Magnetic Resonance in Medicine最新文献

Purpose: Using artificial intelligence neural networks to generate a representation that maps the input directly to neurochemical concentrations and metabolite-level average transverse relaxation times (T₂).

Methods: The proposed model used time-domain JPRESS data as input and was trained to be invariant to phase shifts, frequency offsets, and lineshape variations, using computer-synthesized data without prior knowledge of in vivo metabolite concentration distributions. TE-specific representations were generated using a combination of WaveNet and gated recurrent units (GRUs) and integrated into a unified JPRESS representation.

Results: By focusing solely on target metabolite signals, the model effectively filtered out background signals, including spectral artifacts and unregistered metabolites. The predicted concentrations and metabolite-level average T₂ values were consistent with those reported in the literature. The model demonstrated robustness to phase shifts, frequency offsets, and line broadening. Additionally, it was capable of detecting low-concentration neurochemicals, such as gamma-aminobutyric acid (GABA), without spectral editing.

Conclusion: This study demonstrates that deep learning can be used for automatically quantifying both metabolite concentrations and transverse relaxation times with high practical viability.

Purpose: Blood T₁ is a key parameter for hemodynamic quantification in both non-contrast- and contrast-enhanced imaging. Individual vessel T₁ has been measured using a modified Look-Locker scheme with multi-shot EPI or FLASH in high spatial resolution, requiring ∼1 min. Here, by exploiting the temporal sparsity from the excessive number of inversion delays, we apply Golden Angle rotated Spiral k-t Sparse Parallel imaging (GASSP) to enable blood T₁ measurement in a single shot of 10 s.

Methods: The pulse sequence with single-shot GASSP reconstruction was developed for T₁ measurement from the internal jugular vein (IJV) with 1 × 1 mm² in-plane resolution. On nine healthy volunteers, the single-shot GASSP was compared to the segmented EPI readout and was repeated to assess its intra-scan reproducibility. Another experiment was performed on three patients, during which the 10 s GASSP was obtained at different time points prior to and following the Gadolinium (Gd) administration to assess dynamic changes in blood T₁.

Results: The blood T₁ values measured with the highly undersampled GASSP method were strongly correlated (r = 0.83) with those using the multi-shot EPI readout and exhibited high reproducibility (r = 0.88) within the session. The baseline IJV T₁ values measured were 1700-2000 ms. Following the Gd injection, the T₁ values of IJVs gradually recovered from ∼300-400 to ∼500 ms within 10-15 min.

Conclusion: The feasibility of an ultrafast blood T₁ measurement was demonstrated with high spatial resolution in a single shot of 10 s, applicable to both pre- and post-contrast conditions.

Purpose: To improve cardiac motion representation and reduce artifacts for cardiac- and respiratory-resolved imaging through a synergistic combination of retrospective cardiac phased array RF focusing and rigid respiratory motion compensation (MoCo).

Methods: We incorporated cardiac receive focusing using region-optimized virtual coils (ROVir) and MoCo into cardiac- and respiratory-resolved low-rank tensor (LRT) reconstruction, hypothesizing that the combination of MoCo + ROVir would prioritize the LRT representation of cardiac motion over respiratory motion. We compared LRT, MoCo-LRT, ROVir-LRT, and the proposed MoCo + ROVir-LRT reconstructions of retrospective data from N = 24 pediatric patients with congenital heart disease (CHD) scanned at 3.0 T using ROCK-MUSIC. Technical evaluation metrics included the proportion of cardiac-to-respiratory motion energy in self-gating lines, cardiac motion priority in the temporal basis, flickering energy, and edge sharpness in end-expiratory cardiac cine. Reconstructed cardiac cines were scored by two expert image readers.

Results: MoCo + ROVir significantly increased the proportion of cardiac-to-respiratory motion energy in self-gating lines (p < 0.001) and prioritized cardiac motion in the temporal basis (p < 0.001). MoCo + ROVir reduced flickering energy in cardiac cine images (p < 0.001), sharpened the liver-lung interface (p < 0.001), and improved flickering-specific scores (p = 0.001). Myocardium-blood pool interface sharpness (p = 0.831), cardiac-specific image scores (p = 0.188), and vascular-specific scores (p = 0.901) were not significantly different. Together, these two techniques allowed 3.7-5.2× faster reconstruction times versus LRT-only.

Conclusion: The synergy of MoCo + ROVir successfully prioritized cardiac motion, suppressed respiratory motion, and reduced flickering artifacts, with an added benefit of accelerating reconstruction times. The improved respiratory motion handling may provide an avenue for free-breathing cardiac scans in pediatric patients with CHD.

Purpose: This study aims to develop and evaluate a fully automated deep learning-driven postprocessing pipeline for multiparametric renal MRI, enabling accurate kidney alignment, segmentation, and quantitative feature extraction within a single efficient workflow.

Methods: Our method has three main stages. First, a segmentation network delineates renal structures in high-contrast images. Next, a deep learning-based pairwise image registration algorithm maps the multiparametric image series to a common target and transfers the predicted annotations between the multiparametric images. Finally, clinically relevant quantitative parameters are extracted through region-specific assessment of renal structure and function based on the aligned and segmented multiparametric data. We used five-fold cross-validation to compare the segmentation outcomes and extracted features with manual analyses in 24 patients with prostate cancer or neuroendocrine tumors and 10 healthy subjects, each undergoing repeated scans.

Results: Our automated pipeline achieved high agreement with expert kidney segmentation while delivering significant alignment improvements through registration. Volumetric analysis showed a strong correlation (r $>$ 0.9) with manual results, and feature extraction demonstrated high intraclass correlation coefficients with minimal bias. The complete processing pipeline, encompassing coregistration, segmentation, and feature extraction, required approximately 15 s per scan from raw input to final quantitative output.

Conclusion: The study establishes a reliable automated pipeline for renal multiparametric MRI postprocessing. The achieved accuracy and efficiency can support improved diagnosis and treatment planning for patients with kidney disease.

Purpose: The human brain contains multiple fluid types, including blood, cerebrospinal fluid (CSF), and tissue water. While intravoxel incoherent motion (IVIM) imaging has been used to examine microvascular perfusion, evidence on incoherent flows of CSF is emerging. This study aims to develop in vivo multidimensional MRI methods to investigate potential contributions of CSF in the IVIM regime.

Method: T₁-Diffusion (T₁-D) and T₂-Diffusion (T₂-D) MRI data were acquired from 10 healthy subjects to investigate the relaxivity and diffusion signatures of incoherent fluid flows in the brain. Based on the T₁-D and T₂-D results, T₁/T₂ selective IVIM protocols were developed to map incoherent CSF flows in the human brains.

Results: T₁-D and T₂-D MRI detected incoherent CSF flow in the brain subarachnoid space. Results from four different relaxation selective IVIM methods further support incoherent CSF flows in these regions.

Conclusion: We have shown the feasibility of using T₁-D and T₂-D MRI within the low b-value regime to probe the heterogeneity of IVIM flow components. Designed based on the 2D MRI spectra, relaxation selective 1D IVIM acquisition can be obtained within clinically feasible time frame.

Purpose: The goal of this study was to develop a 5-min 3D MRA acquisition at 0.55 T with predictable scan time, 100% data efficiency, and robust water-fat separation.

Methods: For full data efficiency, the proposed method combined self-gating with retrospective motion correction while ensuring a predictable 5-min scan time. Water-fat separation was implemented using a model-based Dixon reconstruction. Evaluation in 18 volunteers compared results to navigator-gated reference scans with nominal scan times of 5 and 10 min via a Likert scale blinded expert rating. Susceptibility to irregular breathing patterns was also analyzed.

Results: The expert rating for image quality was 4.22 for the proposed method, 3.89 for the 5-min navigator-gated scan and 4.43 for the 10-min navigator-gated scan. Ranking the three methods revealed moderate inter-rater reliability of 0.46, suggesting only minor differences. While navigator-gated acquisitions deviated from the expected scan time by -2.26 to 2.86 min and -3.91 to 4.54 min for the 5- and 10-min protocols respectively, the proposed method deviated only by -0.17 to 0.45 min. The self-gated method further avoided saturation artifacts from the cross-beam navigator, allowing better distinction of the right pulmonary veins. Image quality for the proposed method was also less susceptible to irregular breathing patterns.

Conclusion: Whole-thorax MRA acquisitions with water-fat separation and predictable scan times were successfully acquired in 18 volunteers at 0.55 T. The proposed method demonstrated on average better image quality than navigator-gated acquisitions of the same nominal scan time while mitigating limitations of prospective navigator gating.

Purpose: To develop a robust method for estimating myelin water fraction (MWF) from multi-echo gradient-recalled echo (mGRE) data under acquisition regimes that limit echo-train length and support higher spatial sampling.

Methods: A tensor decomposition-based multi-signal matrix pencil (T-MP) framework is proposed to incorporate data-driven spatial information from neighboring voxels into MWF estimation. By leveraging the reduced temporal sampling requirements of matrix pencil-based approaches, the method enables stable parameter estimation with fewer echoes compared to conventional iterative fitting techniques. The performance of the proposed method was evaluated using numerical simulations across a range of signal-to-noise ratios and echo spacings, as well as in vivo mGRE datasets acquired at different spatial resolutions with shortened echo trains.

Results: Numerical simulations demonstrate that accurate MWF estimation can be achieved with substantially fewer temporal samples, facilitating acquisition protocols that prioritize spatial encoding. In vivo experiments show that the proposed method provides consistent MWF maps across different spatial resolutions without qualitative degradation. Kernel density analysis reveals improved estimation consistency in both white and gray matter compared with conventional voxel-wise fitting approaches. In addition, the proposed framework substantially reduces per-slice computation time.

Conclusion: A tensor decomposition-based multi-signal matrix pencil method for MWF estimation is presented that integrates spatially informed signal structure while reducing temporal sampling requirements. The proposed framework supports spatially efficient mGRE acquisitions and provides improved robustness and computational efficiency compared to existing voxel-wise approaches.

Purpose: A method for simultaneous mapping of static (B₀) and transmit (B₁ ⁺) field inhomogeneities at ultra-high field (UHF) was developed and validated. The utility of accelerating the proposed sequence using deep learning (DL) and joint low-rank tensor completion (TxLR) reconstruction methods was evaluated to enable rapid online implementation.

Methods: A 3D sequence, Combined caLculation of UHF Biases (CLUB)-Sandwich, was developed by incorporating a multi-echo readout into the unsaturated segment of the Sandwich B₁ ⁺ mapping sequence, enabling simultaneous B₀ estimation. Data from 11 healthy volunteers were acquired at 7 T. Estimated ΔB₀ and B₁ ⁺ maps were compared with established, separate reference scans. Retrospectively and prospectively undersampled data were reconstructed using TxLR and a DL-based algorithm. The resulting maps were compared with fully sampled data.

Results: CLUB-Sandwich maps showed strong agreement with reference methods. A strong correlation (r > 0.97) and low mean volumetric root mean squared errors were found for both ΔB₀ (9.5 ± 1.8 Hz) and absolute B₁ ⁺ (3.5° ± 0.3°). Both reconstruction methods enabled acquisitions in under 10 s of acquisition time. DL reconstruction was found to be substantially faster (5 s) than the TxLR algorithm (4 min) while producing comparable map quality. Prospective validation confirmed the feasibility of online mapping with acceptable accuracy.

Conclusion: The CLUB-Sandwich method was developed for fast, accurate, and simultaneous ΔB₀ and B₁ ⁺ mapping. When combined with a DL-based reconstruction, the proposed framework provides maps in under 10 s of acquisition time, presenting a feasible solution for rapid online inhomogeneity estimation in UHF applications.

Purpose: 3D free-breathing, proton, contrast-agent-free MR methods are increasingly used for pulmonary ventilation-weighted measurements. The methods are split between: (1) signal-based, which rely on lung parenchyma signal changes during respiration, and (2) volume-based that utilize the Jacobian determinant of deformation fields from the image registration. This study compares both proton methods using respiratory-resolved images acquired using fermat-looped orthogonally encoded trajectories (FLORET) acquisition.

Methods: Free-breathing FLORET data were acquired from participants with various pulmonary conditions (N = 29) and healthy controls (N = 7), and reconstructed into respiratory phase-resolved images. Signal-based regional ventilation (RVent) was quantified using the 3D phase-resolved functional lung algorithm, and volume-based Jacobian ventilation (JVent) was derived as the Jacobian of the deformation field from the direct image registration of the end-expiratory image to the end-inspiratory image. Differences between the means, coefficients of variation (CoVs), and their ventilation defect percent (VDP) were quantified by Bland-Altman plots. The spatial overlap of the defect maps was determined by multi-class Sørensen-Dice coefficient, and Spearman correlations to ¹²⁹Xe MRI were assessed.

Results: In all study participants, statistically significant differences were found between means/CoVs of RVent and JVent parameters (both p < 0.0001), but not VDP (p = 0.38). The median spatial overlap of the defect maps was 86%. VDP_RVent showed stronger correlation (ρ = 0.78, Meng Z = 4.36, p < 0.0001) to VDP_129Xe than JVent (ρ = 0.34).

Conclusion: Although both proton lung MRI methods successfully identified ventilation defects, the stronger correlation between signal-based and ¹²⁹Xe MRI indicates that RVent may provide a more reliable assessment of lung ventilation in clinical applications in comparison to volume-based parameters.