector fields (MF) and respiratory resolved (RR) maps of the main magnetic field deviations (ΔB₀ ). RR ΔB₀ maps and MC ΔB₀ maps were compared to a reference respiratory phase to assess respiration-induced changes. Subsequently, cardiac binned fat-water images were obtained using a model-based, respiratory motion-corrected image reconstruction.

RESULTS

The 3D cardiac fat-water imaging at 7T was successfully demonstrated. Local respiration-induced frequency shifts in MC ΔB₀ maps are small compared to the chemical shifts used in the multi-peak model. Compared to the reference exhale ΔB₀ map these changes are in the order of 10 Hz on average. Cardiac binned MC fat-water reconstruction reduced respiration induced blurring in the fat-water images, and flow artifacts are reduced in the end-diastolic fat-water separated images.

CONCLUSION

This work demonstrates the feasibility of 3D fat-water imaging at UHF for the entire human heart despite spatial and temporal B 1 + and B₀ variations, as well as respiratory and cardiac motion.

Evaluation of a motion-robust 2D chemical shift-encoded technique for R2* and field map quantification in ferumoxytol-enhanced MRI of the placenta in pregnant rhesus macaques

BACKGROUND

3D chemical shift-encoded (CSE)-MRI techniques enable assessment of ferumoxytol concentration but are unreliable in the presence of motion.

PURPOSE

To evaluate a motion-robust 2D-sequential CSE-MRI for R2* and B0 mapping in ferumoxytol-enhanced MRI of the placenta.

STUDY TYPE

Prospective.

ANIMAL MODEL

Pregnant rhesus macaques.

FIELD STRENGTH/SEQUENCE

3.0T/CSE-MRI.

ASSESSMENT

2D-sequential CSE-MRI was compared with 3D respiratory-gated CSE-MRI in placental imaging of 11 anesthetized animals at multiple timepoints before and after ferumoxytol administration, and in ferumoxytol phantoms (0 μg/mL-440 μg/mL). Motion artifacts of CSE-MRI in 10 pregnant women without ferumoxytol administration were assessed retrospectively by three blinded readers (4-point Likert scale). The repeatability of CSE-MRI in seven pregnant women was also prospectively studied.

STATISTICAL TESTS

Placental R2* and boundary B0 field measurements (ΔB0) were compared between 2D-sequential and 3D respiratory-gated CSE-MRI using linear regression and Bland-Altman analysis.

RESULTS

In phantoms, a slope of 0.94 (r² = 0.99, concordance correlation coefficient ρ = 0.99), and bias of -4.8 s^-1 (limit of agreement [LOA], -41.4 s^-1 , +31.8 s^-1 ) in R2*, and a slope of 1.07 (r² = 1.00, ρ = 0.99) and bias of 11.4 Hz (LOA -12.0 Hz, +34.8 Hz) in ΔB0 were obtained in 2D CSE-MRI compared with 3D CSE-MRI for reference R2* ≤390 s^-1 . In animals, a slope of 0.92 (r² = 0.97, ρ = 0.98) and bias of -2.2 s^-1 (LOA -55.6 s^-1 , +51.3 s^-1 ) in R2*, and a slope of 1.05 (r² = 0.95, ρ = 0.97) and bias of 0.4 Hz (LOA -9.0 Hz, +9.7 Hz) in ΔB0 were obtained. In humans, motion-impaired R2* maps in 3D CSE-MRI (Reader 1: 1.8 ± 0.6, Reader 2: 1.3 ± 0.7, Reader 3: 1.9 ± 0.6), while 2D CSE-MRI was motion-free (Reader 1: 2.9 ± 0.3, Reader 2: 3.0 ± 0, Reader 3: 3.0 ± 0). A mean difference of 0.66 s^-1 and coefficient of repeatability of 9.48 s^-1 for placental R2* were observed in the repeated 2D CSE-MRI.

DATA CONCLUSION

2D-sequential CSE-MRI provides accurate R2* and B0 measurements in ferumoxytol-enhanced placental MRI of animals in the presence of respiratory motion, and motion-robustness in human placental imaging.

LEVEL OF EVIDENCE

1 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2020;51:580-592.

Method of B0 mapping with magnitude-based correction for bipolar two-point Dixon cardiac MRI

PURPOSE

The conventional two-point (2pt) Dixon technique explicitly estimates B0 map by performing phase unwrapping. When signal loss, phase singularity, artifacts, or spatially isolated regions corrupt the measured phase images, this unwrapping-based technique will face difficulty. This work aims to improve the reliability of B0 mapping by performing unwrapping error correction.

METHOD

To detect the unwrapping-caused phase errors, we determined a magnitude-based fat/water mask and used it as reference to identify pixels being mismatched by the phase-based mask, which was derived from the B0-corrected phase term of the Hermitian product between echoes. Then, we corrected the afore-determined phase error on a region-by-region basis. We tested the developed method with nine patients' data, and the results were compared with a well-established region-growing technique.

RESULTS

By adding the step to correct unwrapping-caused error, we improved the robustness of B0 mapping, resulting in better fat-water separation when compared with the conventional 2pt and the phasor-based region-growing techniques.

CONCLUSION

We showed the feasibility of B0 mapping with bipolar 2pt human cardiac data. The software is freely available to the scientific community. Magn Reson Med 78:1862-1869, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Utilization of a balanced steady state free precession signal model for improved fat/water decomposition

PURPOSE

Chemical shift based fat/water decomposition methods such as IDEAL are frequently used in challenging imaging environments with large B0 inhomogeneity. However, they do not account for the signal modulations introduced by a balanced steady state free precession (bSSFP) acquisition. Here we demonstrate improved performance when the bSSFP frequency response is properly incorporated into the multipeak spectral fat model used in the decomposition process.

THEORY AND METHODS

Balanced SSFP allows for rapid imaging but also introduces a characteristic frequency response featuring periodic nulls and pass bands. Fat spectral components in adjacent pass bands will experience bulk phase offsets and magnitude modulations that change the expected constructive and destructive interference between the fat spectral components. A bSSFP signal model was incorporated into the fat/water decomposition process and used to generate images of a fat phantom, and bilateral breast and knee images in four normal volunteers at 1.5 Tesla.

RESULTS

Incorporation of the bSSFP signal model into the decomposition process improved the performance of the fat/water decomposition.

CONCLUSION

Incorporation of this model allows rapid bSSFP imaging sequences to use robust fat/water decomposition methods such as IDEAL. While only one set of imaging parameters were presented, the method is compatible with any field strength or repetition time.

Improving chemical shift encoded water-fat separation using object-based information of the magnetic field inhomogeneity

PURPOSE

The purpose of this work was to improve the robustness of existing chemical shift encoded water-fat separation methods by incorporating object-based information of the B0 field inhomogeneity.

THEORY

The primary challenge in water-fat separation is the estimation of phase shifts that arise from B0 field inhomogeneity, which is composed of the background field and susceptibility-induced field. The susceptibility-induced field can be estimated if the susceptibility distribution is known or can be approximated. In this work, the susceptibility distribution is approximated from the source images using the known susceptibility values of water, fat, and air. The field estimate is then demodulated from the source images before water-fat separation.

METHODS

Chemical shift encoded source images were acquired in anatomical regions that are prone to water-fat swaps. The images were processed using algorithms from the ISMRM Fat-Water Toolbox, with and without the object-based field map information. The estimates were compared to examine the benefit of using the object-based field map information.

RESULTS

Multiple cases are shown in which water-fat swaps were avoided by using the object-based information of the B0 field map.

CONCLUSION

Object-based information of the B0 field may improve the robustness of existing chemical shift encoded water-fat separation methods.